A water and rainfall condition emergency monitoring method and system based on a Beidou inspection robot dog

Abstract

The application discloses a water and rain condition emergency monitoring method and system based on a Beidou inspection robot dog, relates to the technical field of water and rain condition monitoring, and comprises the following steps: acquiring environment point cloud data and positioning data of the inspection robot dog; generating a three-dimensional terrain model according to the environment point cloud data; constructing a three-dimensional topological constraint model based on the three-dimensional terrain model, and combining the three-dimensional topological constraint model and the positioning data to perform abnormal correction to obtain a corrected three-dimensional terrain model; establishing a motion feasibility matrix according to the corrected three-dimensional terrain model, and planning a path for the inspection robot dog based on the motion feasibility matrix; acquiring historical water and rain monitoring data, and collecting real-time water and rain monitoring data of a target area in real time by using the inspection robot dog; constructing a dynamic risk evaluation model by combining the real-time water and rain monitoring data and the historical water and rain monitoring data, and generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model. The application can effectively improve the response efficiency of water and rain condition emergency monitoring.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of water regime monitoring, and in particular to a water regime emergency monitoring method and system based on a Beidou inspection robot dog. BACKGROUND

[0002] Water regime monitoring is a key link for flood control, drought resistance, water resource management, and geological disaster prevention. The core goal is to realize real-time sensing, accurate positioning, and intelligent early warning of parameters such as water level, flow rate, and dam structure hidden dangers under complex meteorological and geographical conditions. Accurate and timely water regime data can effectively reduce the loss caused by natural disasters and ensure the stable development of social economy.

[0003] Traditional water regime monitoring mainly relies on static monitoring means such as fixed sites (such as water level gauges and rainfall stations), but there are significant monitoring blind spots in key areas such as the backwater slope of the dam. Due to the limited distribution density and fixed position of the fixed sites, it is difficult to effectively cover hidden dangers such as piping and seepage, and the monitoring data often cannot accurately reflect the real changes of local water flow. In addition, the mountain torrents caused by heavy rain can carry a large amount of sand and stones, which not only easily destroy fixed monitoring equipment, but also damage wired communication networks and power supply systems, leading to inaccurate monitoring of water regime monitoring data, making it difficult for flood control departments to obtain accurate water regime information in a timely manner, and seriously affecting the efficiency of emergency response.

[0004] At present, there is no good solution to the above problems. SUMMARY

[0005] Embodiments of the present application provide a water regime emergency monitoring method and system based on a Beidou inspection robot dog, which is used to improve the response efficiency of water regime emergency monitoring.

[0006] To achieve the above purpose, embodiments of the present application adopt the following technical solutions:

[0007] In a first aspect, a water regime emergency monitoring method based on a Beidou inspection robot dog is provided, which comprises:

[0008] Collecting environmental point cloud data in a target area using a preset laser radar, and obtaining positioning data of an inspection robot dog in the target area through a Beidou satellite system;

[0009] Generating a three-dimensional terrain model using an iterative closest point algorithm and according to the environmental point cloud data;

[0010] Constructing a three-dimensional topological constraint model based on the three-dimensional terrain model, and combining the three-dimensional topological constraint model and the positioning data to perform abnormal correction on the three-dimensional terrain model to obtain a corrected three-dimensional terrain model;

[0011] establish a motion feasibility matrix of the target area according to the modified three-dimensional terrain model, and perform path planning on the inspection robot dog based on the motion feasibility matrix;

[0012] acquire historical water and rain monitoring data of the target area, and collect real-time water and rain monitoring data and soil state data of the target area by using the inspection robot dog;

[0013] construct a dynamic risk evaluation model by combining the real-time water and rain monitoring data, the soil state data, and the historical water and rain monitoring data, and generate a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

[0014] In a possible implementation manner of the first aspect, the constructing a three-dimensional topological constraint model based on the three-dimensional terrain model comprises the following steps:

[0015] dividing the three-dimensional terrain model into a plurality of three-dimensional grid models, and extracting slope characteristic parameters and curvature characteristic parameters of all the three-dimensional grid models;

[0016] calculating average slope parameters and average curvature parameters of the three-dimensional terrain model based on the slope characteristic parameters and the curvature characteristic parameters, wherein the three-dimensional grid model is a regular grid model;

[0017] taking the average slope parameters and the average curvature parameters as constraint conditions, traversing each three-dimensional grid model, and dividing terrain features of each three-dimensional grid model according to the constraint conditions;

[0018] establishing topological connection rules of the three-dimensional terrain model based on the terrain features, and constructing a topological graph by using a graph theory algorithm;

[0019] constructing a three-dimensional topological constraint model by combining the topological graph, the constraint conditions, and the three-dimensional terrain model.

[0020] In a possible implementation manner of the first aspect, the modifying the three-dimensional terrain model based on the three-dimensional topological constraint model and the positioning data to obtain a modified three-dimensional terrain model comprises the following steps:

[0021] converting a coordinate system of the positioning data from a world coordinate system to a grid coordinate system of the three-dimensional grid model;

[0022] mapping the positioning data after coordinate system conversion and the three-dimensional grid model to a two-dimensional plane grid, and determining historical positioning points in a plurality of plane grids in the plane grid according to historical movement tracks in the positioning data;

[0023] calculating an elevation value corresponding to each of the historical positioning points in each of the planar grids by using a bilinear interpolation algorithm;

[0024] calculating a curvature and a slope corresponding to each of the three-dimensional grid models based on each of the elevation values;

[0025] determining an anomaly type in combination with the curvature, the slope and the constraint condition, and performing anomaly correction on the historical positioning points according to the anomaly type to obtain a corrected three-dimensional terrain model.

[0026] In a possible implementation manner of the first aspect, the establishing of the motion feasibility matrix of the target area according to the corrected three-dimensional terrain model, and the path planning for the inspection robot dog based on the motion feasibility matrix include the following steps:

[0027] calculating an average grid slope and an average grid curvature corresponding to each of the three-dimensional grid models based on the slope feature parameter and the curvature feature parameter in each of the three-dimensional grid models, respectively;

[0028] judging motion feasibility of the inspection robot dog in each of the three-dimensional grid models in combination with the average grid slope and the average grid curvature, respectively, and determining a feasibility value of each of the three-dimensional grid models;

[0029] constructing the motion feasibility matrix of the target area based on each of the feasibility values;

[0030] performing path planning for the inspection robot dog according to the motion feasibility matrix by using Dijkstra algorithm.

[0031] In a possible implementation manner of the first aspect, the path planning for the inspection robot dog according to the motion feasibility matrix by using Dijkstra algorithm includes the following steps:

[0032] transforming the motion feasibility matrix into a graph structure, wherein each matrix element in the motion feasibility matrix is a node in the graph structure;

[0033] determining a connection relationship of each of the nodes according to the feasibility value in the motion feasibility matrix;

[0034] taking an initial position of the inspection robot dog as a starting node, and counting a shortest distance from the starting node to each of the nodes;

[0035] arranging all the shortest distances in ascending order to determine a priority queue of the starting node;

[0036] determining an expansion node of the starting node according to the connection relationship and the priority queue, and counting a priority queue corresponding to the expansion node;

[0037] determining a target expansion node of the expansion node according to the connection relationship and a priority queue corresponding to the expansion node;

[0038] repeating the above steps until all the nodes are traversed, and recording the predecessor nodes of each of the nodes;

[0039] backtracking from the predecessor nodes to the starting node to obtain a shortest path of the inspection robot dog.

[0040] In a possible implementation manner of the first aspect, the constructing a dynamic risk evaluation model according to the real-time water and rain monitoring data, the soil state data and the historical water and rain monitoring data, and generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model comprises the following steps:

[0041] selecting an easy-occurrence index and a disaster-causing index according to the historical water and rain monitoring data by using a correlation analysis algorithm, and determining a rainfall factor;

[0042] performing similarity calculation on the real-time water and rain monitoring data, the easy-occurrence index and the disaster-causing index to determine a target region as an easy-disaster-occurrence area;

[0043] determining a soil state based on the soil state data and by using a preset soil state analysis model;

[0044] calculating membership degrees of the soil state, the easy-disaster-occurrence area and the rainfall factor respectively by using a preset membership function;

[0045] calculating weight values of the soil state, the easy-disaster-occurrence area and the rainfall factor respectively according to the membership degrees of the soil state, the easy-disaster-occurrence area and the rainfall factor by using an analytic hierarchy process;

[0046] integrating and constructing a dynamic risk evaluation model by using fuzzy mathematical operation according to the weight values of the soil state, the easy-disaster-occurrence area and the rainfall factor;

[0047] generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

[0048] In a possible implementation manner of the first aspect, the generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model comprises the following steps:

[0049] for any one of the three-dimensional grid models, generating a risk value corresponding to the three-dimensional grid model by using the dynamic risk evaluation model;

[0050] generate a risk surface of the three-dimensional terrain model by spatial interpolation on each of the risk values through a Kriging method;

[0051] divide the risk surface according to a preset risk level threshold to determine a regional dynamic risk early warning area.

[0052] In a possible implementation manner of the first aspect, the generating the risk surface of the three-dimensional terrain model by spatial interpolation on each of the risk values through the Kriging method comprises the following steps:

[0053] determining a spatial coordinate of a grid center point of the three-dimensional grid model where each of the risk values is located;

[0054] calculating a semi-variogram value of each of the risk values according to the risk value and the spatial coordinate by using a preset empirical variogram function;

[0055] fitting the semi-variogram value according to a preset theoretical variogram function model and a least square method to determine a spatial distribution characteristic of the risk value;

[0056] determining a search neighborhood range according to the spatial distribution characteristic, and determining all interpolation points in the search neighborhood range by using a K-neighbor algorithm;

[0057] calculating a spatial distance between each of the interpolation points based on each of the interpolation points;

[0058] calculating a weight value between each of the interpolation points by using a particle swarm optimization algorithm according to the semi-variogram value and the spatial distance;

[0059] performing weighted summation on all the weight values to obtain an estimated risk value of the interpolation point;

[0060] performing spatial interpolation based on the estimated risk value to obtain the risk surface of the three-dimensional terrain model.

