Self-adaptive voxel grid planning method and device for natural water body and storage medium

Abstract

The invention discloses a self-adaptive voxel grid planning method and device for a natural water body and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: dividing a three-dimensional terrain model into structured initial voxel grids according to a preset initial resolution, and fusing the real-time water quality, flow velocity, historical abnormality and carrier energy consumption multi-source data to obtain a structured initial voxel grid; quantizing a risk value and a risk level of each initial voxel grid through a risk field function, and finally adjusting the initial voxel grids according to a risk level matching differentiation target resolution; scattered multi-dimensional risk information is converted into a voxel-level accurate risk assessment result, so that risk distribution is visualized and quantifiable; through adaptive resolution adjustment of high-risk area refinement and medium and low-risk area coarsening, the high-risk area monitoring precision is guaranteed, the calculation and energy consumption cost of a low-risk area is reduced, and the balance of precision and efficiency is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a natural water body adaptive voxel grid planning method, device and storage medium. BACKGROUND

[0002] When performing water body inspection, the monitoring area is usually divided according to a preset uniform grid resolution, and a fixed monitoring profile and a navigation path are determined. For key areas with high incidence of water quality abnormalities and hidden dangers in historical monitoring, the monitoring density is not targeted to be improved, resulting in insufficient capture of high-value risk information and difficulty in accurately locating abnormalities. For low-risk areas with stable water quality, no historical abnormal records, and gentle hydrodynamic conditions, the same monitoring resources are still invested according to the uniform standard, causing invalid waste of computing power consumption and carrier energy consumption. The resource allocation cannot be optimized according to the regional differences, resulting in high energy consumption cost, insufficient coverage accuracy of key areas, and difficulty in improving the overall monitoring timeliness.

[0003] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0004] The main purpose of the present application is to provide a natural water body adaptive voxel grid planning method, device and storage medium, aiming to solve the technical problem of how to construct a differentiated voxel grid, thereby improving the efficiency and accuracy of water body inspection.

[0005] In order to solve the above problems, the present application provides a natural water body adaptive voxel grid planning method, which comprises: dividing a three-dimensional terrain model of a target water body based on a preset initial resolution to obtain an initial voxel grid; inputting real-time water quality data, flow rate data, historical abnormal data and carrier energy consumption parameters of each initial voxel grid into a preset risk field function to obtain a risk level of each initial voxel grid; adjusting the size of the initial voxel grid based on a target resolution corresponding to the risk level to obtain a target voxel grid.

[0006] In an embodiment, the step of inputting real-time water quality data, flow rate data, historical abnormal data and carrier energy consumption parameters of each initial voxel grid into a preset risk field function to obtain a risk level of each initial voxel grid comprises: based on the voxel grid identifier of the initial voxel grid, obtaining a real-time water quality abnormal coefficient corresponding to the real-time water quality data, a historical abnormal probability corresponding to the historical abnormal data, a flow rate shear coefficient corresponding to the flow rate data, and an energy consumption coefficient corresponding to the carrier energy consumption parameter; inputting the real-time water quality anomaly coefficient, the historical anomaly probability, the flow rate shear coefficient and the energy consumption coefficient into the risk field function for weighted summation to obtain the risk value of the initial voxel grid; comparing the risk value with a preset risk level division rule to obtain the risk level.

[0007] In an embodiment, the step of obtaining the real-time water quality anomaly coefficient corresponding to the real-time water quality data, the historical anomaly probability corresponding to the historical anomaly data, the flow rate shear coefficient corresponding to the flow rate data, and the energy consumption coefficient corresponding to the carrier energy consumption parameter comprises: determining the deviation rate between the real-time water quality data and preset water quality standard data as the real-time water quality anomaly coefficient; obtaining the historical anomaly times and the total monitoring times associated with the initial voxel grid, and determining the historical anomaly probability according to the historical anomaly times and the total monitoring times; obtaining the three-dimensional flow rate vector associated with the initial voxel grid, and determining the maximum absolute value of the partial derivative of the three-dimensional flow rate vector as the flow rate shear coefficient; obtaining the distance between the initial voxel grid and a preset voxel center point, and determining the product of the distance and the energy consumption per unit distance as the energy consumption coefficient.

