Water conservancy project patrol management and protection method and system
By constructing a safety hazard prediction model and a GIS-BIM three-dimensional spatial information model, and combining video surveillance data and A* algorithm to optimize inspection routes, the problem of insufficient data timeliness and relevance in traditional water conservancy project inspection and maintenance has been solved, achieving efficient and accurate inspection and hazard management.
Patent Information
- Application Number
- CN202610127707.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-29
- Publication Date
- 2026-05-12
AI Technical Summary
Traditional water conservancy project inspection and maintenance relies on regular manual inspections, which lack precise spatial feature descriptions and data timeliness, resulting in low inspection accuracy and efficiency, failure to detect potential safety hazards in a timely manner, and lack of specificity, thus increasing manpower and material costs.
By collecting data on shoreline, water area, and reservoir surrounding environment, a safety hazard prediction model and a GIS-BIM three-dimensional spatial information model are constructed. Combined with video surveillance data, real-time updates and simulated inspections are performed. The A* algorithm is used to optimize the inspection route, achieving effective fusion and comprehensive analysis of multi-source data.
It improves the accuracy and efficiency of inspections, enables the timely detection and elimination of potential hazards, reduces costs, and enhances the level of safety management.
Smart Images
Figure CN122022099A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of information technology for water conservancy projects, specifically a method and system for the inspection and maintenance of water conservancy projects. Background Technology
[0002] As a crucial national infrastructure, the safe and stable operation of water conservancy projects directly impacts public interests such as flood control, disaster reduction, and water supply security. Strengthening daily inspection and maintenance of water conservancy projects is a key aspect of ensuring their long-term safe operation. Chinese Patent CN117636357A discloses a big data-based water conservancy project information management system, belonging to the field of management system technology. It includes: a data acquisition module for acquiring image data of paper documents; a processing module for standardizing the acquired image data; an extraction module for extracting features from the image data; a detection module for detecting surface contamination on paper documents; a display module for displaying the detection results; and a storage module for storing the data. This system can accurately detect surface contamination on paper documents by utilizing surface features, and can learn the characteristic representation of paper contamination from large amounts of data, thereby achieving automated contamination detection and improving the quality of information management.
[0003] Traditional water conservancy project inspection and maintenance mainly rely on regular manual inspections and manual records, which has the following main problems: Failure to construct a variogram or proper use of models such as the spherical model will prevent the effective quantification of spatial correlation and variability between different data points; this will lead to vague and imprecise descriptions of spatial features, affecting the accuracy of subsequent analysis and applications; the lack of precise spatial feature descriptions will introduce more errors when constructing GIS-BIM 3D spatial information models, affecting the reliability and application value of the models; failure to constrain weight coefficients by adjusting the model through weight coefficients will prevent observation points with stronger spatial correlation to the target point from receiving greater weights during the estimation process; this will cause interpolation results to deviate from reality, reducing their accuracy. The failure to consider using GIS-BIM 3D spatial information model update formulas in conjunction with real-time monitoring and sensor data for dynamic updates will result in the inability of the model's state and environmental information to reflect actual changes in a timely manner, reducing data timeliness. The failure to consider incorporating virtual inspection technology to simulate inspections will prevent the early detection of potential problems and difficulties during inspections, reducing the accuracy and effectiveness of inspections. Furthermore, the reduced timeliness and accuracy of data, decreased inspection efficiency and accuracy, and insufficient ability to identify and assess potential risks will lead to incomplete or distorted information required for decision-making. Failure to categorize hazard types and levels will result in a lack of focus in inspection work. Inspectors may be unable to accurately identify areas and inspection points that require special attention, leading to blind inspections. Non-optimal inspection paths will increase the manpower, material resources, and time costs of the inspection process, reducing overall inspection efficiency. The inability to promptly detect and eliminate potential safety hazards will lower the overall level of safety management and pose a threat to personnel safety.
[0004] Therefore, in view of the above situation, there is an urgent need to provide a method and system for the inspection and maintenance of water conservancy projects to overcome the shortcomings in current practical applications. Summary of the Invention
[0005] The purpose of this invention is to provide a method and system for the inspection and maintenance of water conservancy projects, which effectively solves the problems mentioned in the background art.
[0006] This invention is implemented as follows: a method for the inspection and maintenance of water conservancy projects, which includes the following steps: S1. Collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data; S2. Preprocess the shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. Then, perform weighted fusion on the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain the comprehensive feature dataset. S3. Construct a safety hazard prediction model, input the comprehensive feature dataset into the safety hazard prediction model, and predict the safety hazard coefficient of the water conservancy project. S4. Compare the predicted safety hazard coefficient of the water conservancy project with the preset safety hazard coefficient threshold of the water conservancy project to determine whether there is a safety hazard in the water conservancy project. S5. If there are no hidden dangers in the water conservancy project, the inspection terminal will issue a monitoring instruction, and the staff will conduct inspections of the pre-set inspection points through video monitoring data. If there are potential hazards in the water conservancy project, the shoreline feature dataset will be associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video surveillance data will be linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios; S6. Construct a hazard type prediction model, input the safety hazard coefficient of the water conservancy project into the hazard type prediction model, and predict the hazard type level; formulate inspection routes and maintenance plans based on the hazard type level in the GIS-BIM three-dimensional spatial information model, and select the best inspection route and optimal maintenance plan through the A* algorithm.