[0061] In a second aspect, the present application provides a machine readable storage medium, the machine readable storage medium storing instructions for causing a machine to execute the above-mentioned water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog.

[0062] In a third aspect, the present application provides a water and rainfall condition emergency monitoring system based on the Beidou inspection robot dog, comprising:

[0063] a memory configured to store instructions; and

[0064] a processor configured to call the instructions from the memory and capable of implementing the above-mentioned water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog when executing the instructions.

[0065] Through the technical solution, by fusing the laser radar point cloud data and the Beidou centimeter-level positioning, the micro-terrain features of the target area can be accurately described, and the terrain model is dynamically updated in real time, providing a reliable basis for subsequent path planning and risk assessment. In combination with the positioning data of the robot dog and the three-dimensional topological constraint model, the terrain model is dynamically corrected, which can eliminate terrain distortion caused by environmental changes and ensure the real-time and accuracy of the model. Based on the corrected three-dimensional terrain model, a motion feasibility matrix is established to plan the optimal path for the robot dog, so that it can autonomously avoid obstacles and choose the safest and fastest route to reach the monitoring point, significantly improving the maneuverability of the robot dog in complex environments, which can shorten the emergency response time and improve the monitoring and emergency response efficiency. In combination with the real-time water and rainfall data and the three-dimensional terrain model, a dynamic risk evaluation model including the time dimension is constructed, which can provide intuitive and accurate basis for emergency decision-making, effectively reducing disaster losses. In combination with the risk assessment results and the distribution of emergency resources, an emergency plan is automatically generated, effectively improving the efficiency and effectiveness of emergency response.

[0066] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific embodiments part. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 A flowchart of a water and rainfall condition emergency monitoring method based on a Beidou inspection robot dog provided by the embodiments of the present application;

[0068] Figure 2 A three-dimensional terrain model profile provided by the embodiments of the present application. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific embodiments described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present application.

[0070] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, motion condition, etc. between components in a certain specific posture (as shown in the drawings), and if the specific posture changes, the directional indications also change accordingly.

[0071] In addition, if the description of "first", "second" and the like is involved in the embodiments of the present application, the description of "first", "second" and the like is only for the purpose of description, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first", "second" can be explicitly or implicitly included at least one of the features. In addition, the technical solutions of various embodiments can be combined with each other, but it must be based on the realization of ordinary skilled in the art, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist, nor in the protection scope claimed by the present application.

[0072] Figure 1 The flowchart of the water and rain emergency monitoring method based on the Beidou inspection robot dog according to the embodiments of the present application is schematically shown. As shown in Figure 1 The embodiments of the present application provide a water and rain emergency monitoring method based on the Beidou inspection robot dog, which can include the following steps.

[0073] S110, collecting the environmental point cloud data in the target area by using the preset laser radar, and obtaining the positioning data of the inspection robot dog in the target area through the Beidou satellite system;

[0074] S120, generating a three-dimensional terrain model according to the environmental point cloud data by using the iterative closest point algorithm;

[0075] S130, constructing a three-dimensional topological constraint model based on the three-dimensional terrain model, and combining the three-dimensional topological constraint model and the positioning data to correct the three-dimensional terrain model to obtain a corrected three-dimensional terrain model;

[0076] S140, establishing a motion feasibility matrix of the target area according to the corrected three-dimensional terrain model, and planning a path for the inspection robot dog based on the motion feasibility matrix;

[0077] S150, obtaining the historical water and rain monitoring data of the target area, and collecting the real-time water and rain monitoring data and soil state data of the target area by using the inspection robot dog;

[0078] S160, constructing a dynamic risk evaluation model by combining the real-time water and rain monitoring data, the soil state data and the historical water and rain monitoring data, and generating a regional dynamic risk warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

[0079] The preset laser radar collects the environmental point cloud data in the target area, and the Beidou satellite system obtains the positioning data of the inspection robot dog in the target area. The laser radar is a technology that obtains distance information by emitting a laser beam and measuring its reflection time, which can quickly and accurately obtain the three-dimensional coordinate information of the target object. The point cloud data is a data set composed of a large number of three-dimensional coordinate points, representing the surface positions of various objects in the target area. Through laser radar scanning, detailed topography, landform and object distribution information in the target area can be obtained. The Beidou satellite navigation system is a global satellite navigation system that can provide high-precision positioning, navigation and timing services. Through the Beidou satellite system, the real-time position information of the inspection robot dog in the target area can be obtained, including longitude, latitude, height and other data. Laser radar and Beidou satellite system work together in this method, laser radar is responsible for obtaining environmental point cloud data, and Beidou satellite system is responsible for providing positioning data of the inspection robot dog. The combination of the two provides comprehensive support for the autonomous navigation and monitoring of the inspection robot dog in the target area.

[0080] Subsequently, a three-dimensional terrain model is generated from the environmental point cloud data using the iterative closest point algorithm. The laser radar emits a laser pulse to scan the target area, and after receiving the echo reflected by the surface of the object, a three-dimensional coordinate data set composed of a large number of discrete points is generated, each point containing X, Y, Z spatial coordinates. The iterative closest point algorithm is a point cloud registration algorithm used to align two sets of point cloud data by iteratively calculating the optimal transformation matrix between the two sets of points to minimize the distance error of the two sets of points. First, multiple frames of point cloud are registered, and ICP algorithm is used to register the environmental point cloud data to obtain high-precision registration results. The accumulated point cloud data is de-duplicated and smoothed to eliminate registration errors and inconsistencies in overlapping areas. Subsequently, Poisson reconstruction, moving cube and other algorithms are used to convert the point cloud data into a continuous three-dimensional mesh model, and then a three-dimensional terrain model is obtained.

[0081] Next, a three-dimensional topological constraint model is constructed based on the three-dimensional terrain model, and the three-dimensional terrain model is corrected based on the three-dimensional topological constraint model and the positioning data to obtain a corrected three-dimensional terrain model. The three-dimensional topological constraint model is a mathematical model of spatial relationships, geometric characteristics and physical rules in the three-dimensional terrain model, and is used to describe the logical consistency between various elements of the terrain (such as elevation, slope, and surface connectivity). Combined with the Beidou positioning data, abnormal data in the three-dimensional terrain model can be dynamically detected and corrected to generate a more reliable terrain representation. First, the geometric and physical laws in the three-dimensional terrain model are converted into calculable constraint conditions for detecting abnormalities. For example, features such as slope, slope direction, and elevation discontinuity in the terrain model can be extracted, and the allowed range of motion (such as a maximum climbing angle of 35°) can be defined. Second, the terrain features of the three-dimensional terrain model are extracted, and a topological graph is constructed based on the terrain features. The three-dimensional topological constraint model is obtained by combining the topological graph, the constraint conditions and the three-dimensional terrain model. After obtaining the three-dimensional topological constraint model, the positioning data of the inspection robot dog is obtained through the Beidou positioning, and the positioning data of the inspection robot dog is compared with the three-dimensional topological constraint model to identify the contradiction points, such as the model showing passable but actually unable to pass. The historical positioning points of the inspection robot dog are used to determine the curvature and slope corresponding to the three-dimensional grid model. According to the curvature and slope corresponding to the three-dimensional grid model and the constraint conditions of the three-dimensional topological constraint model, the type of abnormality is judged, and based on the detected type of abnormality, a targeted method is used to correct the terrain model to ensure its spatial consistency and dynamic accuracy.

[0082] According to the corrected three-dimensional terrain model, a motion feasibility matrix of the target area is established, and path planning for the inspection robot dog based on the motion feasibility matrix refers to generating a motion feasibility matrix through the corrected three-dimensional terrain model, and planning a safe, efficient and dynamically adaptive path based on the matrix, supporting unmanned task execution in complex environments. The three-dimensional terrain model is discretized into a two-dimensional or three-dimensional grid matrix, and each grid cell corresponds to a motion feasibility score, which is used to quantify the passability of the inspection robot dog in that area. Based on the motion feasibility matrix, the optimal path from the starting point to the end point is calculated through an algorithm. First, the corrected three-dimensional terrain model is projected onto a two-dimensional plane and divided into uniform grids. Terrain features (such as average slope and elevation change rate) are extracted for each grid cell. Combined with the performance parameters of the inspection robot dog (such as a maximum climbing angle of 35° and a wading depth of 30 cm), the feasibility score is calculated. Subsequently, based on the path planning of the motion feasibility matrix, the Rapidly-exploring Random Tree algorithm can be used to quickly generate a feasible path. The starting and ending grids can be determined according to the task requirements, and the path planning algorithm is used to search for a path in the motion feasibility matrix, calculate the total feasibility score of the path, and select the path with the highest score that meets the task requirements.

[0083] The historical water and rainfall monitoring data of the target area can be obtained by pre-set system data integration. The historical water and rainfall monitoring data refers to the records of hydrological and meteorological information such as water level and rainfall in the target area in the past period of time. Real-time water and rainfall monitoring data of the target area is collected by the inspection robot dog in real time. The inspection robot dog is a robot with autonomous movement and perception capabilities, which can carry various sensors and equipment for real-time collection of water and rainfall monitoring data in the target area. The water level meter, rain gauge and other sensors are carried on the inspection robot dog to monitor the water level and rainfall and other hydrological and meteorological information in the target area in real time. The inspection robot dog transmits the collected data to the monitoring center or data center in real time through wireless communication technology for further analysis and processing.