[0008] In an embodiment, the step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain a target voxel grid comprises: if the initial voxel is a high-risk voxel, the initial resolution is reduced by a preset multiple to obtain the target resolution; if the initial voxel is a medium-risk voxel, the initial resolution is determined as the target resolution; if the initial voxel is a low-risk voxel, the initial resolution is expanded to a preset multiple to obtain the target resolution; adjusting the initial voxel grid based on the target resolution to obtain the target voxel grid.

[0009] In an embodiment, after the step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain a target voxel grid, the adaptive voxel grid planning method for natural water bodies further comprises: performing path planning based on a preset global planning algorithm and the target voxel grid to obtain a global path; adjusting the global path based on a preset local path planning algorithm and a real-time flow rate gradient to obtain a target path.

[0010] In one embodiment, the step of adjusting the global path based on a preset local path planning algorithm and real-time flow velocity gradient to obtain the target path includes: The area within the preset range is defined as the local planning range. Based on the RRT algorithm, local path planning is performed on the local planning range to obtain the initial local path. If the velocity gradient corresponding to the path node in the initial local path is less than or equal to a preset velocity gradient threshold, the deviation between the initial local path and the global path is less than or equal to a preset deviation threshold, and the local energy consumption increment is less than or equal to a preset local energy consumption increment threshold, then the initial local path is determined to be the target local path. The global path is adjusted based on the target local path to obtain the target path.

[0011] In one embodiment, after the step of adjusting the global path based on a preset local path planning algorithm and real-time flow velocity gradient to obtain the target path, the adaptive voxel grid planning method for natural water bodies further includes: When a new high-risk voxel is detected, and / or the current energy consumption is less than or equal to the preset remaining energy threshold, the global path update process is triggered to obtain the updated global path. When an unmarked obstacle is detected, the local path update process is triggered to obtain the updated local path.

[0012] In one embodiment, after the step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain the target voxel grid, the adaptive voxel grid planning method for natural water bodies further includes: Based on real-time collected water quality index data and a preset water quality anomaly coefficient quantification formula, the water quality anomaly coefficient of each target voxel grid is determined. If the water quality anomaly coefficient is greater than the preset first water quality anomaly coefficient threshold, an anomaly warning is triggered, and the target voxel grid is marked as a highly anomalous voxel. Based on the risk level of the highly anomalous voxels, the revisit cycle is matched from the revisit cycle mapping table, and the revisit tasks are generated and added to the priority queue to obtain the revisit task list.

[0013] Furthermore, to achieve the above objectives, this application also proposes an adaptive voxel grid planning device for natural water bodies, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the adaptive voxel grid planning method for natural water bodies as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the adaptive voxel grid planning method for natural water bodies as described above.

[0015] This application provides an adaptive voxel grid planning method for natural water bodies. A three-dimensional terrain model is divided into a structured initial voxel grid according to a preset initial resolution. Then, multi-source data on real-time water quality, flow velocity, historical anomalies, and carrier energy consumption are integrated. The risk value and risk level of each initial voxel grid are quantified using a risk field function. Finally, the initial voxel grid is adjusted according to the risk level and a differentiated target resolution. This transforms dispersed, multi-dimensional risk information into precise voxel-level risk assessment results, making risk distribution visible and quantifiable. The adaptive resolution adjustment, which refines high-risk areas and coarsens medium- and low-risk areas, ensures monitoring accuracy in high-risk areas while reducing computational and energy costs in low-risk areas, achieving a balance between accuracy and efficiency. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the adaptive voxel grid planning method for natural water bodies in this application; Figure 2 A schematic diagram of the overall system architecture provided for the adaptive voxel grid planning method for natural water bodies in this application; Figure 3 A schematic diagram illustrating the adaptive voxel grid adjustment provided for the adaptive voxel grid planning method for natural water bodies in this application; Figure 4 A schematic diagram of the revisit scheduling timeline provided for the adaptive voxel grid planning method for natural water bodies in this application; Figure 5 A schematic diagram illustrating the data binding and export process provided for the adaptive voxel grid planning method for natural water bodies in this application; Figure 6 This is a schematic diagram of the hardware operating environment involved in the adaptive voxel grid planning method for natural water bodies in the embodiments of this application.