[0007] As a further aspect of the present invention: the shoreline parameter data includes the shoreline length, slope, shoreline material, and the distribution location of water conservancy engineering facilities; The water area parameter data includes water level, water flow velocity, water flow pattern, and water area; The reservoir surrounding environmental data includes reservoir surrounding meteorological data, reservoir surrounding vegetation coverage, and the probability of geological disasters occurring in the reservoir surrounding area; The video surveillance data includes continuous video frame data, the camera's geographical coordinates, the camera's rotation angle, and the timestamp of each video frame.
[0008] As a further aspect of the present invention: In S2, the specific steps for obtaining the comprehensive feature dataset are as follows: The LOF algorithm is used to identify and remove abnormal data in shoreline parameter data, water area parameter data and reservoir surrounding environment data to obtain processed shoreline feature dataset, water area feature dataset and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are normalized and converted into a standard normal distribution to obtain the normalized shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are fused together using a weighted formula to form a comprehensive feature dataset.
[0009] As a further aspect of the present invention: in S3, the training method of the safety hazard prediction model includes: The dataset is divided into training, validation, and test sets for training and evaluating model performance. A fully connected neural network model is constructed as a security risk prediction model, which includes an input layer, a hidden layer, and an output layer. The hidden layer uses the ReLU activation function. The number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset, and is used to input the historical comprehensive feature dataset. The number of neurons in the output layer corresponds to the number of predicted targets, and the safety hazard coefficient of the water conservancy project is output through one neuron. The mean squared error is used as the loss function, and the safety hazard prediction model is trained using the training set data. The loss function is minimized using the SGD optimizer. The performance of the safety hazard prediction model is evaluated and tuned using the validation set, and the performance of the final safety hazard prediction model is evaluated using the test set.
[0010] As a further aspect of the present invention: In S4, the method for determining whether a water conservancy project has potential safety hazards includes: If the predicted safety hazard coefficient of a water conservancy project is greater than or equal to the preset safety hazard coefficient threshold of a water conservancy project, then the water conservancy project is determined to have safety hazards. If the predicted safety hazard coefficient of a water conservancy project is less than the preset threshold for the safety hazard coefficient of a water conservancy project, then the water conservancy project is determined to have no safety hazards.
[0011] As a further aspect of the present invention: In S5, the method for constructing a GIS-BIM three-dimensional spatial information model includes: The spatial correlation between the locations of different data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset is quantified by constructing a variogram function. The mutation function is: ; in, Distance The value of the mutation function; This represents the distance between data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset. For spatial location Observed values; For spatial location Observed values; The distance is The number of point pairs; For example, an observation point in a data point pair; A spherical model is chosen as a suitable variogram model, and the spherical model is as follows: ; in, This is the base value; For variable range; The value of the target point is estimated using known observations through the Kriging estimation formula; The weighting coefficients in the Kriging estimation formula are constrained by adjusting the weighting coefficients in the model. The prediction error of the Kriging estimate is estimated by using the Kriging variance; Kriging interpolation was used to correlate the shoreline feature dataset with the reservoir surrounding environment feature dataset, and cross-validation was used to verify the accuracy of the Kriging estimation formula. The spatial estimates obtained by Kriging estimation were then integrated with the original shoreline feature dataset and the reservoir surrounding environment feature dataset, so that all data were in the same GIS or BIM platform. Let the spatial estimate obtained by Kriging be denoted as The shoreline feature dataset is The database features the following dataset: The integrated dataset is The integrated dataset was generated using GIS and BIM software. Construct a GIS-BIM three-dimensional spatial information model.
[0012] As a further aspect of the present invention: In S5, the method of linking video surveillance data with the GIS-BIM three-dimensional spatial information model and conducting inspections of preset inspection points through simulated scenarios includes: Extracting GIS and BIM data from a GIS-BIM 3D spatial information model using GIS and BIM software tools; Extract each frame of image from video surveillance data; Using camera location and shooting angle information from video surveillance data, camera localization and attitude estimation are performed using the PnP algorithm. The feature points in the video frame are matched with the geographic feature points in the GIS data, and the correspondence between the video frame and the GIS data is found by using the ORB algorithm. The building or facility features in the video frame are matched with the corresponding features in the BIM model, and the correspondence between the video frame and the BIM data is found by using the ICP algorithm. Video surveillance data is overlaid on a GIS-BIM 3D model, and the fused data is displayed in a 3D view using rendering technology. Real-time video monitoring data is collected by sensors, and the status and environmental information in the GIS-BIM three-dimensional spatial information model are continuously updated using the GIS-BIM three-dimensional spatial information model update formula. The influence factors in the GIS-BIM 3D spatial information model update formula are restricted by the influence factor restriction formula; The inspection path is optimized using Dijkstra's algorithm, and inspection simulation is performed using virtual inspection technology. Target detection and tracking algorithms are used to detect and track targets from video surveillance data. By combining video surveillance data and GIS-BIM three-dimensional spatial information model, the three-dimensional positioning of the targets is achieved, and inspections are carried out at preset inspection points.
[0013] As a further aspect of the present invention: in S6, the training method of the hazard type prediction model includes: The dataset is divided into a training set, a validation set, and a test set; a convolutional neural network model is constructed as a hazard type prediction model, which includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; The input layer is used to input the safety hazard coefficients of historical water conservancy projects, and the output layer is used to output the hazard type level. The output layer uses the softmax activation function. Multi-class cross-entropy is used as the loss function of the model, and the hazard type prediction model is trained using the training set. The model parameters are updated through the backpropagation algorithm to minimize the loss function. The performance of the hazard type prediction model is evaluated by calculating the accuracy metric using the validation set. The Adam optimization algorithm was selected as the optimizer, and the model was tuned based on the performance feedback of the validation set. The performance of the model in the prediction task was evaluated using the test set, and the trained hazard type prediction model was used to predict the safety hazard coefficient of the current water conservancy project to obtain the hazard type level.
