Active flood monitoring method, device, electronic device and storage medium

Abstract

This invention relates to the field of flood disaster emergency response technology, and particularly to an active flood monitoring method, device, electronic equipment, and storage medium. It includes: calculating a spatiotemporal risk tensor based on rainfall forecast data and a pre-set cloud-based database using a pre-set flood model; constructing an objective function based on the spatiotemporal risk tensor, regional road network, and distribution of flood-prone points; determining the optimal path for each mobile platform based on constraints and the objective function to control the movement of each mobile platform to the risk monitoring point; determining the water accumulation area at each risk monitoring point using a pre-set target detection algorithm, and then determining the water accumulation depth at each risk monitoring point by combining it with a pre-set digital elevation map; and automatically selecting an appropriate monitoring strategy based on the water accumulation depth at each risk monitoring point to conduct flood monitoring at each risk monitoring point. This solves the problems of weak sensing capabilities, limited coverage, and untimely operational response of single monitoring methods in related technologies.

Description

Technical Field

[0001] This invention relates to the field of flood disaster emergency response technology, and in particular to an active flood monitoring method, device, electronic equipment and storage medium. Background Technology

[0002] With climate change and urbanization, urban flooding is becoming increasingly frequent, posing a serious threat to people's lives and property. The lack of comprehensive monitoring of the complex urban environment and the evolution of flood disasters makes accurate prediction and targeted management of urban flooding impossible, necessitating the use of redundant and resource-intensive defenses to minimize the impact of floods. Therefore, scientifically conducting flood monitoring to guide efficient disaster prevention and mitigation is of significant practical importance.

[0003] Among related technologies, flood monitoring strategies can be broadly categorized into three types: flood monitoring based on satellite remote sensing, flood monitoring based on drone platforms, and flood monitoring based on the Internet of Things.

[0004] However, the monitoring methods of the above-mentioned monitoring strategies are mostly based on historical flood disaster experience to determine the monitoring implementation methods. The monitoring methods are singular, the coverage is limited, and the operation response is not timely, which urgently needs to be addressed. Summary of the Invention

[0005] This invention provides an active flood monitoring method, device, electronic equipment, and storage medium to solve the problems of weak sensing ability, limited coverage, and untimely operation response of single monitoring methods in related technologies, thereby improving the efficiency and accuracy of flood disaster prevention.

[0006] To achieve the above objectives, a first aspect of the present invention provides an active flood monitoring method, comprising the following steps:

[0007] First and second rainfall forecast data for the area to be monitored are obtained. Based on the first and second rainfall forecast data and the preset cloud-based database, the spatiotemporal risk tensor is calculated using a preset flood model.

[0008] Based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points, an objective function is constructed. Based on the preset constraints and the objective function, the optimal path of each preset mobile platform is determined. Based on the optimal path of each preset mobile platform, each preset mobile platform is controlled to move from the initial stopping point to the preset risk monitoring point.

[0009] Based on a preset target detection algorithm, the water accumulation area of ​​each preset risk monitoring point is determined using each preset mobile platform, and the water accumulation depth of each preset risk monitoring point is determined based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map.

[0010] Based on the water depth at each preset risk monitoring point, a monitoring strategy is determined for each preset risk monitoring point. Based on the monitoring strategy for each preset risk monitoring point, flood monitoring is carried out at each preset risk monitoring point to obtain flood monitoring results.

[0011] According to one embodiment of the present invention, determining the monitoring strategy for each preset risk monitoring point based on the water depth at each preset risk monitoring point includes:

[0012] Based on each preset risk monitoring point, at least one first target risk monitoring point is selected where the water depth at the monitoring point is less than the preset depth value.

[0013] The preset mobile platform corresponding to at least one first target risk monitoring point is controlled to move to the target location corresponding to each first target risk monitoring point, and the water depth at the target location corresponding to each first target risk monitoring point is measured using a distance measuring water level gauge.

[0014] According to one embodiment of the present invention, determining the monitoring strategy for each preset risk monitoring point based on the water depth at each preset risk monitoring point further includes:

[0015] Based on each preset risk monitoring point, at least one second target risk monitoring point is selected where the water depth at the monitoring point is greater than or equal to the preset depth value.

[0016] The system controls a preset mobile platform corresponding to at least one second target risk monitoring point to call a preset bionic robot carrying a pressure level gauge to move to the target point corresponding to each second target risk monitoring point, and uses the pressure level gauge to measure the water depth at the target point corresponding to each second target risk monitoring point.

[0017] According to an embodiment of the present invention, the objective function is:

[0018]

[0019] in, To monitor total cost, The number of the preset mobile platforms, , , All are target weights. As a balance factor, For path distance cost, To cover the costs of risk, For energy consumption costs, This is to reduce the cost of repetitive work by multiple machines.

[0020] According to one embodiment of the present invention, before calculating the spatiotemporal risk tensor using the preset flood model based on the first rainfall forecast data, the second rainfall forecast data, and the preset cloud-based database, the method further includes:

[0021] Acquire optical image data and point cloud data of the target scene area;

[0022] Using the preset target detection algorithm, the image elements of the target object in the optical image data are detected to obtain processed optical image data. Then, using the K-nearest neighbor sampling and sparse graph convolution algorithm, the point cloud data and the optical image data are geo-registered, and the geographical coordinates of the point cloud data are assigned to the target object in the optical image data to obtain registered optical image data and registered point cloud data.

[0023] The registered optical image data is mapped to the image feature space using a first preset neural network, and the registered point cloud data is mapped to the point cloud feature space using a second preset neural network.