[0019] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0021] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] To achieve the above objectives, this application proposes an adaptive voxel grid planning method for natural water bodies. The method involves dividing the three-dimensional terrain model of the target water body based on a preset initial resolution to obtain an initial voxel grid. Real-time water quality data, flow velocity data, historical anomaly data, and carrier energy consumption parameters of each initial voxel grid are input into a preset risk field function to obtain the risk level of each initial voxel grid. Based on the target resolution corresponding to the risk level, the size of the initial voxel grid is adjusted to obtain the target voxel grid.

[0023] During water body inspections, the entire area is typically divided into several equal-area monitoring units based on the geographical boundaries of the target water area and a pre-set uniform grid resolution. For monitoring units with stable water quality, no historical anomalies, and mild hydrodynamic conditions, as well as those with frequent water quality anomalies and historical risks, the same testing frequency and sampling point density are set according to uniform standards. This results in the collection of a large amount of normal water quality data, leading to ineffective waste of computational power for water anomaly detection algorithms and energy consumption of inspection equipment.

[0024] This application proposes an adaptive voxel grid planning method for natural water bodies. A three-dimensional terrain model is divided into a structured initial voxel grid according to a preset initial resolution. Then, multi-source data on real-time water quality, flow velocity, historical anomalies, and carrier energy consumption are integrated. The risk value and risk level of each initial voxel grid are quantified using a risk field function. Finally, the initial voxel grid is adjusted according to the risk level and a differentiated target resolution. This transforms dispersed, multi-dimensional risk information into precise voxel-level risk assessment results, making risk distribution visible and quantifiable. The adaptive resolution adjustment, which refines high-risk areas and coarsens medium- and low-risk areas, ensures monitoring accuracy in high-risk areas while reducing computational and energy costs in low-risk areas, achieving a balance between accuracy and efficiency.

[0025] It should be noted that the executing entity in this embodiment can be a computing service device with network communication and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or apparatus capable of performing the above functions. The following description uses an adaptive voxel grid planning device for natural water bodies as an example to illustrate this embodiment and the subsequent embodiments.

[0026] Based on this, embodiments of this application provide an adaptive voxel grid planning method for natural water bodies, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the adaptive voxel grid planning method for natural water bodies in this application.

[0027] In this embodiment, the adaptive voxel grid planning method for natural water bodies is applied to an adaptive voxel grid planning device for natural water bodies, and the method includes steps S10~S30: Step S10: Divide the three-dimensional terrain model of the target water body based on the preset initial resolution to obtain the initial voxel mesh.

[0028] In this embodiment, the spatial range of the 3D model is divided into uniform cubes, i.e., voxels, according to the rule resolution; each voxel corresponds to a unique spatial location, realizing the discretization and structuring of the 3D space.

[0029] Read the spatial boundary parameters of the 3D terrain model to determine the 3D coordinate range of the modeling area: X-axis range, Y-axis range, and Z-axis range. Extend outwards by one voxel edge length as a buffer zone to prevent terrain edges from being truncated. Based on the preset initial voxel resolution Δ0 (Δx×Δy×Δz), calculate the number of voxels in the voxel mesh, where the number of voxels in the X direction is Nx = ceil ((X_max-X_min) / Δx), the number of voxels in the Y direction is Ny = ceil ((Y_max-Y_min) / Δy), and the number of voxels in the Z direction is Nz = ceil ((Z_max - Z_min) / Δz), where ceil is the floor function. Assign a unique identifier ID to each voxel, with each ID corresponding one-to-one with the voxel's spatial coordinate range. Based on the calculated number of voxels and coordinate range, generate a uniformly distributed voxel mesh in 3D space, with each voxel recording its own ID and X / Y / Z coordinate range. Check if each voxel intersects with the 3D terrain model, mark voxels containing terrain and empty voxels, ensure that the voxel mesh completely covers the terrain area without redundancy; output the final 3D terrain model and the initial uniform voxel mesh.

[0030] Please refer to Figure 2 , Figure 2A system architecture diagram is provided. The cloud computing platform undertakes core tasks such as modeling and indexing optimization, automatic knowledge reasoning, element association and tracing, risk calculation, heterogeneous data governance, and interpretability adaptation. The shore station control terminal is responsible for task interpretation and verification and situational awareness display. The underwater ROV, as the execution terminal, integrates multimodal sensors (such as sonar, lidar, water quality sensors, etc.), control and navigation modules (such as propulsion control, navigation and positioning, communication), and execution units (sampling, robotic arm). The three are connected through public / private networks and data links to realize the interaction of task commands and monitoring data, forming an integrated underwater inspection system of cloud-based intelligent decision-making, shore station situational control, and underwater vehicle execution.