[0014] As a further aspect of the present invention: in S6, the method for selecting the optimal inspection route and the optimal maintenance plan includes: The types and levels of hazards are divided into hazard types and hazard levels; Based on the type and level of the hazard, key areas and inspection points that need to be inspected are marked in the GIS-BIM three-dimensional spatial information model to form a set of inspection points; Based on the inspection results and the level of the hazard, a maintenance plan associated with the inspection point is formulated, forming a set of maintenance plans; In the set of inspection points, an inspection start point and an end point are preset, and the shortest path from the start point to the end point is found as the optimal inspection route using the A* algorithm. Define the set of costs required for each maintenance plan. Within this set, a maintenance start point and an end point are preset. The A* algorithm is used to find the minimum path cost from the maintenance start point to the maintenance end point. The corresponding maintenance plan is the optimal maintenance plan.
[0015] The present invention also provides a water conservancy project inspection and maintenance system for implementing the above-mentioned water conservancy project inspection and maintenance method, comprising: The data acquisition module is used to collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data. The data processing module is used to preprocess shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain shoreline feature datasets, water area feature datasets, and reservoir surrounding environment feature datasets; and to perform weighted fusion of the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain a comprehensive feature dataset. The safety hazard prediction module is used to build a safety hazard prediction model. The comprehensive feature dataset is input into the safety hazard prediction model to predict the safety hazard coefficient of the water conservancy project. The safety hazard assessment module is used to compare the predicted safety hazard coefficient of water conservancy projects with the preset safety hazard coefficient threshold of water conservancy projects to determine whether there are safety hazards in water conservancy projects. In the inspection management module, if there are no hidden dangers in the water conservancy project, the inspection terminal issues a monitoring command, and the staff conducts inspections of the pre-set inspection points through video monitoring data; if there are hidden dangers in the water conservancy project, the shoreline feature dataset is associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video monitoring data is linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios; The inspection route planning module is used to build a hazard type prediction model. The safety hazard coefficient of the water conservancy project is input into the hazard type prediction model to predict the hazard type level. Based on the hazard type level, the inspection route and maintenance plan are formulated in the GIS-BIM three-dimensional spatial information model, and the best inspection route and optimal maintenance plan are selected by A* algorithm. The modules are connected to each other by wired and / or wireless means.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: By constructing a variogram and applying Kriging interpolation, the shoreline feature dataset and the reservoir surrounding environment feature dataset are correlated and integrated to construct a GIS-BIM three-dimensional spatial information model. The use of the variogram and spherical model effectively quantifies the spatial correlation and variability between different data points, making the description of spatial features more accurate. This helps to deepen the understanding of the interaction between the shoreline and the reservoir surrounding environment. By adjusting the weight coefficients of the model, the weight coefficients are restricted to ensure that observation points with stronger spatial correlation to the target point receive greater weights in the estimation process, thereby improving the accuracy of the interpolation results. Reasonable weights are assigned to different data sources, thereby achieving effective fusion and comprehensive analysis of multi-source data. Through real-time acquisition of video surveillance and sensor data, the GIS-BIM 3D spatial information model can reflect the changes in the surrounding environment in real time. Utilizing the GIS-BIM 3D spatial information model update formula, combined with real-time monitoring and sensor data, the model dynamically updates state and environmental information, ensuring data timeliness and accuracy. The Dijkstra algorithm is used to optimize inspection paths, and combined with virtual inspection technology, the inspection process is simulated, improving inspection efficiency and accuracy. During inspections and emergency response, potential risks can be identified and assessed in advance, allowing for effective measures to reduce the likelihood of accidents. By subdividing hazard types and levels, we can accurately pinpoint areas and inspection points requiring special attention, avoiding blind inspections and improving the targeting and efficiency of inspections. Utilizing the A* algorithm to plan the shortest path from the inspection start point to the end point ensures optimal inspection routes, reducing inspection time and labor costs. By calculating the required cost of maintenance plans and using the A* algorithm to find the lowest-cost maintenance path, we achieve economic optimization of maintenance plans, reducing maintenance costs. This helps to promptly identify and eliminate potential safety hazards, improving safety management levels. Attached Figure Description
[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of a water conservancy project inspection and maintenance method according to the present invention; Figure 2 This is a schematic diagram of the structure of a water conservancy project inspection and maintenance system according to the present invention. Detailed Implementation
[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] The present invention will be further explained below with reference to specific embodiments.
[0021] Example 1: Please see Figure 1 and Figure 2 As shown in the figure, this embodiment provides a method for the inspection and maintenance of water conservancy projects, including: S1. Collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data; S2. Preprocess the shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. Then, perform weighted fusion on the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain the comprehensive feature dataset. S3. Construct a safety hazard prediction model, input the comprehensive feature dataset into the safety hazard prediction model, and predict the safety hazard coefficient of the water conservancy project. S4. Compare the predicted safety hazard coefficient of the water conservancy project with the preset safety hazard coefficient threshold of the water conservancy project to determine whether there is a safety hazard in the water conservancy project. S5. If there are no hidden dangers in the water conservancy project, the inspection terminal will issue a monitoring command, and the staff will inspect the pre-set inspection points through video monitoring data; if there are hidden dangers in the water conservancy project, the shoreline feature dataset will be associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video monitoring data will be linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios. S6. Construct a hazard type prediction model, input the safety hazard coefficient of water conservancy project into the hazard type diagnosis model, and predict the hazard type level; formulate inspection routes and maintenance plans based on the hazard type level in the GIS-BIM three-dimensional spatial information model, and select the best inspection route and optimal maintenance plan through the A* algorithm.