[0024] Using a pre-defined feature mapping network, the image feature space and the point cloud feature space are aligned to obtain the target feature vector, and a target segmentation model is trained based on the target feature vector;

[0025] The geometric boundary information of the target object is interpreted using the target segmentation model, and the geographic coordinate information and geometric size attributes of the target object are extracted. The geometric boundary information, the geographic coordinate information and the geometric size attributes are then updated to the preset cloud-based base database.

[0026] The active flood monitoring method proposed in this invention calculates a spatiotemporal risk tensor using a pre-defined flood model based on rainfall forecast data and a pre-defined cloud-based database. An objective function is constructed based on the spatiotemporal risk tensor, regional road network, and distribution of flood-prone locations. The optimal path for each mobile platform is determined based on constraints and the objective function to control its movement to pre-defined risk monitoring points. A pre-defined target detection algorithm is used to determine the water accumulation area at each risk monitoring point, and combined with a pre-defined digital elevation map, the water accumulation depth at each risk monitoring point is determined. Based on the water accumulation depth at each risk monitoring point, an appropriate monitoring strategy is automatically selected to monitor flooding at each point. This solves the problems of weak sensing capabilities, limited coverage, and untimely operational response of single monitoring methods in related technologies, improving the efficiency and accuracy of flood disaster prevention.

[0027] To achieve the above objectives, a second aspect of the present invention provides an active flood monitoring device, comprising:

[0028] The acquisition module is used to acquire the first and second rainfall forecast data of the area to be monitored, and to calculate the spatiotemporal risk tensor based on the first rainfall forecast data, the second rainfall forecast data and the preset cloud-based database using a preset flood model.

[0029] The control module is used to construct an objective function based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points, and to determine the optimal path of each preset mobile platform based on preset constraints and the objective function, and to control each preset mobile platform to move from the initial stop point to the preset risk monitoring point based on the optimal path of each preset mobile platform.

[0030] The determination module is used to determine the water accumulation area of ​​each preset risk monitoring point based on a preset target detection algorithm and each preset mobile platform, and to determine the water accumulation depth of each preset risk monitoring point based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map;

[0031] The monitoring module is used to determine the monitoring strategy for each preset risk monitoring point based on the water depth of each preset risk monitoring point, and to perform flood monitoring on each preset risk monitoring point based on the monitoring strategy of each preset risk monitoring point, so as to obtain the flood monitoring results.

[0032] According to one embodiment of the present invention, the monitoring module is specifically used for:

[0033] Based on each preset risk monitoring point, at least one first target risk monitoring point is selected where the water depth at the monitoring point is less than the preset depth value.

[0034] The preset mobile platform corresponding to at least one first target risk monitoring point is controlled to move to the target location corresponding to each first target risk monitoring point, and the water depth at the target location corresponding to each first target risk monitoring point is measured using a distance measuring water level gauge.

[0035] According to one embodiment of the present invention, the monitoring module is further configured to:

[0036] Based on each preset risk monitoring point, at least one second target risk monitoring point is selected where the water depth at the monitoring point is greater than or equal to the preset depth value.

[0037] The system controls a preset mobile platform corresponding to at least one second target risk monitoring point to call a preset bionic robot carrying a pressure level gauge to move to the target point corresponding to each second target risk monitoring point, and uses the pressure level gauge to measure the water depth at the target point corresponding to each second target risk monitoring point.

[0038] According to an embodiment of the present invention, the objective function is:

[0039]

[0040] in, To monitor total cost, The number of the preset mobile platforms, , , All are target weights. As a balance factor, For path distance cost, To cover the costs of risk, For energy consumption costs, This is to reduce the cost of repetitive work by multiple machines.

[0041] According to one embodiment of the present invention, before calculating the spatiotemporal risk tensor using the preset flood model based on the first rainfall forecast data, the second rainfall forecast data, and the preset cloud-based database, the obtaining module is further configured to:

[0042] Acquire optical image data and point cloud data of the target scene area;

[0043] Using the preset target detection algorithm, the image elements of the target object in the optical image data are detected to obtain processed optical image data. Then, using the K-nearest neighbor sampling and sparse graph convolution algorithm, the point cloud data and the optical image data are geo-registered, and the geographical coordinates of the point cloud data are assigned to the target object in the optical image data to obtain registered optical image data and registered point cloud data.

[0044] The registered optical image data is mapped to the image feature space using a first preset neural network, and the registered point cloud data is mapped to the point cloud feature space using a second preset neural network.

[0045] Using a pre-defined feature mapping network, the image feature space and the point cloud feature space are aligned to obtain the target feature vector, and a target segmentation model is trained based on the target feature vector;

[0046] The geometric boundary information of the target object is interpreted using the target segmentation model, and the geographic coordinate information and geometric size attributes of the target object are extracted. The geometric boundary information, the geographic coordinate information and the geometric size attributes are then updated to the preset cloud-based base database.

[0047] The active flood monitoring device proposed in this embodiment of the invention calculates a spatiotemporal risk tensor based on rainfall forecast data and a pre-set cloud-based database using a pre-set flood model. Based on the spatiotemporal risk tensor, road network distribution, and distribution of flood-prone locations, an objective function is constructed. The optimal path for each mobile platform is determined based on constraints and the objective function to control the movement of each platform to a pre-set risk monitoring point. A pre-set target detection algorithm is used to determine the water accumulation area at each risk monitoring point, and combined with a pre-set digital elevation map, the water accumulation depth at each risk monitoring point is determined. Based on the water accumulation depth at each risk monitoring point, an appropriate monitoring strategy is automatically selected to monitor flooding at each point. This solves the problems of weak sensing capabilities, limited coverage, and untimely operational response of single monitoring methods in related technologies, improving the efficiency and accuracy of flood disaster prevention.