[0084] Next, a dynamic risk evaluation model is constructed based on real-time water and rainfall monitoring data and historical water and rainfall monitoring data, and a regional dynamic risk warning area is generated based on the dynamic risk evaluation model and the three-dimensional terrain model. A spatio-temporal continuous risk quantification model is constructed by real-time water and rainfall monitoring data and three-dimensional terrain features, and risk level areas (such as high risk / low risk) are dynamically divided, providing minute-level update accurate early warning for disaster prevention decision-making, significantly improving disaster response efficiency and resource allocation rationality. The relevant indexes of water and rainfall monitoring data are selected by historical water and rainfall monitoring data, and the membership degrees of the relevant indexes are calculated by using the membership function. Then the weight values of the relevant indexes are calculated by using the pre-set analytic hierarchy process, and the dynamic risk evaluation model is constructed by using fuzzy mathematical operation to integrate the weight values of the relevant indexes. After obtaining the dynamic risk evaluation model, a regional dynamic risk warning area is generated based on the dynamic risk evaluation model and the three-dimensional terrain model. The dynamic risk evaluation model is a mathematical model based on real-time data, historical data and environmental variables to dynamically evaluate potential risks in a specific area. The risk level data calculated by the dynamic risk model is superimposed with the spatial position data of the three-dimensional terrain model, and the target area is divided into different risk warning areas according to the risk level and terrain features to determine the regional dynamic risk warning area.

[0085] By fusing lidar point cloud data with Beidou centimeter-level positioning, the micro-topographic features of the target area can be accurately described, and the terrain model can be dynamically updated in real time, providing a reliable basis for subsequent path planning and risk assessment. By combining the positioning data of the robot dog with the three-dimensional topological constraint model, the terrain model can be dynamically corrected, which can eliminate terrain distortion caused by environmental changes and ensure the real-time and accuracy of the model. Based on the corrected three-dimensional terrain model, a motion feasibility matrix is established to plan the optimal path for the robot dog, enabling it to autonomously avoid obstacles and choose the safest and fastest route to reach the monitoring point, significantly improving the robot dog's mobility in complex environments, reducing emergency response time, and improving monitoring and emergency response efficiency. By combining real-time water and rainfall data with the three-dimensional terrain model, a dynamic risk assessment model is established, which can provide intuitive and accurate basis for emergency decision-making, effectively reducing disaster losses. By combining risk assessment results and emergency resource distribution, an emergency plan is automatically generated, effectively improving the efficiency and effectiveness of emergency response.

[0086] In one embodiment of the present embodiment, constructing a three-dimensional topological constraint model based on a three-dimensional terrain model includes the following steps:

[0087] S210, dividing the three-dimensional terrain model into a plurality of three-dimensional grid models, extracting the slope feature parameters and curvature feature parameters of all three-dimensional grid models, and calculating the average slope parameters and average curvature parameters of the three-dimensional terrain model based on the slope feature parameters and curvature feature parameters, wherein the three-dimensional grid model is a regular grid model;

[0088] S220, using the average slope parameters and average curvature parameters as constraint conditions;

[0089] S230, traversing each three-dimensional grid model, and dividing the terrain features of each three-dimensional grid model according to the constraint conditions;

[0090] S240, establishing topological connection rules of the three-dimensional terrain model based on the terrain features, and constructing a topological graph using graph theory algorithm;

[0091] S250, constructing a three-dimensional topological constraint model combining the topological graph, constraint conditions and three-dimensional terrain model.

[0092] The three-dimensional terrain model is divided into a plurality of three-dimensional grid models, the slope characteristic parameters and the curvature characteristic parameters of all three-dimensional grid models are extracted, and the average slope parameters and the average curvature parameters of the three-dimensional terrain model are calculated based on the slope characteristic parameters and the curvature characteristic parameters, wherein the three-dimensional grid model is a regular grid model. In this embodiment, the three-dimensional grid model is a representation manner of discretizing the three-dimensional space into cubic units (grid) of the same size and arranged in order. First, the continuous three-dimensional terrain model is divided into regular cubic grid units (three-dimensional grid model), each grid unit has a fixed size, and the quantitative analysis of the terrain characteristics is realized through the discretization process. After obtaining a plurality of three-dimensional grid models, the slope characteristic parameters and the curvature characteristic parameters of all three-dimensional grid models are extracted, which can be completed by using a geographic information system tool. For each grid unit, the slope can also be calculated based on the elevation values of the surrounding grids. Similarly, for each grid unit, the curvature can also be calculated based on the elevation values of the surrounding grids. The slope characteristic parameters represent the terrain inclination degree at the position of the grid unit; the curvature characteristic parameters represent the terrain bending degree at the position of the grid unit. Then, the average slope parameters and the average curvature parameters are calculated according to the slope characteristic parameters and the curvature characteristic parameters. The average slope parameters refer to the arithmetic mean of the slope characteristic parameters of all grid units, and the average curvature parameters refer to the arithmetic mean of the curvature characteristic parameters of all grid units. The average slope formula and the average curvature formula can be used, as follows:

[0093] Average slope = Σ (slope value of each grid) / total number of grids;

[0094] Average curvature = Σ (curvature value of each grid) / total number of grids;

[0095] The average slope parameters and the average curvature parameters are used as constraint conditions, which means that restrictions or requirements are imposed on variables or parameters in the optimization, modeling or decision-making process. The slope parameters and the average curvature parameters as constraint conditions mean that specific terrain steepness and bending degree requirements must be met in the analysis, design or simulation process. The average curvature parameters represent the overall bending degree of the terrain. The average slope parameters are used as constraint conditions, that is, the maximum curvature limit and the minimum curvature limit are set. For example, the maximum curvature limit of the path through which the inspection robot dog can pass is not more than 0.2 m - ¹, and the minimum curvature limit is not less than 0.1 m - ¹. The average slope parameters are used as constraint conditions, that is, the maximum slope limit and the minimum slope limit are set. For example, the maximum slope limit of the path through which the inspection robot dog can pass is not more than 10%, and the minimum slope limit is not less than 0.5%.

[0096] Subsequently, each three-dimensional grid model is traversed to divide the terrain features of each three-dimensional grid model according to the constraint conditions, which means checking each grid cell in three-dimensional space one by one, and assigning corresponding terrain feature labels to each grid based on preset constraint conditions. In three-dimensional space modeling, the three-dimensional grid model divides the target area into regular cubic grids (grids), and each grid represents the terrain or environmental features within a certain spatial range. According to the constraint conditions, the grid is traversed and the constraint parameters are calculated. For each grid, the inclination angle of the grid is calculated by the height difference of the adjacent grid. The degree of curvature of the surface is calculated by the height change rate of the local area of the grid, and one or more feature labels are assigned to each grid to describe the terrain properties of the area. For example, if the slope of a certain grid is ≤20° and there are no obstacles, it is marked as "passable flat ground"; if the slope is >30° and the curvature is >0.4 / m, it is marked as "steep and rugged terrain"; if the grid is occupied by rocks, it is marked as "obstacle".

[0097] Based on the topological connection rules of the three-dimensional terrain model established based on terrain features, a topological graph is constructed using graph theory algorithms, and each grid is traversed to calculate terrain parameters such as slope and curvature. According to the preset constraint conditions, the grid is divided into different terrain features. Subsequently, according to the terrain features (such as passability, slope, obstacles, etc.), the connection allowed conditions of adjacent areas in three-dimensional space are defined, for example, if two adjacent grids are both "passable flat ground" (such as slope ≤20° and no obstacles), then connection is allowed; if one of the grids is "steep slope" (slope >30°), then connection is prohibited; in the vertical direction, only when the height difference between the upper and lower layers of the grid is ≤ the leg height of the inspection robot dog (such as 0.2m), vertical connection is allowed. After obtaining the topological connection rules, a topological graph is constructed using graph theory algorithms, which is a collection of algorithms for studying graphs and related problems. Graph theory algorithms solve path planning, network optimization, connectivity analysis and other tasks in practical problems by analyzing the relationship between vertices (nodes) and edges. Each grid cell or terrain category in the three-dimensional terrain model is traversed, and according to the topological connection rules, the corresponding vertices and edges are added to the graph, which can be represented by adjacency matrix or adjacency list, etc. Data structure to represent the graph, thus obtaining the topological graph.

[0098] A three-dimensional topological constraint model is constructed by combining a topological graph, constraint conditions and a three-dimensional terrain model. The three-dimensional terrain model is taken as a spatial base, the spatial connectivity of the terrain unit is described by the topological graph, and the spatial relationship is quantitatively limited by combining the constraint conditions, and finally a three-dimensional topological constraint model that is computable and verifiable is formed. The fusion of the topological graph and the three-dimensional terrain model includes spatial mapping and attribute association. The spatial mapping refers to mapping the nodes and edges in the topological graph to specific positions in the three-dimensional terrain model. The attribute association refers to giving the terrain attributes (such as the slope value of the edge and the elevation of the node) to the edges and nodes of the topological graph, so that the topological graph has spatial and geographical characteristics. Then, the constraint conditions are converted into edge weights or node attribute limits of the topological graph.

[0099] Through the constructed three-dimensional topological constraint model, accurate terrain analysis can be performed, the optimal path can be provided for the inspection robot dog, obstacles can be avoided, the navigation efficiency can be improved, and the reliability and intelligent level of spatial decision-making can be significantly improved.