[0022] Shoreline parameter data includes shoreline length, slope, shoreline material, and the distribution of water conservancy facilities; water area parameter data includes water level, flow velocity, flow pattern, and water area; reservoir surrounding environment data includes reservoir surrounding meteorological data, reservoir surrounding vegetation coverage, and the probability of reservoir surrounding geological disasters; reservoir surrounding meteorological data includes annual rainfall, rainfall distribution, wind direction and speed, temperature, and humidity; video surveillance data includes continuous video frame data, camera geographic coordinates, camera rotation angle, and timestamps for each video frame.
[0023] The length of the shoreline was obtained through drone aerial photography; the slope was obtained through total station measurement; the shoreline material was obtained through on-site observation and sampling by staff; the distribution of water conservancy facilities was obtained through spatial querying on the GIS platform; the water level of the water area was estimated through changes in water body characteristics in satellite remote sensing images; the water flow velocity was measured by an acoustic Doppler current meter; the water flow pattern was obtained through observation by deploying buoys; and the water area was calculated by extracting water bodies from satellite remote sensing images. Meteorological data around the reservoir was estimated using satellite remote sensing data; vegetation coverage around the reservoir was calculated using high-resolution remote sensing images; the probability of geological disasters around the reservoir was predicted using a GIS platform; and video surveillance data was obtained through video surveillance cameras installed at key locations in the reservoir area. Methods for preprocessing shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain shoreline feature datasets, water area feature datasets, and reservoir surrounding environment feature datasets, and then weightedly fusing these datasets to obtain a comprehensive feature dataset, include: The LOF algorithm is used to identify and remove abnormal data in shoreline parameter data, water area parameter data and reservoir surrounding environment data to obtain processed shoreline feature dataset, water area feature dataset and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are normalized and converted into a standard normal distribution to obtain the normalized shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are fused using a weighted formula to form a comprehensive feature dataset; the shoreline feature dataset is denoted as... The water feature dataset is denoted as The database surrounding environment feature dataset is denoted as ; The weighted formula is: ; in, For comprehensive feature datasets; These are the weighting coefficients for the shoreline feature dataset; These are the weighting coefficients for the water feature dataset; These are the weighting coefficients for the database's surrounding environment feature dataset.
[0024] Training methods for safety hazard prediction models include: The dataset is divided into training, validation, and test sets for training and evaluating model performance. A security risk prediction model is constructed, which includes an input layer, hidden layers, and an output layer. The hidden layer uses the ReLU activation function. The input layer of the model is a historical comprehensive feature dataset, and the number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset; the output layer of the model is the safety hazard coefficient of water conservancy projects, and the number of neurons in the output layer corresponds to the number of prediction targets, with a single neuron outputting the prediction value; the safety hazard prediction model is a fully connected neural network model. Mean squared error is used as the loss function to measure the difference between the model's predicted values and the actual values; the mean squared error loss function is: ; in, Number of datasets; For data points in the dataset The actual value; For data points in the dataset The predicted value; The model is trained using the training set data, and the loss function is minimized using the SGD optimizer. The model's performance is evaluated using the validation set, the model is tuned, and the model's hyperparameters are adjusted based on the performance feedback until the model's performance no longer improves significantly or the preset stopping condition is met. The performance of the model in the prediction task is evaluated using a test set. The current comprehensive feature dataset is input into the trained safety hazard prediction model to obtain the safety hazard coefficient of water conservancy projects.
[0025] Methods for determining whether a water conservancy project has safety hazards by comparing the predicted safety hazard coefficient with the preset safety hazard coefficient threshold include: If the predicted safety hazard coefficient of a water conservancy project is greater than or equal to the preset safety hazard coefficient threshold of a water conservancy project, then the water conservancy project is determined to have safety hazards. If the predicted safety hazard coefficient of a water conservancy project is less than the preset threshold for the safety hazard coefficient of a water conservancy project, then the water conservancy project is determined to have no safety hazards.