[0048] To achieve the above objectives, a third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the active flood monitoring method as described in the above embodiments.

[0049] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the active flood monitoring method as described in the above embodiments.

[0050] To achieve the above objectives, a fifth aspect of the present invention provides a computer program product comprising a computer program that, when executed by a processor, is used to implement the active flood monitoring method as described in the above embodiments.

[0051] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0052] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0053] Figure 1 A flowchart of an active flood monitoring method provided according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of a multimodal data processing flow according to an embodiment of the present invention;

[0055] Figure 3 A block diagram of a preset mobile platform according to an embodiment of the present invention;

[0056] Figure 4 A flowchart of a preset mobile platform dynamic path planning algorithm according to an embodiment of the present invention;

[0057] Figure 5 This is a block diagram of an active flood monitoring device provided according to an embodiment of the present invention;

[0058] Figure 6 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0059] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0060] The active flood monitoring method, apparatus, electronic device, and storage medium according to embodiments of the present invention will be described below with reference to the accompanying drawings. First, the active flood monitoring method according to embodiments of the present invention will be described with reference to the accompanying drawings.

[0061] Figure 1 This is a flowchart of an embodiment of the active flood monitoring method of the present invention.

[0062] For example, such as Figure 1 As shown, this active flood monitoring method includes the following steps:

[0063] In step S101, the first and second rainfall forecast data of the area to be monitored are obtained. Based on the first and second rainfall forecast data and the preset cloud-based database, the spatiotemporal risk tensor is calculated using the preset flood model.

[0064] It is understandable that the first rainfall forecast data refers to the 0-2 hour short-term rainfall forecast based on statistical extrapolation from meteorological radar; the second rainfall forecast data refers to the 2-6 hour short-term rainfall forecast based on numerical model calculations. The pre-defined cloud-based base database is a real-time dynamic database built on a cloud platform, used to store and manage multimodal sensing data required for urban flood modeling, including geometric boundaries of urban infrastructure (such as bridge and culvert boundaries, road boundaries), geographic coordinates (including elevation), and geometric dimensions; the latest environmental sensing data (such as water depth and risk point information) collected in real time and wirelessly transmitted through sensors (such as LiDAR and action cameras) on mobile platforms; and static geographic information such as high-resolution digital elevation base maps and historical flood-prone point distribution. The spatiotemporal risk tensor is used to describe the probability of flooding and its potential severity within a specific time and spatial range.

[0065] Specifically, to monitor and issue early warnings of flood risk in a specific area, the first step is to acquire primary and secondary rainfall forecast data for that area. This data, typically provided by meteorological departments, contains predictions of rainfall over a future period. Next, combined with a pre-built cloud-based database, the primary and secondary rainfall forecast data can be used to drive a pre-defined flood model. Through comprehensive analysis and calculation, a spatiotemporal risk tensor reflecting road flooding and traffic disruptions can be obtained. This spatiotemporal risk tensor can represent each path node within the next 6 hours. At different times The predicted level of risk is of great reference value for formulating response measures and carrying out emergency responses.

[0066]

[0067] in, Path node At any moment The depth of the accumulated water can be calculated using a pre-set flood model; The median critical water depth at which the preset mobile platform stops is generally taken as 0.3 meters. That is, it is assumed that when the water depth on the path exceeds 0.3 meters, the preset mobile platform (the mobile platform equipped with sensors) cannot pass through. The speed attenuation coefficient of the preset mobile platform is generally set to 4.

[0068] To make it easier to understand, the following details how to obtain the preset cloud-based base database.

[0069] In some embodiments, before calculating the spatiotemporal risk tensor using a preset flood model based on first rainfall forecast data, second rainfall forecast data, and a preset cloud-based database, the method further includes: acquiring optical image data and point cloud data of the target scene area; using a preset target detection algorithm to detect image features of target objects in the optical image data to obtain processed optical image data; and using K-nearest neighbor sampling and sparse graph convolution algorithms to georegister the point cloud data with the optical image data, and assigning the geographical coordinates of the point cloud data to the target objects in the optical image data to obtain registered optical image data. Based on the registered point cloud data; using a first preset neural network to map the registered optical image data to the image feature space, and using a second preset neural network to map the registered point cloud data to the point cloud feature space; using a preset feature mapping network to align the image feature space and the point cloud feature space to obtain the target feature vector, and training a target segmentation model based on the target feature vector; using the target segmentation model to interpret the geometric boundary information of the target object, and extracting the geographic coordinate information and geometric size attributes of the target object, and updating the geometric boundary information, geographic coordinate information and geometric size attributes to a preset cloud-based base database.