[0031] Step S20: Input the real-time water quality data, flow velocity data, historical anomaly data, and carrier energy consumption parameters of each initial voxel grid into a preset risk field function to obtain the risk level of each initial voxel grid.

[0032] In this embodiment, four heterogeneous risk factors—real-time water quality data, historical anomaly data, flow velocity data, and carrier energy consumption parameters—reflect real-time environmental safety, historical risk correlation, dynamic environmental interference, and carrier operating costs, respectively. These risk factors are standardized and quantified, then integrated into a unified voxel risk value using preset weighting coefficients. Risk levels are then categorized based on thresholds, ultimately forming a visualized spatiotemporal risk field distribution. Real-time water quality data directly reflects the current pollution state of the water body, signs of algal blooms, and other immediate risks. The risk field function quantifies the deviation of these data from water quality standards, quickly identifying high-risk voxels. Flow velocity data determines the diffusion speed and range of pollutants; for example, high-velocity shearing easily leads to rapid spread of pollution plumes, making it a key environmental factor for judging the evolution trend of abnormal areas. Historical anomaly data reflects both inherent regional risks and avoids the invalidation of long-term data due to environmental changes. Distance is a core factor affecting carrier energy consumption; combined with the equipment's rated energy consumption parameters, the energy cost of monitoring this voxel can be quantified.

[0033] Specifically, step S20 includes steps S21 to S23: Step S21: Based on the voxel grid identifier of the initial voxel grid, obtain the real-time water quality anomaly coefficient corresponding to the real-time water quality data, the historical anomaly probability corresponding to the historical anomaly data, the flow velocity shear coefficient corresponding to the flow velocity data, and the energy consumption coefficient corresponding to the carrier energy consumption parameter.

[0034] Step S22: Input the real-time water quality anomaly coefficient, the historical anomaly probability, the flow velocity shear coefficient, and the energy consumption coefficient into the risk field function and perform a weighted summation to obtain the risk value of the initial voxel grid.

[0035] Step S23: Compare the risk value with the preset risk level classification rules to obtain the risk level.

[0036] In this embodiment, the preset weighting coefficients reflect the contribution of each factor to the risk and can be adjusted according to the monitoring objectives. The importance of different risk factors is reflected through weight allocation, with core factors assigned higher weights and secondary factors assigned lower weights, ultimately merging into a single risk value to achieve a comprehensive assessment of multi-dimensional risk. The risk value of each voxel is calculated according to the preset formula: f = α×f1 + β×f2 + γ×f3 + δ×f4, where α, β, γ, and δ are preset weights. Based on the preset risk level classification rules, the coefficients f1 to f4 of each voxel are substituted into the formula to calculate the corresponding risk value.

[0037] Iterate through the risk values ​​of each voxel and classify them into levels according to the threshold. For example, high risk: f≥0.7, the overall risk is extremely high and should be monitored or avoided first; medium risk: 0.3<f<0.7, the overall risk is moderate and should be monitored as needed; low risk: f≤0.3, the overall risk is relatively low and the monitoring frequency can be reduced; add a risk level label to each voxel.

[0038] Using a voxel grid as the spatial framework, the risk value f of each voxel is mapped to its 3D coordinate position. A heatmap color system is adopted, with high-risk voxels marked in red, medium-risk in yellow, and low-risk in green; or a gradient color scheme from 0 to 1, with darker colors indicating higher risk. A timestamp t for the current calculation is added to the heatmap to clarify the temporal attribute of the risk field. The heatmap is exported to a standard visualization format, and an interactive 3D risk field model is generated, supporting rotation and zooming to view the risk distribution in different areas.

[0039] In one feasible implementation, step S21 may include: The following steps are taken: 1. Obtain the real-time water quality data associated with the initial voxel grid; 2. Determine the deviation rate between the real-time water quality data and the preset water quality standard data as the real-time water quality anomaly coefficient; 3. Obtain the historical anomaly count and total monitoring count associated with the initial voxel grid; 4. Determine the historical anomaly probability based on the historical anomaly count and the total monitoring count; 5. Obtain the three-dimensional velocity vector associated with the initial voxel grid; 6. Determine the maximum absolute value of the partial derivative of the three-dimensional velocity vector as the velocity shear coefficient; 7. Obtain the distance between the initial voxel grid and the preset voxel center point; 8. Determine the energy consumption coefficient by multiplying the distance by the energy consumption per unit distance.