[0026] If there are potential hazards in a water conservancy project, the method of constructing a GIS-BIM three-dimensional spatial information model by associating the shoreline feature dataset with the reservoir surrounding environmental feature dataset includes: The spatial correlation between the locations of different data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset is quantified by constructing a variogram function. The variability function is: ; in, Distance The value of the mutation function; This represents the distance between data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset. For spatial location Observed values; For spatial location Observed values; The distance is The number of point pairs; For example, an observation point in a data point pair; A spherical model is chosen as the appropriate variogram model. The spherical model is as follows: ; in, The sill value represents the stable value of the variogram when the distance is sufficiently large; For variable range, it represents the effective distance of spatial autocorrelation; The Kriging estimation formula is used to estimate the value of the target point using known observations. The Kriging estimation formula is as follows: ;in, For target location Kriging's estimate; For position Observed values; For the first The weight coefficient of each observation point reflects the observation point Contribution to the target point estimate; The number of observations used for Kriging estimation; The weight coefficients in the Kriging estimation formula are constrained by a weight coefficient adjustment model, which is as follows: ; in, For the first Kriging weighting coefficients for each observation point; For observation point With the target point The values of the variation function between them; For target point The average of the variogram values of all surrounding observation points; This represents the total number of observation points; This formula uses the target point The average variogram value of all surrounding observation points With specific observation points variogram value The ratio is calculated to obtain the result for each observation point. Weighting coefficients Due to the value of the variogram in the molecule It may become smaller, causing the weighting coefficient to decrease. The increase indicates that the closer the observation point is to the target point, the greater its impact on the estimation. For example, suppose we have the following information: number of observation points The value is 3, representing the variogram value between the observation point and the target point. , , The average of the variogram values of all observation points surrounding the target point. It is 0.4; For the first observation point: ; For the second observation point: ; For the third observation point: ; Assume the observed values at the observation point are: , , ; Substituting the weighting coefficients and observations into the Kriging estimation formula: ; The prediction error of the Kriging estimate is estimated using the Kriging variance. The Kriging variance is: ; in, For target point Kriging variance; The variogram value represents the variance of the target point itself. For observation point With the target point The variogram values between the observation points reflect the variogram values between the observation points. Impact on target point estimates; For observation point With observation point The variogram values between the observation points reflect the spatial correlation between them. For the first Weighting coefficients for each observation point; Kriging interpolation was used to correlate the shoreline feature dataset with the reservoir surrounding environment feature dataset, and cross-validation was used to verify the accuracy of the Kriging estimation formula. The spatial estimates obtained by Kriging estimation were then integrated with the original shoreline feature dataset and the reservoir surrounding environment feature dataset, so that all data were in the same GIS or BIM platform. For example, the following data points were observed on the same day, and their locations are distributed at different points around the reservoir; the shoreline feature dataset is shown in Table 1: Table 1. Shoreline Feature Dataset
[0027] The database environment feature dataset is shown in Table 2: Table 2. Database Surrounding Environment Characteristics Dataset
[0028] For slope data, calculate the variogram values at different distances; for vegetation cover data, calculate the variogram values at different distances; set the spherical model parameters: D=0.1, a=3; set the target point. We calculate the weighting coefficients, use shoreline slope and vegetation cover data to perform kriging estimation, and obtain the estimated values of slope and vegetation cover at the target point; we calculate the kriging variance of the target point to assess the uncertainty of the estimated values.
[0029] Let the spatial estimate obtained by Kriging be denoted as The shoreline feature dataset is The database features the following dataset: The integrated dataset is The integrated dataset was generated using GIS and BIM software. Construct a GIS-BIM three-dimensional spatial information model.
[0030] Methods that link video surveillance data with GIS-BIM 3D spatial information models to conduct inspections of pre-set inspection points through simulated scenarios include: GIS and BIM data are extracted from the GIS-BIM 3D spatial information model using GIS and BIM software tools. The GIS data includes reservoir perimeter meteorological data, reservoir perimeter vegetation coverage, and the probability of geological disasters occurring around the reservoir. The reservoir perimeter meteorological data includes annual rainfall, rainfall distribution, wind direction and speed, temperature, and humidity. The BIM data includes shoreline length, slope, shoreline material, and the distribution and location of water conservancy engineering facilities. Extracting each frame of image from video surveillance data: ; in, For the first Frame image; Total number of frames; Using camera location and shooting angle information from video surveillance data, the PnP algorithm is used for camera localization and attitude estimation. ; in, It is a rotation matrix; It is a translation vector; For three-dimensional points; For the corresponding two-dimensional image points; This formula indicates that the camera pose is calculated from 3D points and 2D image points using the PnP algorithm; feature points in the video frame are matched with geographic feature points in GIS data, and the ORB algorithm is used to find the correspondence between the video frame and the GIS data; the ORB algorithm is as follows: ; in, It is a homography matrix; A 3D point cloud in a video frame; A 3D point cloud in GIS data; A function for calculating the homography matrix; The system matches building or facility features in video frames with corresponding features in the BIM model, using the ICP algorithm to find the correspondence between video frames and BIM data. The ICP algorithm is as follows: ; in, The transformation matrix; A 3D point cloud in a video frame; For 3D point clouds in BIM data; A function for calculating the transformation matrix; Video surveillance data is overlaid on a GIS-BIM 3D model, and the merged data is displayed in a 3D view using rendering technology. ; in, For displaying the merged video; For video surveillance data; For GIS-BIM three-dimensional spatial information model; This is a rendering function that integrates multiple data sources and displays them in a 3D view. Real-time video monitoring data is collected by sensors, and the status and environmental information in the GIS-BIM three-dimensional spatial information model are continuously updated using the GIS-BIM three-dimensional spatial information model update formula. The formula for updating the GIS-BIM 3D spatial information model is: ; in, For time The status of the GIS-BIM 3D spatial information model at any given time; For time The status of the GIS-BIM 3D spatial information model at any given time; For time Real-time video surveillance data; For time Sensor data at any given moment; Influence factors in updating formulas for GIS-BIM 3D spatial information models; The influence factors in the GIS-BIM 3D spatial information model update formula are restricted using an influence factor constraint formula; the influence factor constraint formula is as follows: ; in, The time required to collect video surveillance data and sensor data; The number of times video surveillance data and sensor data are collected; For the required time Weighting coefficients; For the number of times Weighting coefficients; For example, if the time required to collect video surveillance data and sensor data is 10 seconds, and the number of times the data is collected is 5, with a weighting factor of 0.4 for the time required and a weighting factor of 0.6 for the number of times, then the influence factor... Approximately 0.29; The larger the numerator and the smaller the denominator, the greater the impact factor. This means that longer time periods and fewer data collection frequencies usually result in a greater impact on model updates. The Dijkstra algorithm is used to optimize the inspection path to ensure efficient coverage of inspection points. Virtual inspection technology is used to simulate inspections. Target detection and tracking algorithms are used to detect and track targets from video surveillance data. Combined with video surveillance data and GIS-BIM 3D spatial information model, the 3D positioning of targets is achieved, and inspections are carried out at preset inspection points.