[0070] Specifically, such as Figure 2As shown, embodiments of the present invention can utilize relevant technical means to acquire optical image data and point cloud data of the target scene area. For example, an optical camera mounted on a motion camera, a drone, or a satellite remote sensing device can be used to collect optical image data (i.e., high-resolution surface images) through RGB (Red, Green, Blue) or near-infrared bands. Simultaneously, a lidar can be used for three-dimensional spatial scanning to generate point cloud data with centimeter-level precision. After obtaining the optical image data and point cloud data, a preset target detection algorithm (such as YOLOv5 (YouOnly Look Once version 5) algorithm) can be used. The YOLOv5 network is used as the backbone network for image feature extraction. Through its unique Focus structure (used to efficiently extract the initial features of the image) and CSP (CrossStage Partial Network) module, multi-scale feature fusion of high-resolution optical images can be achieved. This process not only improves processing speed but also enhances the network's ability to capture details. Therefore, the YOLOv5 network can quickly and automatically detect and label target objects in optical image data, i.e., various basic features, including bridge and culvert boundary lines, road boundary lines, storm drains, and river outlets, thus obtaining processed optical image data. Simultaneously, K-nearest neighbor sampling and sparse graph convolution algorithms can be used to encode and label the geometric features of target objects from point cloud data, thus obtaining processed point cloud data. Georegistration is then performed on the processed optical image data and the processed point cloud data; that is, the data from different sensors (optical images and point cloud data) are aligned in a geographic coordinate system, thereby accurately locating the target to be analyzed and making the two types of data spatially comparable.

[0071] Next, the registered optical image data is encoded into the image feature space using a first pre-defined neural network (such as a convolutional neural network), and the registered point cloud data is encoded into the point cloud feature space using a second pre-defined neural network (such as a neural network). Then, a feature mapping network based on a lightweight transformer network (Transformer) (i.e., the pre-defined feature mapping network) is constructed, and this pre-defined feature mapping network is used to align the image feature space and the point cloud feature space, forming a feature vector (i.e., the target feature vector) based on a unified hidden layer space representation. Based on this, a target segmentation model based on the unified hidden layer space representation can be trained. This model can quickly interpret the geometric boundaries of the target object and extract information such as the target object's geographic coordinates (including elevation) and geometric dimensions. Finally, through wireless transmission technology, the latest identified information parameters are synchronously updated to a pre-defined cloud-based base database, achieving real-time data updates and storage.

[0072] In step S102, an objective function is constructed based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points. Based on the preset constraints and the objective function, the optimal path of each preset mobile platform is determined. Based on the optimal path of each preset mobile platform, each preset mobile platform is controlled to move from the initial stopping point to the preset risk monitoring point.

[0073] It is understandable that the target area road network refers to the topological structure data of the road network within the area to be monitored, including road nodes (such as intersections and locations of key facilities), road segment connectivity (whether the road is passable), and dynamic traffic capacity (such as the actual traffic status after water accumulation). The distribution of historical flood-prone locations refers to areas with high incidence of urban flooding identified through historical monitoring data or disaster records.

[0074] Among them, such as Figure 3 As shown, the pre-set mobile platform refers to a type of autonomous monitoring vehicle (such as an intelligent monitoring platform based on a vehicle) that integrates multimodal sensors and intelligent decision-making systems. In this embodiment of the invention, the pre-set mobile platform is equipped with monitoring devices such as lidar, action cameras, drones, and level gauges (i.e., range-measuring water level gauges and pressure water level gauges). Simultaneously, the pre-set mobile platform also integrates a small bionic robot and a microcomputer system (including memory, processor, and communication interface). The microcomputer system is pre-loaded with autonomous driving electronic maps, high-resolution digital elevation base maps, and base map resources such as the distribution of historical flood-prone points in urban areas, used to assist the pre-set mobile platform in dynamic path planning, flood risk identification, and rapid collection and intelligent processing of terrain and water-related element information. The pre-set constraints are set based on an objective function, mainly constraining the access load, starting position, ending position, and multi-sensor collaboration of each pre-set mobile platform, which will be elaborated later.

[0075] Specifically, after obtaining the spatiotemporal risk tensor, an objective function can be constructed based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points. This objective function is determined by considering the path distance cost, risk-bearing cost, and energy efficiency of multiple pre-defined mobile platforms. Combined with pre-defined constraints, and using a dynamic flood optimization swarm algorithm based on reinforcement learning and the objective function, the pre-defined mobile platforms can collaborate to obtain the optimal path from their initial stopping point to the pre-defined risk monitoring point. Different pre-defined mobile platforms typically correspond to different optimal paths, and while each platform shares the same initial stopping point, their pre-defined risk monitoring points differ.

[0076] It should be noted that, before path planning, the embodiments of the present invention can also divide the area to be monitored into multiple evenly distributed grids, each grid representing a node, thereby constructing a grid map of the area to be monitored, making path planning more intuitive.

[0077] Optionally, in some embodiments, the objective function is:

[0078]

[0079] in, To monitor total cost, To preset the number of mobile platforms, , , All are target weights. As a balance factor, For path distance cost, To cover the costs of risk, For energy consumption costs, This is to reduce the cost of repetitive work by multiple machines.

[0080] The path cost can be:

[0081]

[0082] in, For nodes To the node distance, Indicates the default mobile platform Whether it is possible to traverse this path (depending on the connectivity between road segments and the risk level of the road segments) (joint control) For mobile platforms Total travel time The mission time limit is controlled by the energy storage of the mobile platform. This is the time penalty coefficient (the degree of impact of the additional path distance cost caused by the mobile platform needing to be retrieved by other vehicles due to energy depletion, which can be taken as 103 here). The total number of nodes. , For nodes.

[0083] The cost of triggering the risk can be:

[0084]

[0085]

[0086] in, The attenuation coefficient is... For the expected arrival node The maximum time, For this moment, This is a time decay factor. By introducing a time decay factor... (The closer the predicted risk is to the current time, the higher the risk weight is) The higher the risk level, the more the route planning will focus on the upcoming high-risk sections.