[0100] In one of the embodiments of the present embodiment, the three-dimensional terrain model is modified by combining the three-dimensional topological constraint model and the positioning data to obtain a modified three-dimensional terrain model, including the following steps:

[0101] S310, converting the coordinate system of the positioning data from the world coordinate system to the grid coordinate system of the three-dimensional grid model;

[0102] S320, mapping the three-dimensional grid model and the coordinate system converted positioning data to a two-dimensional plane grid, and determining the historical positioning points in the plurality of plane grids in the plane grid according to the historical movement track in the positioning data;

[0103] S330, calculating the elevation value corresponding to the historical positioning point in each plane grid by using a bilinear interpolation algorithm;

[0104] S340, calculating the curvature and slope corresponding to each three-dimensional grid model based on each elevation value;

[0105] S350, determining the abnormal type by combining the curvature, slope and constraint condition, and performing abnormal correction on the historical positioning points according to the abnormal type to obtain a modified three-dimensional terrain model.

[0106] First, converting the positioning data into the grid coordinate system of the three-dimensional grid model refers to converting the positioning data into the grid coordinate system of the three-dimensional grid model. Real-time three-dimensional spatial position information (longitude, latitude, and elevation) obtained through positioning technology can be converted into the coordinate system of a predefined three-dimensional grid model (or three-dimensional voxel model), so that the positioning data can correspond to a specific grid unit in the grid model. Convert geographic coordinates to plane coordinates. If the positioning data is longitude and latitude, such as coordinates (116.391°, 39.907°, 50m), it needs to be converted to plane coordinates first, such as (4043150m, 215500m, 50m) through UTM projection or Gauss-Kruger projection.

[0107] Second, mapping the grid coordinate system to the three-dimensional terrain model to obtain historical positioning points refers to mapping the grid coordinates (i, j, k) in the three-dimensional grid model to the corresponding plane grid (such as the X-Y plane or any horizontal slice) of the model to obtain historical positioning points. The grid coordinate system refers to the three-dimensional grid model, which discretizes space into regular grid units, and each grid unit is identified by an integer index (i, j, k). The plane grid is a horizontal slice or projection plane of the three-dimensional grid model (such as the X-Y plane of the terrain grid). That is, the grid coordinate system is mapped to the horizontal slice of the fixed layer.

[0108] Calculating the elevation value corresponding to each historical positioning point in each plane grid using a predetermined bilinear interpolation algorithm refers to using a bilinear interpolation method to estimate the accurate elevation value of the positioning point based on the four known elevation values around the historical positioning point in the plane grid. Bilinear interpolation is a two-dimensional interpolation method used to estimate the elevation value of a target point based on the elevation values of four known points, suitable for elevation estimation at any position in the plane grid. That is, the elevation value corresponding to each historical positioning point is calculated by the bilinear interpolation algorithm.

[0109] After obtaining the elevation value, the curvature and slope corresponding to each three-dimensional grid model are calculated based on each elevation value. The elevation value is the vertical height of a point in three-dimensional space relative to the reference surface, and is the most basic geometric property in the terrain model. In the three-dimensional grid model, each grid unit (grid) corresponds to an elevation value, forming a discretized representation of the terrain surface. For each grid unit, the first derivative (slope) and second derivative (curvature) of the point are calculated using the elevation values of its adjacent grids, such as the 8-neighborhood points in a 3x3 window, to obtain the corresponding curvature. For each grid unit, the gradient in the x and y directions (i.e. the rate of change of slope) is calculated using the elevation value of the grid unit, and then converted to an angle through the arctangent function to obtain the corresponding slope.

[0110] The abnormal types are determined in combination with the curvature, the slope and the constraint condition, the historical positioning points are corrected according to the abnormal types to obtain a corrected three-dimensional terrain model, in the embodiment, the abnormal types are curvature abnormal and slope abnormal, the curvature abnormal refers to a point whose curvature value exceeds a normal range or does not conform to a terrain logic, and the slope abnormal refers to a point whose slope value exceeds a reasonable range or does not conform to the terrain logic. The curvature and the slope values of each point are compared with the constraint condition to determine whether the point is abnormal, the curvature abnormal refers to a point whose curvature value exceeds the constraint condition, and the slope abnormal refers to a point whose slope value exceeds the constraint condition. Subsequently, the points of the curvature abnormal and the slope abnormal are marked as points to be corrected, different correction strategies are adopted according to the abnormal types. The curvature abnormal correction can be corrected by interpolation using the curvature values of surrounding normal points (such as inverse distance weighted interpolation, Kriging interpolation, etc.). The slope abnormal correction can be corrected by interpolation using the slope values of surrounding normal points. For each point to be corrected, the corresponding correction strategy is applied to obtain corrected curvature and slope values. According to the corrected curvature and slope values, the corrected elevation data is inversed, the elevation data is recalculated by integrating or least square method, etc., in combination with the known elevation data and the corrected curvature and slope values, the three-dimensional terrain model is generated using the corrected elevation data.

[0111] The three-dimensional terrain model is corrected by the three-dimensional topological constraint model and the positioning data, which can significantly improve the accuracy and authenticity of the terrain data, optimize terrain analysis and application, improve data processing efficiency, enhance the robustness of the terrain model, support multi-scale terrain analysis, promote the sharing and reuse of terrain data, and reduce the cost of terrain data acquisition and processing.

[0112] In one of the embodiments of the present embodiment, a motion feasibility matrix of the target region is established according to the corrected three-dimensional terrain model, and path planning for the inspection robot dog is based on the motion feasibility matrix, including the following steps:

[0113] S410, calculating the average grid slope and the average grid curvature corresponding to each three-dimensional grid model based on the slope feature parameter and the curvature feature parameter in each three-dimensional grid model;

[0114] S420, judging the motion feasibility of the inspection robot dog in each three-dimensional grid model in combination with the average grid slope and the average grid curvature, and determining the feasibility value of each three-dimensional grid model;

[0115] S430, constructing a motion feasibility matrix of the target region based on each feasibility value;

[0116] S440, path planning for the inspection robot dog according to the motion feasibility matrix by using Dijkstra algorithm.

[0117] The average grid slope and average grid curvature corresponding to each three-dimensional grid model are calculated based on the slope characteristic parameter and the curvature characteristic parameter in each three-dimensional grid model. The slope is the inclination degree of the terrain surface at a point, which is usually expressed by an angle or a percentage. In the three-dimensional grid model, each grid unit (or pixel) has a corresponding slope value, which represents the inclination degree of the terrain of the grid unit. The curvature is the bending degree of the terrain surface at a point, which is divided into plane curvature (horizontal curvature), profile curvature (vertical curvature) and total curvature. In the three-dimensional grid model, each grid unit has a corresponding curvature value, which represents the bending degree of the terrain of the grid unit. The average grid slope refers to the average value of the slope of all grid units in a three-dimensional grid model region, which is used to reflect the average inclination degree of the terrain of the region. The average grid curvature refers to the average value of the curvature of all grid units in a three-dimensional grid model region, which is used to reflect the average bending degree of the terrain of the region. The slope values of all grid units in a three-dimensional grid model region are summed up, and then divided by the total number of grid units to obtain the average grid slope. The curvature values of all grid units in a three-dimensional grid model region are summed up, and then divided by the total number of grid units to obtain the average grid curvature.

[0118] As shown in Figure 2 As shown in Figure 2 is a profile view of the three-dimensional terrain model in this embodiment. In Figure 2 the three-dimensional grid model with a black triangle in the figure is a ground three-dimensional grid model. Figure 2 It is only a schematic diagram, and the specific can be determined according to the actual situation.

[0119] Next, the feasibility of each three-dimensional grid model is determined by combining the average grid slope corresponding to each three-dimensional grid model with the preset grid slope and the average grid curvature with the preset grid curvature. In this embodiment, the grid slope and grid curvature of the three-dimensional grid model that is not a ground surface are not considered. The average grid slope reflects the average value of the slope of all grid cells in a three-dimensional grid model region, indicating the average inclination of the terrain in the region. The preset grid slope is a slope threshold value preset according to specific application requirements or standards, used to determine whether the terrain meets the requirements. The average grid curvature reflects the average value of the curvature of all grid cells in a three-dimensional grid model region, indicating the average bending degree of the terrain in the region. The preset grid curvature is a curvature threshold value preset according to specific application requirements or standards, used to determine whether the terrain meets the requirements. The feasibility determination is to determine whether the three-dimensional grid model meets specific conditions or requirements by comparing the average grid slope with the preset grid slope and the average grid curvature with the preset grid curvature. The feasibility value is a quantitative index for indicating the degree to which the three-dimensional grid model meets specific conditions or requirements. The higher the feasibility value, the more the terrain meets the requirements; the lower the feasibility value, the less the terrain meets the requirements. For each three-dimensional grid model, compare its average grid slope with the preset grid slope threshold and its average grid curvature with the preset grid curvature threshold. If both the average grid slope and the average grid curvature meet the preset threshold (or are within the preset range), the three-dimensional grid model is determined to be feasible; otherwise, it is determined to be infeasible. According to the determination result, a feasibility value is calculated for each three-dimensional grid model. The feasibility value can be calculated in various ways, such as weighted scoring method, fuzzy comprehensive evaluation method, etc.