[0031] Training methods for hazard type prediction models include: The dataset is divided into training, validation, and test sets; a hazard type prediction model is constructed, which includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; the input layer is the safety hazard coefficient of historical water conservancy projects, and the output layer is the hazard type level; softmax is used as the activation function; the hazard type prediction model is a convolutional neural network model. Multi-class cross-entropy is used as the model's loss function to measure the difference between the model's predicted values and the actual values; the multi-class cross-entropy loss function is: ; in, The average loss of the dataset; This represents the total number of samples in the dataset. The number of hazard type levels; For the first For the nth sample Real labels of various types; The model predicts the first The sample belongs to the first The probability of each type; The hazard type prediction model is trained using the training set, and the model parameters are updated using the backpropagation algorithm to minimize the loss function. The performance of the hazard type prediction model is evaluated by calculating the accuracy metric using the validation set. The Adam optimization algorithm was selected as the optimizer. The model was tuned based on the performance feedback of the validation set. The model parameters were adjusted until the performance no longer improved significantly or the preset stopping condition was reached. The performance of the model in the prediction task was evaluated using the test set. The trained hazard type prediction model was used to predict the safety hazard coefficient of the current water conservancy project and obtain the hazard type level.
[0032] In a GIS-BIM 3D spatial information model, methods for developing inspection routes and maintenance plans based on the type and level of potential hazards, and for selecting the optimal inspection route and maintenance plan using the A* algorithm, include: The hazard type is divided into hazard type and hazard level, and the different hazard types are combined into a hazard type set. Different levels of hazards constitute a set of hazard levels. Based on the type and level of the hazard, mark all key areas and inspection points that need to be inspected in the GIS-BIM three-dimensional spatial information model, and plan the inspection route. A set of inspection points is preset, and each inspection point The set of inspection points is associated with the type and level of hazard. ; in, Gather at the inspection point; This refers to the number of inspection points; Based on the inspection results and the level of potential hazards, develop corresponding maintenance plans. These plans should include the necessary materials, equipment, personnel, and maintenance procedures. A set of maintenance plans should be pre-defined, with each plan... With inspection points Related; The maintenance plan is as follows: ; in, A collection of maintenance plans; The number of maintenance plans; grouped at pre-set inspection points. In the middle, the preset inspection point starting point is The preset inspection point endpoint is Find the source using the A* algorithm arrive The shortest path is the optimal inspection route; Define the set of costs required for the maintenance plan as follows: In the cost set required for the maintenance plan In the middle, the preset maintenance starting point is The preset maintenance endpoint is Find the maintenance starting point using the A* algorithm. To the end of the maintenance The minimum path cost is the lowest cost required for the maintenance plan, and the maintenance plan at this point is the optimal maintenance plan.
[0033] The preset threshold for the safety hazard coefficient of water conservancy projects is set by the staff. The safety hazard coefficients of different water conservancy projects are collected through the water conservancy project management terminal, and the average value of multiple water conservancy project safety hazard coefficients is taken as the preset threshold for the safety hazard coefficient of water conservancy projects. In this embodiment, a GIS-BIM three-dimensional spatial information model is constructed by constructing a variogram and applying Kriging interpolation to correlate and integrate the shoreline feature dataset with the reservoir surrounding environment feature dataset. The use of the variogram and spherical model effectively quantifies the spatial correlation and variability between different data points, making the description of spatial features more accurate. This helps to deepen the understanding of the interaction between the shoreline and the reservoir surrounding environment. By adjusting the weight coefficients of the model, the weight coefficients are restricted to ensure that observation points with stronger spatial correlation with the target point receive greater weights in the estimation process, thereby improving the accuracy of the interpolation results. Reasonable weights are assigned to different data sources, thereby achieving effective fusion and comprehensive analysis of multi-source data. Through real-time acquisition of video surveillance and sensor data, the GIS-BIM 3D spatial information model can reflect the changes in the surrounding environment in real time. Utilizing the GIS-BIM 3D spatial information model update formula, combined with real-time monitoring and sensor data, the model dynamically updates state and environmental information, ensuring data timeliness and accuracy. The Dijkstra algorithm is used to optimize inspection paths, and combined with virtual inspection technology, the inspection process is simulated, improving inspection efficiency and accuracy. During inspections and emergency response, potential risks can be identified and assessed in advance, allowing for effective measures to reduce the likelihood of accidents. By subdividing hazard types and levels, we can accurately pinpoint areas and inspection points requiring special attention, avoiding blind inspections and improving the targeting and efficiency of inspections. Utilizing the A* algorithm to plan the shortest path from the inspection start point to the end point ensures optimal inspection routes, reducing inspection time and labor costs. By calculating the required cost of maintenance plans and using the A* algorithm to find the lowest-cost maintenance path, we achieve economic optimization of maintenance plans, reducing maintenance costs. This helps to promptly identify and eliminate potential safety hazards, improving safety management levels.
[0034] Example 2: Please see Figure 2 As shown, this embodiment provides a water conservancy project inspection and maintenance system, including: The data acquisition module is used to collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data. The data processing module is used to preprocess shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain shoreline feature datasets, water area feature datasets, and reservoir surrounding environment feature datasets; and to perform weighted fusion of the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain a comprehensive feature dataset. The safety hazard prediction module is used to build a safety hazard prediction model. The comprehensive feature dataset is input into the safety hazard prediction model to predict the safety hazard coefficient of the water conservancy project. The safety hazard assessment module is used to compare the predicted safety hazard coefficient of water conservancy projects with the preset safety hazard coefficient threshold of water conservancy projects to determine whether there are safety hazards in water conservancy projects. In the inspection management module, if there are no hidden dangers in the water conservancy project, the inspection terminal issues a monitoring command, and the staff conducts inspections of the pre-set inspection points through video monitoring data; if there are hidden dangers in the water conservancy project, the shoreline feature dataset is associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video monitoring data is linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios; The inspection route planning module is used to build a hazard type prediction model. The safety hazard coefficient of the water conservancy project is input into the hazard type diagnosis model to predict the hazard type level. Based on the hazard type level, the inspection route and maintenance plan are formulated in the GIS-BIM three-dimensional spatial information model, and the best inspection route and optimal maintenance plan are selected by A* algorithm.