[0087] Energy consumption costs can be:

[0088]

[0089] in, This refers to the vehicle's driving power. For the power of the sensing system, For battery energy conversion efficiency, For mobile platforms Driving distance.

[0090] The calculation formula is:

[0091]

[0092] in, The coefficient of friction, For the quality of the vehicle, It is the acceleration due to gravity. For the speed of the vehicle, The ground inclination angle, The air drag coefficient, air density, This refers to the frontal area of ​​the car.

[0093] The cost of multiple machines performing repetitive tasks can be:

[0094]

[0095] in, For mobile platforms The area under responsibility for monitoring For mobile platforms The area under responsibility for monitoring This is the penalty coefficient for overlapping areas (which can be set to 2 here). By introducing this cost, we guide the pre-set mobile platforms to minimize overlapping monitoring, thereby effectively improving monitoring coverage.

[0096] Optionally, the preset constraints may include several of the following: the preset mobile platform is allowed to access each target location node once; the preset mobile platform's starting position is located at a given docking center; the preset mobile platform's travel time satisfies the following formula:

[0097]

[0098] in, For the preset mobile platform arrival node Time, For the preset mobile platform arrival node Time, To pre-set the mobile platform at the node The length of stay For preset mobile platform slave nodes To the node Travel time.

[0099] Furthermore, this embodiment of the invention employs a dynamic flood optimization population reinforcement learning algorithm to solve the above-mentioned problem of minimizing the total monitoring cost. For optimization problems with objectives, such as Figure 4 As shown, the process may include the following steps: Initializing with a randomly generated path on a grid map by a preset mobile platform, iterative updates are driven by navigation behavior, dynamic weight adjustments are used to calculate fitness values, and a solution to the objective function is selected based on the fitness values. When the population meets the convergence condition, the optimal paths for multiple preset mobile platforms to move from the initial docking point to the preset risk monitoring point are output. The convergence condition includes reaching a preset number of iterations, fitness values ​​being less than a preset threshold, or emergency termination when the flood risk value of all paths exceeds the threshold. The preset number of iterations or the preset threshold can be set according to actual needs and are not specifically limited here.

[0100] Meanwhile, during the solution process, the dynamic flood optimization population reinforcement learning algorithm establishes a navigation mechanism of "global search + local optimization + population evolution" to optimize the path scheme:

[0101] (1) Refined initial path trimming: Randomly generate path schemes, refine the path trimming to ensure that the initial solution meets the basic constraints, and calculate each fitness value, i.e. the total cost of the path.

[0102] (2) Calculate fitness value: Calculate the total travel cost of each path and select the best solution based on the fitness value; retain the best solution and the top 30% of the best solutions as the population basis for the next generation.

[0103] (3) Dynamic multi-objective trade-off: The dynamic flood optimization algorithm can effectively escape local optima, dynamically trade off multiple objectives, find the global optimal path, and dynamically adjust the weights according to real-time data.

[0104] (4) Adaptive mutation: For high-risk or high-energy-consumption nodes in the path, replace them with adjacent safe grids according to probability.

[0105] It should be noted that in determining the optimal path for each preset mobile platform, the navigation simulation method can include at least one of global search, local optimization, and population evolution. Global search involves applying large-scale perturbations to each path to break through local optima and discover potential high-quality areas. Local optimization involves fine-tuning the path. Population evolution involves selecting the top 30% of high-quality paths based on fitness values ​​as parents, cutting and splicing the two paths at random nodes through crossover and recombination, and replacing high-risk or high-energy-consumption nodes in the path with adjacent safe grids based on probability through adaptive mutation.

[0106] Understandably, compared to existing particle swarm optimization algorithms and the classic dung beetle algorithm, the dynamic flood optimization swarm reinforcement learning algorithm has a lower overall monitoring cost. The internal system considers factors such as path distance, flood risk, energy efficiency, and balanced multi-vehicle collaborative operation, and uses dynamically adjustable weights. This enables adaptive allocation of real-time priorities for multiple objectives, effectively addressing the limitations of particle swarm optimization (where fixed weights in single-objective optimization make it difficult to cope with dynamic multi-objectives, such as the real-time balancing needs of distance, risk, and energy consumption, and it is prone to getting trapped in local optima) and the classic dung beetle algorithm (which is based on fixed behavioral rules, such as single-objective optimization logic like rolling a ball, foraging, and reproduction, and lacks multi-objective collaboration and real-time decision-making capabilities).

[0107] Furthermore, the dynamic flood optimization swarm reinforcement learning algorithm establishes a navigation mechanism of "global search + local optimization + population evolution," applying large-scale perturbations to the movement path to escape local optima and replacing high-risk or high-energy-consumption nodes with adjacent safe grids according to probability, achieving adaptive mutation. In addition, by randomly cutting and splicing possible paths, combined with flood risk projections, it avoids getting stuck in high-risk areas, effectively overcoming the limitations of ant colony algorithms (low initial search rate, making it difficult to update paths in real time during high-risk, dynamically changing flood conditions) and classic dung beetle algorithms (fixed behavior rules under single-objective optimization conditions cannot cope with sudden situations such as path failures, such as a sudden increase in water depth at a node).

[0108] In step S103, based on a preset target detection algorithm, the water accumulation area of ​​each preset risk monitoring point is determined using each preset mobile platform, and the water accumulation depth of each preset risk monitoring point is determined based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map.