[0120] The motion feasibility matrix of the target region is constructed based on each feasibility value, which aims to evaluate the feasibility of different positions or grid cells in the target region for a specific motion (such as walking, vehicle driving, robot moving, etc.) through quantitative indicators. The motion feasibility matrix is a two-dimensional matrix, where each element represents the motion feasibility value of the corresponding grid cell in the target region. Through this matrix, the motion feasibility of different positions in the target region can be intuitively understood. Collect the feasibility values of all grid cells in the target region, which can be calculated by comparing the average grid slope, average grid curvature and preset threshold. According to the grid division of the target region, initialize a two-dimensional matrix with a size matching the number of grid cells. The number of rows and columns of the matrix correspond to the number of grid cells in the horizontal and vertical directions of the target region, respectively. Fill the feasibility value of each grid cell into the corresponding position of the motion feasibility matrix. Usually, the row index and column index of the matrix can correspond to the row number and column number of the grid cell in the target region, respectively.

[0121] The path planning of the inspection robot dog according to the motion feasibility matrix refers to using the quantified information stored in the previously constructed motion feasibility matrix to plan an optimal or feasible path for the inspection robot dog in the target area to meet specific motion requirements. By analyzing the motion feasibility matrix, the motion feasibility of different positions in the target area is understood. The grid element with a higher feasibility value indicates that it is easier or safer for the inspection robot dog to move in this position; the Dijkstra algorithm is a classical graph algorithm, which is a single-source shortest path algorithm, and is used to calculate the shortest path from a starting point to all other nodes in a weighted directed graph or undirected graph. The grid element with a lower feasibility value indicates that it is more difficult or risky to move. According to the specific requirements of the inspection robot dog (such as the fastest arrival, the lowest energy consumption, the highest safety, etc.), the path planning target is set, which will be the basis for subsequent path search and optimization, and an optimal or feasible path that meets the path planning target is searched.

[0122] By establishing the motion feasibility matrix of the target area through the corrected three-dimensional terrain model and planning the path of the inspection robot dog based on the matrix, accurate motion feasibility evaluation, efficient and safe path planning, adaptation to complex terrain environment, reduction of energy consumption and improvement of efficiency, enhancement of the autonomy and adaptability of the inspection robot dog, support for real-time path adjustment, and improvement of the success rate of task execution can be achieved.

[0123] In one of the embodiments of the present embodiment, the path planning of the inspection robot dog according to the motion feasibility matrix by using the Dijkstra algorithm includes the following steps:

[0124] S510, converting the motion feasibility matrix into a graph structure, wherein each matrix element in the motion feasibility matrix is a node in the graph structure;

[0125] S520, determining the connection relationship of each node according to the feasibility value in the motion feasibility matrix;

[0126] S530, taking the initial position of the inspection robot dog as the starting node and counting the shortest distance from the starting node to each node;

[0127] S540, arranging all the shortest distances from small to large to determine the priority queue of the starting node;

[0128] S550, determining the expansion node of the starting node according to the connection relationship and the priority queue, and counting the priority queue corresponding to the expansion node;

[0129] S560, determining the target expansion node of the expansion node according to the connection relationship and the priority queue corresponding to the expansion node;

[0130] S570, reiterate the above steps until all nodes are traversed, and record the predecessor node of each node;

[0131] S580, according to the predecessor node, trace back to the starting node to obtain the shortest path of the inspection robot dog.

[0132] First, the motion feasibility matrix is converted into a graph structure, where each matrix element in the motion feasibility matrix is a node in the graph structure. The motion feasibility matrix is a two-dimensional matrix, where each element represents the motion feasibility value of the corresponding grid cell in the target area. This value reflects the difficulty, safety, or efficiency of the inspection robot dog moving on that grid cell. The graph structure is a data structure composed of nodes and edges, used to represent the relationship between objects. Each element in the motion feasibility matrix (i.e. each grid cell) is mapped to a node in the graph structure, and each node can contain its corresponding motion feasibility value, as well as other related information (such as coordinates, terrain features, etc.). According to the adjacency relationship between grid cells, edges between nodes are established in the graph structure. The weight of the edge can be determined according to the difference in motion feasibility between grid cells or other factors (such as distance, energy consumption, etc.). For example, the weight of the edge can be set to some function (such as the average value, difference value, etc.) of the motion feasibility values of the two adjacent grid cells. All nodes and edges are combined to form a complete graph structure. This graph structure can reflect the connection relationship and motion feasibility difference between different positions in the target area. In the target area, the inspection robot dog needs to move from the starting point to the ending point. In order to plan an optimal path, the node and edge information in the graph structure needs to be used for path search.

[0133] Determining the connection relationship of each node according to the feasibility value in the motion feasibility matrix means that in the process of converting the motion feasibility matrix into a graph structure, the feasibility value of each element in the matrix is used to determine which nodes should be connected (edges) and how to define the weight or properties of these connections. First, determine which nodes are adjacent. In a two-dimensional grid map, adjacent nodes are usually up, down, left, right, or diagonally adjacent grids, for example, in a 4-neighbor model, each node has at most 4 adjacent nodes; in an 8-neighbor model, at most 8 adjacent nodes. For each node, check the feasibility values of its adjacent nodes. The feasibility value can be used to determine whether to establish a connection and how to define the weight of the connection. Then, set a feasibility threshold. Only when the feasibility value of the adjacent node is higher than the threshold, a connection is established, for example, if the threshold is 50, only the adjacent nodes with a feasibility value ≥ 50 will be connected to the current node. If the feasibility value of a certain grid is 0 or very low (indicating an obstacle or impassable area), no connection is established with any node. In a dynamic environment, the feasibility value may change over time, and the connection relationship needs to be updated in real time.

[0134] The initial position of the inspection robot dog is taken as the starting node, and the shortest distance from the starting node to each node is calculated. After converting the motion feasibility matrix into a graph structure, the node corresponding to the initial position of the inspection robot dog is taken as the starting point, and the shortest path distance from this node to all other nodes in the graph structure is calculated. In the graph structure, the shortest distance refers to the sum of the weights of all edges on the shortest path from one node to another. According to the motion feasibility matrix, a graph structure is constructed, where each grid corresponds to a node, and edges are established between adjacent grids according to the feasibility value, and the weight of the edge is defined (such as the inverse of the feasibility value, distance, etc.). The node corresponding to the initial position of the inspection robot dog is taken as the starting node, for example, if the initial position is in grid (0, 0), the starting node is (0, 0). The shortest distance from the initial position to all other nodes needs to be calculated to provide a basis for path planning. The terrain data is converted into a graph structure, and the shortest distance from the starting node to all other nodes is calculated using the shortest path algorithm. According to the shortest distance information, an optimal path is planned.

[0135] In complex and variable inspection environments, the inspection robot dog faces challenges such as rugged terrain and numerous obstacles. In order to efficiently complete the inspection task, the robot dog needs to rely on advanced artificial intelligence algorithms and powerful computing power to plan an optimal path from the initial position to the target position. In this process, the Floyd-Warshall algorithm and other shortest path algorithms play a key role. These algorithms can calculate the shortest path between all node pairs in the graph, providing comprehensive path planning information for the robot dog. By using these algorithms, the robot dog can quickly determine the shortest distance from the starting node to all other nodes, and thus select a safe and efficient path. Specifically, when the robot dog receives an inspection task, it will first use its sensors and cameras and other devices to perceive and model the surrounding environment, constructing a map containing information such as terrain features and obstacle locations. Then, the robot dog will use advanced algorithms such as the Floyd-Warshall algorithm to analyze and calculate this map in depth. During the operation of the algorithm, powerful computing power ensures the efficiency and accuracy of the calculation. Through continuous iteration and optimization, the algorithm can quickly find the shortest path from the starting node to the target node, taking into account factors such as terrain undulations and obstacle avoidance, ensuring the feasibility and optimality of the path. Finally, the inspection robot dog will navigate to the target position according to the optimal path calculated by the algorithm, completing the inspection task. This process not only reflects the important role of artificial intelligence algorithms in path planning, but also demonstrates the key support of powerful computing power for efficient and accurate inspection.

[0136] The priority queue of all shortest distances arranged from small to large determines the starting node. After calculating the shortest distances from the starting node to all other nodes, arrange these shortest distances in order from small to large, and build a priority queue based on this. The construction of the priority queue is to prioritize the nodes closer to the starting node in path planning or other scenarios that require processing nodes by priority. The priority queue is a data structure where each element has a priority. Elements with higher priority (i.e., nodes closer to the starting node) are processed first. In path planning, the priority queue is usually used to store nodes to be explored and sorted by shortest distance to prioritize exploring nodes with shorter distances. Using the shortest path algorithm to calculate the shortest distance from the starting node to all other nodes, the result is usually a dictionary or list recording the shortest distance from each node to the starting node. Extract all nodes and their corresponding shortest distances to form node-distance pairs, and sort the nodes by distance from small to large, for example, for the node set [(A, 5), (B, 3), (C, 7)], the sorted result is [(B, 3), (A, 5), (C, 7)]. Insert the sorted nodes into the priority queue. The priority queue will automatically adjust the order according to the priority of the nodes (i.e., the shortest distance) to ensure that the node with the shortest distance is always at the front of the queue. In Python, the heapq module can be used to implement the priority queue.

[0137] The extended node of the starting node is determined according to the connection relationship and the priority queue, which refers to selecting the next node to be expanded in the path planning or graph search process based on the connection relationship (i.e. adjacent nodes) of the current node and the priority information in the priority queue. The extended node refers to the node taken out from the priority queue that is closer to the starting node, and processing these nodes first helps to quickly find the optimal path. The extended node is the node taken out from the priority queue in the graph search process, which is used to further explore its adjacent nodes. The extended node is usually the node that is closer to the starting node. The priority queue is a data structure used to store nodes to be explored and is sorted according to the priority (such as the shortest distance) of the nodes. Nodes with high priority are processed first. The graph structure is constructed according to the motion feasibility matrix, where each node represents a position or state, and the edge represents the connection relationship between nodes, and the weight of the edge is defined. The starting node is inserted into the priority queue, and its priority is set to 0 (because the distance from the starting node to itself is 0). The node with the highest priority (i.e. the node closest to the starting node) is taken out from the priority queue as the current extended node. All adjacent nodes of the extended node are explored, and the shortest distance from the starting node to these adjacent nodes is calculated. If the shortest distance of the adjacent node can be updated (i.e. a shorter path is found), its priority is updated and it is inserted into the priority queue. During the process of expanding the node, the state of the priority queue is recorded, including the nodes in the current queue and their priorities, which can be used to count the changes of the priority queue through logs, debugging output or visualization tools.