[0035] Since the electronic device involved in this embodiment is the electronic device used in implementing the water conservancy project inspection and maintenance method and system described in this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the water conservancy project inspection and maintenance method and system described in this application. Therefore, how the electronic device implements the method in this application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the water conservancy project inspection and maintenance method and system described in this application, it falls within the scope of protection of this application.
[0036] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.
[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for inspection and maintenance of water conservancy projects, characterized in that, The method includes the following steps: S1. Collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data; S2. Preprocess the shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. Then, perform weighted fusion on the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain the comprehensive feature dataset. S3. Construct a safety hazard prediction model, input the comprehensive feature dataset into the safety hazard prediction model, and predict the safety hazard coefficient of the water conservancy project. S4. Compare the predicted safety hazard coefficient of the water conservancy project with the preset safety hazard coefficient threshold of the water conservancy project to determine whether there is a safety hazard in the water conservancy project. S5. If there are no hidden dangers in the water conservancy project, the inspection terminal will issue a monitoring instruction, and the staff will conduct inspections of the pre-set inspection points through video monitoring data. If there are potential hazards in the water conservancy project, the shoreline feature dataset will be associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video surveillance data will be linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios; S6. Construct a hazard type prediction model, input the safety hazard coefficient of the water conservancy project into the hazard type prediction model, and predict the hazard type level; formulate inspection routes and maintenance plans based on the hazard type level in the GIS-BIM three-dimensional spatial information model, and select the best inspection route and optimal maintenance plan through the A* algorithm.
2. The method for inspection and maintenance of water conservancy projects according to claim 1, characterized in that, The shoreline parameter data includes the shoreline length, slope, shoreline material, and the distribution location of water conservancy engineering facilities; The water area parameter data includes water level, water flow velocity, water flow pattern, and water area; The reservoir surrounding environmental data includes reservoir surrounding meteorological data, reservoir surrounding vegetation coverage, and the probability of geological disasters occurring in the reservoir surrounding area; The video surveillance data includes continuous video frame data, the camera's geographical coordinates, the camera's rotation angle, and the timestamp of each video frame.
3. The method for inspection and maintenance of water conservancy projects according to claim 2, characterized in that, In S2, the specific steps to obtain the comprehensive feature dataset are as follows: The LOF algorithm is used to identify and remove abnormal data in shoreline parameter data, water area parameter data and reservoir surrounding environment data to obtain processed shoreline feature dataset, water area feature dataset and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are normalized and converted into a standard normal distribution to obtain the normalized shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset. The shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset are fused together using a weighted formula to form a comprehensive feature dataset.
4. The method for inspection and maintenance of water conservancy projects according to claim 3, characterized in that, In S3, the training method for the safety hazard prediction model includes: The dataset is divided into training, validation, and test sets for training and evaluating model performance. A fully connected neural network model is constructed as a security risk prediction model, which includes an input layer, a hidden layer, and an output layer. The hidden layer uses the ReLU activation function. The number of neurons in the input layer matches the number of features in the historical comprehensive feature dataset, and is used to input the historical comprehensive feature dataset. The number of neurons in the output layer corresponds to the number of predicted targets, and the safety hazard coefficient of the water conservancy project is output through one neuron. The mean squared error is used as the loss function, and the safety hazard prediction model is trained using the training set data. The loss function is minimized using the SGD optimizer. The performance of the safety hazard prediction model is evaluated and tuned using the validation set, and the performance of the final safety hazard prediction model is evaluated using the test set.
5. The method for inspection and maintenance of water conservancy projects according to claim 4, characterized in that, In S4, the methods for determining whether a water conservancy project has potential safety hazards include: If the predicted safety hazard coefficient of a water conservancy project is greater than or equal to the preset safety hazard coefficient threshold of a water conservancy project, then the water conservancy project is determined to have safety hazards. If the predicted safety hazard coefficient of a water conservancy project is less than the preset threshold for the safety hazard coefficient of a water conservancy project, then the water conservancy project is determined to have no safety hazards.
6. The method for inspection and maintenance of water conservancy projects according to claim 5, characterized in that, In S5, the methods for constructing GIS-BIM 3D spatial information models include: The spatial correlation between the locations of different data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset is quantified by constructing a variogram function. The mutation function is: ； in, Distance The value of the mutation function; This represents the distance between data points in the shoreline feature dataset and the reservoir surrounding environment feature dataset. For spatial location Observed values; For spatial location Observed values; The distance is The number of point pairs; For example, an observation point in a data point pair; A spherical model is chosen as a suitable variogram model, and the spherical model is as follows: ； in, This is the base value; For variable range; The value of the target point is estimated using known observations through the Kriging estimation formula; The weighting coefficients in the Kriging estimation formula are constrained by adjusting the weighting coefficients in the model. The prediction error of the Kriging estimate is estimated by using the Kriging variance; Kriging interpolation was used to correlate the shoreline feature dataset with the reservoir surrounding environment feature dataset, and cross-validation was used to verify the accuracy of the Kriging estimation formula. The spatial estimates obtained by Kriging estimation were then integrated with the original shoreline feature dataset and the reservoir surrounding environment feature dataset, so that all data were in the same GIS or BIM platform. The spatial estimate obtained by Kriging is denoted as... The shoreline feature dataset is The database features the following dataset: The integrated dataset is The integrated dataset was generated using GIS and BIM software. Construct a GIS-BIM three-dimensional spatial information model.