[0109] Specifically, once the preset mobile platform reaches the preset risk monitoring point, it can accurately identify the water accumulation area (water accumulation range) of the preset risk monitoring point from the optical image using a preset target detection algorithm. Combined with the preset digital elevation base map (i.e., the high-resolution digital elevation base map pre-set in the microcomputer system of the preset mobile platform), it can accurately calculate the water accumulation depth of the preset risk monitoring point.

[0110] In step S104, a monitoring strategy for each preset risk monitoring point is determined based on the water depth of each preset risk monitoring point, and flood monitoring is carried out on each preset risk monitoring point based on the monitoring strategy of each preset risk monitoring point to obtain flood monitoring results.

[0111] In other words, after obtaining the water depth at each preset risk monitoring point, corresponding monitoring strategies can be determined based on this depth information to address water depths at different levels, ensuring timely and effective risk warnings and management. Based on the monitoring strategy corresponding to each preset risk monitoring point, flood monitoring work can be carried out at the respective preset risk monitoring points, thereby obtaining the corresponding flood monitoring results.

[0112] As one possible implementation method, in some embodiments, a monitoring strategy for each preset risk monitoring point is determined based on the water depth of each preset risk monitoring point, including: based on each preset risk monitoring point, selecting at least one first target risk monitoring point whose water depth is less than a preset depth value; controlling a preset mobile platform corresponding to at least one first target risk monitoring point to move to the target location corresponding to each first target risk monitoring point, and using a distance-measuring water level gauge to measure the water depth of the target location corresponding to each first target risk monitoring point.

[0113] The preset depth value can be pre-set by researchers in the field, obtained through a limited number of experiments, or obtained through a limited number of computer simulations. No specific limitation is made here. Preferably, in this embodiment of the invention, the preset depth value is set to 0.15m.

[0114] Specifically, for the first target risk monitoring point among the preset risk monitoring points where the water depth is less than the preset depth value (e.g., 0.15m), the preset mobile platform corresponding to each first target risk monitoring point can be controlled to travel to the target point (a specific location in the water accumulation area, used to collect water depth information) corresponding to each first target risk monitoring point, and a distance measuring water level gauge can be used to further measure the actual water depth.

[0115] As one possible implementation, in some other embodiments, the monitoring strategy for each preset risk monitoring point is determined based on the water depth of each preset risk monitoring point. This further includes: based on each preset risk monitoring point, selecting at least one second target risk monitoring point whose water depth is greater than or equal to a preset depth value; controlling a preset mobile platform corresponding to at least one second target risk monitoring point to call a preset bionic robot carrying a pressure level gauge to move to the target point corresponding to each second target risk monitoring point, and using the pressure level gauge to measure the water depth at the target point corresponding to each second target risk monitoring point.

[0116] Specifically, for the second target risk monitoring point among the preset risk monitoring points where the water depth is greater than or equal to the preset depth value (e.g., 0.15m), the preset mobile platform corresponding to each second target risk monitoring point can be controlled to call the preset bionic robot (i.e., the small bionic robot integrated into the preset mobile platform), organize the preset bionic robot to carry the pressure level gauge, and move to the target point corresponding to each second target risk monitoring point in a semi-submersible mode. Then, the pressure level gauge is released to measure the water depth at the target point corresponding to each second target risk monitoring point.

[0117] In summary, the active flood monitoring method proposed in this invention has at least the following beneficial effects:

[0118] (1) The embodiments of the present invention can realize active perception and targeted monitoring of urban floods, improve the timeliness and support of flood monitoring, and serve the construction of livable, resilient and smart cities.

[0119] (2) In this invention, an objective function is constructed based on the travel path distance, flood risk cost, and energy efficiency of multiple preset mobile platforms. Multiple constraints are set for the objective function to obtain the actual path planning results. Based on a dynamic flood optimization population reinforcement learning algorithm, path planning is performed on the multiple preset mobile platforms to obtain the target path for the preset mobile platforms to move from the initial stopping point to the preset risk monitoring point. This algorithm can effectively escape local optima, dynamically weigh multiple objectives, find the globally optimal path, and dynamically adjust weights based on real-time data. This not only improves the accuracy and efficiency of optimal path search but also ensures the practical feasibility and dynamic adaptability of the path scheme, guaranteeing its adaptability to complex urban road network environments and demonstrating high applicability.

[0120] The active flood monitoring method proposed in this invention calculates a spatiotemporal risk tensor based on rainfall forecast data and a pre-set cloud-based database using a pre-set flood model. An objective function is constructed based on the spatiotemporal risk tensor, regional road network, and distribution of flood-prone areas. The optimal path for each mobile platform is determined based on constraints and the objective function to control the movement of each platform to a pre-set risk monitoring point. A pre-set target detection algorithm is used to determine the water accumulation area at each risk monitoring point, and combined with a pre-set digital elevation map, the water accumulation depth at each risk monitoring point is determined. Based on the water accumulation depth at each risk monitoring point, an appropriate monitoring strategy is automatically selected to monitor flooding at each point. This solves the problems of weak sensing capabilities, limited coverage, and untimely operational response of single monitoring methods in related technologies, improving the efficiency and accuracy of flood disaster prevention.

[0121] Next, the active flood monitoring device according to an embodiment of the present invention is described with reference to the accompanying drawings.

[0122] Figure 5 This is a block diagram of an active flood monitoring device according to an embodiment of the present invention.

[0123] like Figure 5 As shown, the active flood monitoring device 10 includes: an acquisition module 100, a control module 200, a determination module 300, and a monitoring module 400.