[0138] The target expansion node of the extended node is determined according to the connection relationship and the priority queue corresponding to the extended node, which refers to selecting the next target expansion node to be explored in the path planning or graph search process based on the adjacent nodes of the current extended node and the priority information in the priority queue. The target expansion node is the target node selected from the adjacent nodes of the extended node according to the priority information of the priority queue. The node with the highest priority (i.e. the node closest to the starting node) is taken out from the priority queue as the current extended node, and all adjacent nodes of the extended node are explored, and the shortest distance from the starting node to these adjacent nodes is calculated. According to the priority information of the priority queue, the node with the highest priority is selected from the adjacent nodes as the target expansion node. If the shortest distance of the target expansion node can be updated, its priority is updated and it is inserted into the priority queue. Repeat the above process until the priority queue is empty or the target node is found.

[0139] Subsequently, the above steps are repeated until all nodes are processed, which means that the process of taking a node out of the priority queue, exploring its adjacent nodes, updating the shortest distance, and inserting the unprocessed adjacent nodes into the priority queue continues until the priority queue is empty. During the path planning process, whenever a shorter path to a node is found, the predecessor node of the node (i.e., the node from which the path to the current node is shorter) needs to be recorded. The record of the predecessor node helps subsequent path backtracking. The starting node is inserted into the priority queue, and its predecessor node is set. The node with the highest priority (the expansion node) is taken out of the priority queue, and all adjacent nodes of the expansion node are explored. If a shorter path to an adjacent node is found, the shortest distance of the adjacent node is updated, its predecessor node is set to the current expansion node, and the adjacent node is inserted into the priority queue. The above steps are repeated until the priority queue is empty (i.e., all nodes are processed).

[0140] Starting from the target node, the predecessor node information is recorded, and the path is backtracked to the starting node step by step. In this way, the shortest path from the starting node to the target node can be reconstructed. After backtracking is completed, all nodes on the path are arranged in order, which is the shortest path of the inspection robot dog. Starting from the target node, the predecessor node of the target node is found according to the predecessor node of the target node, and the predecessor node is repeated, until the starting node is backtracked to the starting node. Since the path is backtracked from the target node to the starting node, the route needs to be reversed to obtain the correct order from the starting node to the target node.

[0141] By converting the motion feasibility matrix into a graph structure and using a priority queue and predecessor node record to implement path planning for the inspection robot dog, the method has high efficiency, accuracy, and scalability. The method can quickly find the shortest path and provide guidance for motion control of the inspection robot dog, improving the emergency response rate of the inspection robot dog.

[0142] In one of the embodiments of the present embodiment, a dynamic risk evaluation model is constructed by combining real-time water and rain monitoring data and historical water and rain monitoring data, and a regional dynamic risk warning area is generated based on the dynamic risk evaluation model and a three-dimensional terrain model, including the following steps:

[0143] S610, selecting the prone index and disaster-causing index using a correlation analysis algorithm according to the historical water and rain monitoring data, and determining the rainfall factor;

[0144] S620, performing similarity calculation on the real-time water and rain monitoring data and the prone index and disaster-causing index to determine the target area as a disaster-prone area;

[0145] S630, determining the soil state based on the soil state data and using a preset soil state analysis model;

[0146] S640, calculate the membership degrees of the soil state, the disaster-prone area and the rainfall factor respectively by using the preset membership function;

[0147] S650, calculate the weight values of the soil state, the disaster-prone area and the rainfall factor respectively by using the analytic hierarchy process according to the membership degrees of the soil state, the disaster-prone area and the rainfall factor;

[0148] S660, integrate and construct a dynamic risk evaluation model by using fuzzy mathematical operation according to the weight values of the soil state, the disaster-prone area and the rainfall factor;

[0149] S670, generate a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

[0150] According to historical water and rainfall monitoring data, the preset correlation analysis algorithm is used to select the disaster-prone index and the disaster-causing index. The historical water and rainfall monitoring data includes long time series data such as rainfall, water level, flow, flow rate, soil moisture, and weather conditions. The preset correlation analysis algorithm can be Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information, etc. The disaster-prone index refers to a variable that is highly correlated with the frequency of disaster occurrence. In this embodiment, the disaster-prone index can be vegetation coverage, slope structure type, etc. The disaster-causing index is a variable that is directly related to the intensity or loss of disaster, such as maximum flood peak flow, highest water level, etc., reflecting the destructive power of disaster. The correlation between each index and the frequency of disaster occurrence (disaster-prone index) and the intensity of disaster (disaster-causing index) is calculated by Pearson correlation coefficient. According to the correlation analysis result, the threshold value of the correlation coefficient is set, and the index with significant correlation is selected. According to the historical water and rainfall monitoring data, the correlation analysis algorithm is used to determine the rainfall factor. The rainfall factor is a rainfall characteristic parameter that is highly correlated with a specific target (such as disaster occurrence, water resource change, ecological environment influence, etc.), such as maximum rainfall intensity, cumulative rainfall, etc. The correlation between each rainfall factor and the target variable (such as disaster occurrence, water level change, etc.) is calculated. According to the correlation analysis result, the threshold value of the correlation coefficient is set, and the rainfall factor with significant correlation is selected. According to the direction and intensity of correlation, the rainfall factor that has the greatest impact on the target variable is determined, such as maximum rainfall intensity, cumulative rainfall, etc.

[0151] The real-time water and rainfall monitoring data in the embodiment refers to feature data in the spatial dimension, that is, the distribution characteristics of the real-time water and rainfall monitoring data in different geographical spaces. The real-time water and rainfall monitoring data in different geographical locations and soil types and topographical features have different influences on disaster-prone areas. For example, when the soil type is soft soil with large pores and the topographical feature is prone to water storage, the area is more prone to seepage than other locations. Therefore, the real-time water and rainfall monitoring data in the embodiment includes spatial dimension features, that is, real-time water and rainfall monitoring data in different topographical features and soil types. The similarity calculation is performed on the real-time water and rainfall monitoring data, the disaster-prone index, and the disaster-causing index to determine the disaster-prone area. In the embodiment, the similarity calculation can be performed by the Euclidean distance, cosine similarity, dynamic time warping (DTW), and the like. The Euclidean distance between the disaster-prone index and the disaster-causing index and the real-time water and rainfall monitoring data is calculated to obtain a similarity result. According to the similarity result, the area highly similar to the historical disaster mode is screened out, and the area with a similarity score higher than a threshold value is marked as the disaster-prone area.

[0152] The membership degrees of the soil state, the disaster-prone area, and the rainfall factor are calculated by using a preset membership function. The membership function is a core concept in fuzzy mathematics, which is used to quantify fuzzy concepts (such as “soil saturation”, “disaster-prone area”, etc.) into membership values (0-1). The membership value represents the degree to which an element belongs to a fuzzy set. For example, a membership degree of 0.8 represents “highly belongs to”, and a membership degree of 0.2 represents “lowly belongs to”. The soil state membership degree represents the degree to which the current state of the soil (such as dry, wet, and saturated) belongs to a fuzzy set (such as “disaster-prone soil state”). The disaster-prone area membership degree represents the degree to which an area belongs to the “disaster-prone area”. The rainfall factor membership degree represents the degree to which the rainfall condition (such as rainfall intensity and duration) belongs to the “disaster-prone rainfall condition”. The membership degrees of the soil state, the disaster-prone area, and the rainfall factor are calculated by using different membership functions.

[0153] According to the membership of soil state, disaster-prone area and rainfall factor, the weight values of these factors are calculated using the preset analytic hierarchy process (AHP), which refers to determining the relative importance of each factor in disaster assessment by constructing a hierarchical structure model, constructing a judgment matrix, calculating weights and conducting consistency test. Through the analytic hierarchy process, the weight values of soil state, disaster-prone area and rainfall factor can be calculated to determine the evaluation factors and evaluation target as soil state, disaster-prone area and rainfall factor, and the evaluation target as the water and rainfall disaster assessment result. The evaluation target is taken as the highest layer, the evaluation factors as the intermediate layer, and the evaluation result as the lowest layer. Then, the judgment matrix is constructed, the characteristic vector of the judgment matrix is calculated using mathematical methods (such as sum-product method, square root method, etc.), and the relative importance of each evaluation factor in disaster assessment is analyzed according to the calculated weight values.

[0154] The dynamic risk assessment model is constructed by integrating the weight values of soil state, disaster-prone area and rainfall factor using fuzzy mathematical operations, taking the weight values of soil state, disaster-prone area and rainfall factor calculated by the analytic hierarchy process as input. Each evaluation factor is fuzzified and converted into a fuzzy set or membership function for fuzzy mathematical operations. Each evaluation factor is integrated using the operation rules in fuzzy mathematics (such as fuzzy addition, fuzzy multiplication, etc.), and a comprehensive risk assessment index or risk level is calculated. Since soil state, disaster-prone area and rainfall factor may change over time, the dynamic risk assessment model needs to adjust the risk assessment results in real time according to these changes.