7. The method for inspection and maintenance of water conservancy projects according to claim 6, characterized in that, In S5, the methods for linking video surveillance data with GIS-BIM 3D spatial information models and conducting inspections of pre-set inspection points through simulated scenarios include: Extracting GIS and BIM data from a GIS-BIM 3D spatial information model using GIS and BIM software tools; Extract each frame of image from video surveillance data; Using camera location and shooting angle information from video surveillance data, camera localization and attitude estimation are performed using the PnP algorithm. The feature points in the video frame are matched with the geographic feature points in the GIS data, and the correspondence between the video frame and the GIS data is found by using the ORB algorithm. The building or facility features in the video frame are matched with the corresponding features in the BIM model, and the correspondence between the video frame and the BIM data is found by using the ICP algorithm. Video surveillance data is overlaid on a GIS-BIM 3D model, and the fused data is displayed in a 3D view using rendering technology. Real-time video monitoring data is collected by sensors, and the status and environmental information in the GIS-BIM three-dimensional spatial information model are continuously updated using the GIS-BIM three-dimensional spatial information model update formula. The influence factors in the GIS-BIM 3D spatial information model update formula are restricted by the influence factor restriction formula; The inspection path is optimized using Dijkstra's algorithm, and inspection simulation is performed using virtual inspection technology. Target detection and tracking algorithms are used to detect and track targets from video surveillance data. By combining video surveillance data and GIS-BIM three-dimensional spatial information model, the three-dimensional positioning of the targets is achieved, and inspections are carried out at preset inspection points.
8. The method for inspection and maintenance of water conservancy projects according to claim 7, characterized in that, In S6, the training method for the hazard type prediction model includes: The dataset is divided into a training set, a validation set, and a test set; a convolutional neural network model is constructed as a hazard type prediction model, which includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; The input layer is used to input the safety hazard coefficients of historical water conservancy projects, and the output layer is used to output the hazard type level. The output layer uses the softmax activation function. Multi-class cross-entropy is used as the loss function of the model, and the hazard type prediction model is trained using the training set. The model parameters are updated through the backpropagation algorithm to minimize the loss function. The performance of the hazard type prediction model is evaluated by calculating the accuracy metric using the validation set. The Adam optimization algorithm was selected as the optimizer, and the model was tuned based on the performance feedback of the validation set. The performance of the model in the prediction task was evaluated using the test set, and the trained hazard type prediction model was used to predict the safety hazard coefficient of the current water conservancy project to obtain the hazard type level.
9. The method for inspection and maintenance of water conservancy projects according to claim 8, characterized in that, In S6, the methods for selecting the optimal inspection route and the optimal maintenance plan include: The types and levels of hazards are divided into hazard types and hazard levels; Based on the type and level of the hazard, key areas and inspection points that need to be inspected are marked in the GIS-BIM three-dimensional spatial information model to form a set of inspection points; Based on the inspection results and the level of the hazard, a maintenance plan associated with the inspection point is formulated, forming a set of maintenance plans; In the set of inspection points, an inspection start point and an end point are preset, and the shortest path from the start point to the end point is found as the optimal inspection route using the A* algorithm. Define the set of costs required for each maintenance plan. Within this set, a maintenance start point and an end point are preset. The A* algorithm is used to find the minimum path cost from the maintenance start point to the maintenance end point. The corresponding maintenance plan is the optimal maintenance plan.
10. A water conservancy project inspection and maintenance system, used to implement the water conservancy project inspection and maintenance method according to any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect shoreline parameter data, water area parameter data, reservoir surrounding environment data, and video surveillance data. The data processing module is used to preprocess shoreline parameter data, water area parameter data, and reservoir surrounding environment data to obtain shoreline feature datasets, water area feature datasets, and reservoir surrounding environment feature datasets; and to perform weighted fusion of the shoreline feature dataset, water area feature dataset, and reservoir surrounding environment feature dataset to obtain a comprehensive feature dataset. The safety hazard prediction module is used to build a safety hazard prediction model. The comprehensive feature dataset is input into the safety hazard prediction model to predict the safety hazard coefficient of the water conservancy project. The safety hazard assessment module is used to compare the predicted safety hazard coefficient of water conservancy projects with the preset safety hazard coefficient threshold of water conservancy projects to determine whether there are safety hazards in water conservancy projects. In the inspection management module, if there are no hidden dangers in the water conservancy project, the inspection terminal issues a monitoring command, and the staff conducts inspections of the pre-set inspection points through video monitoring data; if there are hidden dangers in the water conservancy project, the shoreline feature dataset is associated with the reservoir surrounding environment feature dataset to construct a GIS-BIM three-dimensional spatial information model; the video monitoring data is linked with the GIS-BIM three-dimensional spatial information model to conduct inspections of the pre-set inspection points through simulated scenarios; The inspection route planning module is used to build a hazard type prediction model. The safety hazard coefficient of the water conservancy project is input into the hazard type prediction model to predict the hazard type level. Based on the hazard type level, the inspection route and maintenance plan are formulated in the GIS-BIM three-dimensional spatial information model, and the best inspection route and optimal maintenance plan are selected by A* algorithm. The modules are connected to each other by wired and / or wireless means.