[0124] The module 100 is used to acquire the first and second rainfall forecast data of the area to be monitored. Based on the first and second rainfall forecast data and the preset cloud-based database, the spatiotemporal risk tensor is calculated using the preset flood model.

[0125] The control module 200 is used to construct an objective function based on the spatiotemporal risk tensor, the road network of the target area and the distribution of historical flood-prone points, and to determine the optimal path of each preset mobile platform based on preset constraints and the objective function. Based on the optimal path of each preset mobile platform, it controls each preset mobile platform to move from the initial stopping point to the preset risk monitoring point.

[0126] The determination module 300 is used to determine the water accumulation area of ​​each preset risk monitoring point based on a preset target detection algorithm and each preset mobile platform, and to determine the water accumulation depth of each preset risk monitoring point based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map.

[0127] The monitoring module 400 is used to determine the monitoring strategy for each preset risk monitoring point based on the water depth of each preset risk monitoring point, and to conduct flood monitoring for each preset risk monitoring point based on the monitoring strategy of each preset risk monitoring point, so as to obtain the flood monitoring results.

[0128] Optionally, in some embodiments, the monitoring module 400 is specifically used for:

[0129] Based on each preset risk monitoring point, at least one first target risk monitoring point is selected where the water depth at the monitoring point is less than the preset depth value.

[0130] Control at least one preset mobile platform corresponding to a first target risk monitoring point to move to the target location corresponding to each first target risk monitoring point, and use a distance measuring water level gauge to measure the water depth at the target location corresponding to each first target risk monitoring point.

[0131] Optionally, in some embodiments, the monitoring module 400 is further configured to:

[0132] Based on each preset risk monitoring point, at least one second target risk monitoring point is selected, whose water depth is greater than or equal to the preset depth value.

[0133] Control at least one preset mobile platform corresponding to a second target risk monitoring point to call a preset bionic robot carrying a pressure level gauge to move to the target point corresponding to each second target risk monitoring point, and use the pressure level gauge to measure the water depth at the target point corresponding to each second target risk monitoring point.

[0134] Optionally, in some embodiments, the objective function is:

[0135]

[0136] in, To monitor total cost, To preset the number of mobile platforms, , , All are target weights. As a balance factor, For path distance cost, To cover the costs of risk, For energy consumption costs, This is to reduce the cost of repetitive work by multiple machines.

[0137] Optionally, in some embodiments, before calculating the spatiotemporal risk tensor using a preset flood model based on the first rainfall forecast data, the second rainfall forecast data, and a preset cloud-based database, the obtaining module 100 is further configured to:

[0138] Acquire optical image data and point cloud data of the target scene area;

[0139] Using a pre-defined target detection algorithm, the image elements of the target object in the optical image data are detected to obtain the processed optical image data. Then, using K-nearest neighbor sampling and sparse graph convolution algorithms, the point cloud data and the optical image data are geo-registered, and the geographical coordinates of the point cloud data are assigned to the target object in the optical image data to obtain the registered optical image data and the registered point cloud data.

[0140] The registered optical image data is mapped to the image feature space using a first preset neural network, and the registered point cloud data is mapped to the point cloud feature space using a second preset neural network.

[0141] Using a pre-defined feature mapping network, the image feature space and point cloud feature space are aligned to obtain the target feature vector, and a target segmentation model is trained based on the target feature vector;

[0142] The geometric boundary information of the target object is interpreted using the target segmentation model, and the geographic coordinate information and geometric size attributes of the target object are extracted. The geometric boundary information, geographic coordinate information and geometric size attributes are then updated to the preset cloud-based base database.

[0143] It should be noted that the foregoing explanation of the active flood monitoring method embodiment also applies to the active flood monitoring device of this embodiment, and will not be repeated here.

[0144] The active flood monitoring device proposed in this embodiment of the invention calculates a spatiotemporal risk tensor based on rainfall forecast data and a pre-set cloud-based database using a pre-set flood model. Based on the spatiotemporal risk tensor, regional road network, and distribution of flood-prone areas, an objective function is constructed. The optimal path for each mobile platform is determined based on constraints and the objective function to control the movement of each platform to a pre-set risk monitoring point. A pre-set target detection algorithm is used to determine the water accumulation area at each risk monitoring point, and combined with a pre-set digital elevation map, the water accumulation depth at each risk monitoring point is determined. Based on the water accumulation depth at each risk monitoring point, an appropriate monitoring strategy is automatically selected to monitor flooding at each risk monitoring point. This solves the problems of weak sensing capabilities, limited coverage, and untimely operational response of single monitoring methods in related technologies, improving the efficiency and accuracy of flood disaster prevention.

[0145] Figure 6 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. The electronic device may include:

[0146] The memory 601, the processor 602, and the computer program stored on the memory 601 and capable of running on the processor 602.

[0147] When the processor 602 executes the program, it implements the active flood monitoring method provided in the above embodiments.

[0148] Furthermore, the electronic device further includes:

[0149] Communication interface 603 is used for communication between memory 601 and processor 602.

[0150] The memory 601 is used to store computer programs that can run on the processor 602.

[0151] The memory 601 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0152] If the memory 601, processor 602, and communication interface 603 are implemented independently, then the communication interface 603, memory 601, and processor 602 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0153] Optionally, in a specific implementation, if the memory 601, processor 602, and communication interface 603 are integrated on a single chip, then the memory 601, processor 602, and communication interface 603 can communicate with each other through an internal interface.