[0155] Based on the dynamic risk assessment model and the three-dimensional terrain model, the regional dynamic risk warning area is generated by data fusion of the risk assessment index or risk level output by the dynamic risk assessment model and the three-dimensional terrain model. Spatial analysis is performed on the fused data using spatial analysis tools such as geographic information system (GIS) to identify high-risk areas. According to the results of spatial analysis, the high-risk areas are divided into different warning areas, each corresponding to a specific risk level or risk assessment index range. The visualization function of the three-dimensional terrain model is used to display the warning areas on the map with different colors or symbols, so that decision-makers can intuitively understand the spatial distribution and morphology of the risk areas. The generated regional dynamic risk warning area can be used to guide disaster prevention and mitigation work, such as evacuation plan development, resource allocation, etc.

[0156] By constructing a dynamic risk assessment model based on real-time and historical water and rainfall monitoring data, and generating a regional dynamic risk warning area based on the model and a three-dimensional terrain model, the accuracy and scientificity of risk assessment can be improved, the loss caused by disasters can be reduced, the efficiency of disaster prevention and mitigation can be improved, emergency plans can be developed in advance, and the speed and effectiveness of emergency response can be improved.

[0157] In one embodiment of the present embodiment, generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model includes the following steps:

[0158] S710, for any one three-dimensional grid model, generating a risk value corresponding to the three-dimensional grid model using the dynamic risk evaluation model;

[0159] S720, generating a risk surface of the three-dimensional terrain model by spatial interpolation of each risk value using the Kriging method;

[0160] S730, dividing the risk surface according to a preset risk level threshold to determine the regional dynamic risk early warning area.

[0161] For any one three-dimensional grid model, generating a risk value corresponding to the three-dimensional grid model using the dynamic risk evaluation model is to apply the dynamic risk evaluation model to the space represented by the three-dimensional grid model, quantifying the risk degree of each grid unit occurring a specific disaster (such as flood, landslide, etc.), providing basic data for subsequent risk analysis, early warning area division, etc. By discretizing the three-dimensional space into grid units, it is convenient to digitally describe and analyze the space, so that the risk assessment can be specific to each spatial location, providing a spatial carrier for subsequent risk value generation. The dynamic risk evaluation model is a mathematical model that comprehensively considers multiple factors (such as rainfall, soil state, terrain, etc.) for assessing the likelihood and severity of a disaster occurring in a specific area. It usually contains multiple sub-models, such as rainfall prediction sub-model, soil stability analysis sub-model, disaster impact range prediction sub-model, etc., each sub-model is related to each other through specific algorithms and parameters, and together constitutes a complete evaluation system. Input data preparation, three-dimensional grid model data is used to extract relevant attribute information of each grid in the three-dimensional grid model, such as elevation, slope, slope direction, soil type, vegetation coverage, etc. Real-time monitoring data is used to obtain real-time monitoring data related to risk assessment, such as current rainfall intensity, water level, weather conditions, etc., which can reflect the current environmental conditions and have an important influence on risk assessment. The processed data is input into the dynamic risk evaluation model, and the model calculates each grid according to the preset algorithm and parameters to obtain the risk value of the grid occurring a disaster. The calculation of risk value usually considers multiple factors, such as the relationship between rainfall intensity and soil stability, the influence of terrain on water flow, etc.

[0162] The risk surface of the three-dimensional terrain model generated by spatial interpolation of each risk value by the Kriging method is to convert discrete risk value data into a continuous risk distribution surface by using the Kriging interpolation algorithm, so as to intuitively show the change trend of the risk in the three-dimensional terrain space, and provide more comprehensive spatial information support for risk assessment and early warning. The Kriging method is a spatial interpolation method based on statistics, which assumes that the variables in space have randomness and structure, that is, the variables in space change not only contains random error, but also follows a certain spatial correlation rule. The risk surface data calculated by interpolation is stored in the three-dimensional terrain model to form a three-dimensional terrain data set containing risk information.

[0163] The regional dynamic risk warning area is determined by dividing the risk surface according to the preset risk level threshold. By setting the numerical limits corresponding to different risk levels, the continuous risk surface data is converted into partitions with clear risk levels, so as to intuitively show the spatial distribution and severity of disaster risk, and provide key basis for disaster prevention decision-making. The risk value of each grid in the risk surface is compared with the preset risk level threshold. For example, for a grid with a risk value of 65, it is compared with each level threshold, and it is found that 65 is in the range of 51-80, so the grid is divided into a high risk level. By comparing the threshold values of all grids, the grids with the same risk level are classified into a class, so as to divide the entire risk surface into different risk areas. For example, all low-risk-level grids are combined into a low-risk area, and all high-risk-level grids are combined into a high-risk area.

[0164] The regional dynamic risk warning area is generated by the dynamic risk evaluation model and the three-dimensional terrain model, which can effectively improve the accuracy of risk assessment, improve the efficiency of resource utilization, and reduce the cost of disaster prevention and mitigation.

[0165] In one of the embodiments of the present embodiment, generating the risk surface of the three-dimensional terrain model by spatial interpolation of each risk value by the Kriging method includes the following steps:

[0166] S810, determining the spatial coordinates of the grid center point of the three-dimensional grid model where each risk value is located;

[0167] S820, calculating the semi-variation function value of each risk value according to the combined risk value and spatial coordinates by using a preset empirical variation function;

[0168] S830, combining the preset theoretical variation function model and the least square method to fit the semi-variation function value to determine the spatial distribution characteristics of the risk value;

[0169] S840, determining the search neighborhood range according to the spatial distribution characteristics, and determining all the interpolation points in the search neighborhood range by using the K-neighbor algorithm;

[0170] S850, calculate the spatial distance between each interpolation point based on each to-be-interpolated point;

[0171] S860, combine the semi-variation function value and the spatial distance to dynamically calculate the weight value between each to-be-interpolated point using a particle swarm optimization algorithm;

[0172] S870, weighted sum all weight values to obtain the estimated risk value of the to-be-interpolated point;

[0173] S880, based on the estimated risk value, perform spatial interpolation to obtain the risk surface of the three-dimensional terrain model.

[0174] First, determine the spatial coordinates of the grid center point of the three-dimensional grid model where each risk value is located. The grid center point refers to the geometric center point of each three-dimensional grid in space. In a three-dimensional coordinate system, the center point coordinates can be represented as (x, y, z), where x and y represent the horizontal position, and z represents the vertical position. Each grid risk value needs to be associated with the actual geographic spatial position. The center point coordinates are the spatial anchor points of the grid in the three-dimensional terrain, which are used for subsequent spatial analysis. The grid center point coordinates can be extracted directly through the "Grid to Point" tool in GIS software (such as ArcGIS, QGIS).

[0175] Subsequently, the semi-variation function value of each risk value is calculated based on the combined risk value and spatial coordinates using a pre-set empirical variation function. The semi-variation function is a core tool for describing the spatial correlation of spatial variables (such as risk values). The semi-variation function value calculated using the pre-set empirical variation function in combination with the risk value and spatial coordinates quantifies the structural differences and random noise of spatial data through mathematical methods, revealing the distribution rules and dependence relationships of risk values in space. The pre-set empirical variation function is as follows:

[0176]

[0177] where h is the spatial lag distance, N(h) is the number of sample pairs with a distance of h, x i , y i , z i are spatial coordinates, and r is the risk value.

[0178] The semi-variation function value of each risk value is calculated by the pre-set empirical variation function and the risk value and spatial coordinates. The semi-variation function value is mainly used to analyze the correlation and difference of variables in space to reveal the spatial structure characteristics of data.

[0179] Next, the spatial distribution characteristics of the risk value are determined by fitting the semi-variation function value in combination with a preset theoretical variation function model and the least square method. In this embodiment, the preset theoretical variation function model can be a Gaussian model. The least square method is used to minimize the sum of squares of errors between the model predicted value and the actual observed value, to determine the optimal parameters of the theoretical model, so that the theoretical model can fit the experimental semi-variation function value as much as possible, and the accuracy of the spatial distribution prediction is improved. Specifically, a suitable model can be selected according to the data characteristics, and then the least square method is used to adjust the parameters of the theoretical model, so that the semi-variation function value predicted by the model is closest to the experimental value.

[0180] The search neighborhood range is determined according to the spatial distribution characteristics, and all the to-be-interpolated points in the search neighborhood range are determined by using the K-neighbor algorithm. The spatial distribution characteristics refer to the distribution mode of the data points in the geographical or feature space. The neighborhood range is a spatial region dynamically determined according to the distribution characteristics, such as a circular radius, a rectangular window, etc. The K-neighbor algorithm is used to infer the attribute of the to-be-predicted sample by measuring the distance between the to-be-predicted sample and the known category sample, and using the characteristics of the neighboring samples. Specifically, the similarity between the target point and each point in the neighborhood is measured by using the Euclidean distance method, etc., then the points are sorted according to the distance, and the first k points are taken. The attribute values of the neighbors are weighted and averaged.

[0181] The spatial distance between each to-be-interpolated point and other to-be-interpolated points is calculated based on each to-be-interpolated point. For each position (to-be-interpolated point) that needs to estimate the attribute value, the geometric distance between the point and all other to-be-interpolated points in space is calculated, which can be realized by calculating the Euclidean distance. The to-be-interpolated point refers to a position point in space whose attribute value is unknown and needs to be estimated by an interpolation method.

[0182] The weight value between each to-be-interpolated point is dynamically calculated by using the particle swarm optimization algorithm in combination with the semi-variation function value and the spatial distance. The particle swarm optimization algorithm is used to find the optimal solution by cooperation and information sharing between individuals in the group. Each particle represents a potential solution, and the global optimal solution is approached by iteratively updating the speed and position. In the interpolation weight calculation, the particle swarm optimization algorithm is used to optimize the weight value, so that the error between the interpolation result and the actual observed value is minimized. Specifically, the particle swarm optimization algorithm is used to optimize the parameters (such as nugget value, sill value, and range) of the semi-variation function model, so that the model can better fit the actual data. In each iteration, the particle updates the speed and position according to the individual optimal solution and the global optimal solution, and dynamically adjusts the weight parameter. The weight value not only depends on the spatial distance, but also combines the semi-variation function value, so that the points with closer distance and similar attribute values obtain higher weight. Based on the optimized weight value, the to-be-interpolated points are weighted and averaged to obtain the final interpolation result.

[0183] The estimation risk value is obtained by weighted summation of all weight values, and the weight is determined by the spatial distance between the data point and the to-be-interpolated point, and the semi-variogram value. The closer the known point to the to-be-interpolated point, the greater the weight. The observation value of all known data points is multiplied by the corresponding weight, and then accumulated to obtain the estimation value of the to-be-interpolated point. The estimation risk value is the result of weighted summation, that is, the risk prediction value of the to-be-interpolated point.

[0184] Based on the estimation risk value, the risk surface of the three-dimensional terrain model is obtained by spatial interpolation, that is, whether the to-be-interpolated point is at the grid center point is judged by the estimation risk value. If the to-be-interpolated point is exactly at the grid center point, the risk value is directly assigned. If the to-be-interpolated point is not at the grid center point, the risk value needs to be distributed to the nearest grid center point through coordinate matching (such as the nearest neighbor method). The risk value of each grid center point is optimally estimated by directly using the fitted theoretical variogram model and the semi-variogram value, and a continuous three-dimensional risk body is generated, that is, the risk surface of the three-dimensional terrain model (X-Y-Z are spatial dimensions, and the risk value is an attribute).

[0185] The embodiment also provides a machine readable storage medium, which stores instructions for causing a machine to execute the water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog.

[0186] The embodiment of the application also provides an electronic device, which comprises:

[0187] a memory configured to store instructions; and

[0188] a processor configured to call the instructions from the memory and capable of implementing the water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog when the instructions are executed.

[0189] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system or a computer program product. Therefore, the application can be in the form of a completely hardware embodiment, a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, the application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0190] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0191] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0192] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0193] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0194] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, for storing instructions and data used and / or generated by the computing device. The memory can also include non-volatile memory, such as read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other non-volatile memory.

[0195] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0196] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0197] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A water and rainfall condition emergency monitoring method based on a Beidou inspection robot dog, characterized in that, The method comprises the following steps: Collecting environmental point cloud data in a target area by a preset laser radar, and obtaining positioning data of a patrol robot dog in the target area through a Beidou satellite system; Generating a three-dimensional terrain model according to the environmental point cloud data by using an iterative closest point algorithm; Constructing a three-dimensional topological constraint model based on the three-dimensional terrain model, and combining the three-dimensional topological constraint model and the positioning data to correct the three-dimensional terrain model to obtain a corrected three-dimensional terrain model; Establishing a motion feasibility matrix of the target area according to the corrected three-dimensional terrain model, and planning a path for the patrol robot dog based on the motion feasibility matrix; Obtaining historical water and rain monitoring data of the target area, and collecting real-time water and rain monitoring data and soil state data of the target area by using the patrol robot dog; Combining the real-time water and rain monitoring data, the soil state data and the historical water and rain monitoring data to construct a dynamic risk evaluation model, and generating a regional dynamic risk early warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

2. The method of claim 1, wherein, The method comprises the following steps: Dividing the three-dimensional terrain model into a plurality of three-dimensional grid models, and extracting slope characteristic parameters and curvature characteristic parameters of all the three-dimensional grid models; Calculating average slope parameters and average curvature parameters of the three-dimensional terrain model based on the slope characteristic parameters and the curvature characteristic parameters, wherein the three-dimensional grid model is a regular grid model; Taking the average slope parameters and the average curvature parameters as constraint conditions, traversing each three-dimensional grid model, and dividing the terrain features of each three-dimensional grid model according to the constraint conditions; Establishing topological connection rules of the three-dimensional terrain model based on the terrain features, and constructing a topological graph using a graph theory algorithm; Combining the topological graph, the constraint conditions and the three-dimensional terrain model to construct a three-dimensional topological constraint model.

3. The method of claim 2, wherein, The method comprises the following steps: Converting the coordinate system of the positioning data from a world coordinate system to a grid coordinate system of the three-dimensional grid model; Mapping the positioning data after coordinate system conversion and the three-dimensional grid model to a two-dimensional plane grid, and determining historical positioning points in a plurality of plane grids in the plane grid according to historical movement trajectories in the positioning data; Calculating elevation values corresponding to the historical positioning points in each plane grid by using a bilinear interpolation algorithm; Calculating curvatures and slopes corresponding to each three-dimensional grid model based on each elevation value; Determining an abnormal type by combining the curvatures, the slopes and the constraint conditions, and correcting the historical positioning points according to the abnormal type to obtain a corrected three-dimensional terrain model.

4. The method of claim 2, wherein, The method comprises the following steps: Calculate the average grid slope and average grid curvature corresponding to each three-dimensional grid model based on the slope feature parameter and the curvature feature parameter in each three-dimensional grid model respectively; Determine the motion feasibility of the inspection robot dog in each three-dimensional grid model by combining the average grid slope and the average grid curvature respectively, and determine the feasibility value of each three-dimensional grid model; Construct a motion feasibility matrix of the target area based on each feasibility value; Use Dijkstra algorithm to plan a path for the inspection robot dog according to the motion feasibility matrix.

5. The method of claim 4, wherein, The path planning for the inspection robot dog according to the motion feasibility matrix using Dijkstra algorithm includes the following steps: Convert the motion feasibility matrix into a graph structure, where each matrix element in the motion feasibility matrix is a node in the graph structure; Determine the connection relationship of each node according to the feasibility value in the motion feasibility matrix; Take the initial position of the inspection robot dog as the starting node, and count the shortest distance from the starting node to each node; Arrange all the shortest distances from small to large to determine the priority queue of the starting node; Determine the expansion node of the starting node according to the connection relationship and the priority queue, and count the priority queue corresponding to the expansion node; Determine the target expansion node of the expansion node according to the connection relationship and the priority queue corresponding to the expansion node; Repeat the above steps until all nodes are traversed, and record the predecessor node of each node; Backtrack from the predecessor node to the starting node to get the shortest path of the inspection robot dog.

6. The method of claim 1, wherein, The steps for constructing a dynamic risk evaluation model based on the real-time water and rain monitoring data, the soil state data and the historical water and rain monitoring data, and generating a regional dynamic risk warning area based on the dynamic risk evaluation model and the three-dimensional terrain model include: Select the prone index and disaster-causing index according to the historical water and rain monitoring data using correlation analysis algorithm, and determine the rainfall factor; Calculate the similarity of the real-time water and rain monitoring data, the prone index and the disaster-causing index to determine the target area as a disaster-prone area; Determine the soil state based on the soil state data and using a preset soil state analysis model; Calculate the membership degree of the soil state, the disaster-prone area and the rainfall factor respectively using a preset membership function; Calculate the weight value of the soil state, the disaster-prone area and the rainfall factor respectively using the analytic hierarchy process according to the membership degree of the soil state, the disaster-prone area and the rainfall factor; Integrate the soil state, the disaster-prone area and the rainfall factor using fuzzy mathematical operation to construct a dynamic risk evaluation model according to the weight value of the soil state, the disaster-prone area and the rainfall factor; Generate a regional dynamic risk warning area based on the dynamic risk evaluation model and the three-dimensional terrain model.

7. The method of claim 2, wherein, The steps for generating a regional dynamic risk warning area based on the dynamic risk evaluation model and the three-dimensional terrain model include: For any one of the three-dimensional grid models, a risk value corresponding to the three-dimensional grid model is generated by using the dynamic risk evaluation model; A risk surface of the three-dimensional terrain model is generated by spatial interpolation of each of the risk values by using the Kriging method; The risk surface is divided according to a preset risk level threshold to determine a regional dynamic risk early warning area.

8. The method of claim 7, wherein, The risk surface of the three-dimensional terrain model is generated by spatial interpolation of each of the risk values by using the Kriging method, including the following steps: A spatial coordinate of a grid center point of the three-dimensional grid model where each of the risk values is located is determined; A semi-variogram value of each of the risk values is calculated by using a preset empirical variogram function in combination with the risk value and the spatial coordinate; A spatial distribution characteristic of the risk value is determined by fitting the semi-variogram value in combination with a preset theoretical variogram function model and a least square method; A search neighborhood range is determined according to the spatial distribution characteristic, and all interpolation points in the search neighborhood range are determined by using a K-neighbor algorithm; A spatial distance between each of the interpolation points is calculated; A weight value between each of the interpolation points is dynamically calculated by using a particle swarm optimization algorithm in combination with the semi-variogram value and the spatial distance; An estimated risk value of each of the interpolation points is obtained by weighted summation of all the weight values; The risk surface of the three-dimensional terrain model is obtained by spatial interpolation based on the estimated risk value.

9. A machine-readable storage medium, characterized in that, The machine readable storage medium has instructions stored thereon for causing a machine to perform the water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog according to any one of claims 1 to 8.

10. A water and rainfall condition emergency monitoring system based on a Beidou inspection robot dog, characterized in that, Comprise: a memory configured to store instructions; and a processor configured to call the instructions from the memory and enable the water and rainfall condition emergency monitoring method based on the Beidou inspection robot dog according to any one of claims 1 to 8 when the instructions are executed.

Images