[0154] Processor 602 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention.

[0155] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described active flood monitoring method.

[0156] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described active flood monitoring method.

[0157] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0158] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0159] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An active flood monitoring method, characterized in that, Includes the following steps: First and second rainfall forecast data for the area to be monitored are obtained. Based on the first and second rainfall forecast data and the preset cloud-based database, the spatiotemporal risk tensor is calculated using a preset flood model. Based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points, an objective function is constructed. Based on the preset constraints and the objective function, the optimal path of each preset mobile platform is determined. Based on the optimal path of each preset mobile platform, each preset mobile platform is controlled to move from the initial stopping point to the preset risk monitoring point. Based on a preset target detection algorithm, the water accumulation area of ​​each preset risk monitoring point is determined using each preset mobile platform, and the water accumulation depth of each preset risk monitoring point is determined based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map. Based on the water depth at each preset risk monitoring point, a monitoring strategy is determined for each preset risk monitoring point. Based on the monitoring strategy for each preset risk monitoring point, flood monitoring is carried out at each preset risk monitoring point to obtain flood monitoring results.

2. The active flood monitoring method according to claim 1, characterized in that, The step of determining the monitoring strategy for each preset risk monitoring point based on the water depth at each preset risk monitoring point includes: Based on each preset risk monitoring point, at least one first target risk monitoring point is selected where the water depth at the monitoring point is less than the preset depth value. The preset mobile platform corresponding to at least one first target risk monitoring point is controlled to move to the target location corresponding to each first target risk monitoring point, and the water depth at the target location corresponding to each first target risk monitoring point is measured using a distance measuring water level gauge.

3. The active flood monitoring method according to claim 1, characterized in that, The step of determining the monitoring strategy for each preset risk monitoring point based on the water depth at each preset risk monitoring point further includes: Based on each preset risk monitoring point, at least one second target risk monitoring point is selected where the water depth at the monitoring point is greater than or equal to the preset depth value. The system controls a preset mobile platform corresponding to at least one second target risk monitoring point to call a preset bionic robot carrying a pressure level gauge to move to the target point corresponding to each second target risk monitoring point, and uses the pressure level gauge to measure the water depth at the target point corresponding to each second target risk monitoring point.

4. The active flood monitoring method according to claim 1, characterized in that, The objective function is: in, To monitor total cost, The number of the preset mobile platforms, , , All are target weights. As a balance factor, For path distance cost, To cover the costs of risk, For energy consumption costs, This is to reduce the cost of repetitive work by multiple machines.

5. The active flood monitoring method according to claim 1, characterized in that, Before calculating the spatiotemporal risk tensor using the preset flood model based on the first rainfall forecast data, the second rainfall forecast data, and the preset cloud-based database, the process further includes: Acquire optical image data and point cloud data of the target scene area; Using the preset target detection algorithm, the image elements of the target object in the optical image data are detected to obtain processed optical image data. Then, using the K-nearest neighbor sampling and sparse graph convolution algorithm, the point cloud data and the optical image data are geo-registered, and the geographical coordinates of the point cloud data are assigned to the target object in the optical image data to obtain registered optical image data and registered point cloud data. The registered optical image data is mapped to the image feature space using a first preset neural network, and the registered point cloud data is mapped to the point cloud feature space using a second preset neural network. Using a pre-defined feature mapping network, the image feature space and the point cloud feature space are aligned to obtain the target feature vector, and a target segmentation model is trained based on the target feature vector; The geometric boundary information of the target object is interpreted using the target segmentation model, and the geographic coordinate information and geometric size attributes of the target object are extracted. The geometric boundary information, the geographic coordinate information and the geometric size attributes are then updated to the preset cloud-based base database.

6. An active flood monitoring device, characterized in that, include: The acquisition module is used to acquire the first and second rainfall forecast data of the area to be monitored, and to calculate the spatiotemporal risk tensor based on the first rainfall forecast data, the second rainfall forecast data and the preset cloud-based database using a preset flood model. The control module is used to construct an objective function based on the spatiotemporal risk tensor, the road network of the target area, and the distribution of historical flood-prone points, and to determine the optimal path of each preset mobile platform based on preset constraints and the objective function, and to control each preset mobile platform to move from the initial stop point to the preset risk monitoring point based on the optimal path of each preset mobile platform. The determination module is used to determine the water accumulation area of ​​each preset risk monitoring point based on a preset target detection algorithm and each preset mobile platform, and to determine the water accumulation depth of each preset risk monitoring point based on the water accumulation area of ​​each preset risk monitoring point and a preset digital elevation base map; The monitoring module is used to determine the monitoring strategy for each preset risk monitoring point based on the water depth of each preset risk monitoring point, and to perform flood monitoring on each preset risk monitoring point based on the monitoring strategy of each preset risk monitoring point, so as to obtain the flood monitoring results.

7. The active flood monitoring device according to claim 6, characterized in that, The monitoring module is used for: Based on each preset risk monitoring point, at least one first target risk monitoring point is selected where the water depth at the monitoring point is less than the preset depth value. The preset mobile platform corresponding to at least one first target risk monitoring point is controlled to move to the target location corresponding to each first target risk monitoring point, and the water depth at the target location corresponding to each first target risk monitoring point is measured using a distance measuring water level gauge.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the active flood monitoring method as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the active flood monitoring method as described in any one of claims 1-5.

10. A computer program product, characterized in that, The method includes a computer program, which, when executed by a processor, is used to implement the active flood monitoring method according to any one of claims 1-5.

Images