[0040] In this embodiment, core information for each voxel is extracted to construct a voxel information table, including voxel ID, X / Y / Z coordinate range, and voxel center point coordinates (x0, y0, z0). Water quality data is then organized by collection timestamp and collection location coordinates to form a water quality data record table, including time t, collection coordinates (x, y, z), pH value, DO value (dissolved oxygen), and ORP value (oxidation-reduction potential). Flow velocity data is also organized by timestamp and collection location coordinates to form a flow velocity data record table, including time t, collection coordinates (x, y, z), and a three-dimensional flow velocity vector (u, v, w). All historical anomaly records are extracted, and each anomaly is associated with its corresponding voxel based on its coordinate range. The number of past anomalies and the total number of monitoring times for each voxel are then counted. Carrier parameters record the carrier's current location coordinates (x_c, y_c, z_c) and energy consumption per unit distance E, where E is a preset fixed value. Based on the coordinates of the voxel center point, water quality data and flow velocity data are assigned to the corresponding voxels according to the intersection of the collection coordinates and the range of voxel coordinates. That is, a data collection point is associated with the voxel ID if it falls within the coordinate range of a certain voxel. Based on the latest timestamp t of the real-time data, historical abnormal data and carrier location data are time-aligned. If a voxel has no real-time water quality / flow velocity data, the average value of neighboring voxels is used to fill the gap. For example, the average value of the corresponding data of the eight adjacent voxels around the voxel is taken to avoid interruption of risk calculation due to missing data.

[0041] For the real-time water quality anomaly coefficient f1, the greater the deviation of the water quality indicator from the normal range, the greater the risk coefficient. Real-time water quality data linked to voxels is read, and values ​​are assigned according to preset rules. For example, if DO < 5 mg / L (abnormal), then f1 = 0.8; if DO ≥ 5 mg / L (normal), then f1 = 0.2. The anomaly coefficient for each indicator is calculated separately, and then the average of the coefficients is taken as the final f1 to ensure comprehensive evaluation of multiple water quality indicators. The real-time water quality anomaly coefficient f1 can also be calculated by the deviation rate between the measured and standard values ​​of water quality parameters, such as f1 = 1 - (measured DO / standard DO), where standard DO is set according to the "Surface Water Environmental Quality Standard," such as Class III water with DO ≥ 5 mg / L.

[0042] For the historical anomaly probability f2, the higher the frequency of past anomalies of a voxel, the greater the current risk probability. Optionally, the past anomaly count N1 and total monitoring count N2 of the corresponding voxel can be extracted from the historical anomaly database; the anomaly probability f2 = N1 / N2 can be calculated. Alternatively, the historical anomaly probability can be calculated based on Bayes' theorem: f2 = P(anomaly | voxel) = (historical anomaly count of that voxel + 1) / (total monitoring count of that voxel + 2), avoiding zero probability.

[0043] For the velocity shear coefficient f3, the larger the velocity gradient, the worse the stability of the carrier operation and the greater the risk factor. Read the three-dimensional velocity vector (u, v, w) associated with the voxel. Based on the velocity difference between adjacent voxels, calculate the partial derivatives of the velocity in the X, Y, and Z directions, and take the maximum absolute value of each partial derivative as the velocity gradient of the voxel. Based on the preset threshold rule, determine the velocity shear coefficient corresponding to the velocity gradient.

[0044] For the energy consumption coefficient f4, the greater the distance between the voxel and the current position of the carrier, the higher the total energy consumption accumulated per unit distance E, and the greater the risk of operating costs. The Euclidean distance D between the voxel center point (x0, y0, z0) and the current position of the carrier (x_c, y_c, z_c) is calculated. The original energy consumption value E_raw is calculated using the formula: E_raw = D × E. A voxel and risk factor quantification table is obtained, with each voxel ID corresponding to four standardized coefficients: f1, f2, f3, and f4.

[0045] Step S30: Based on the target resolution corresponding to the risk level, adjust the size of the initial voxel grid to obtain the target voxel grid.

[0046] In this embodiment, the voxel mesh in high-risk areas is refined, the voxel mesh in medium-risk areas is kept at the baseline, and the voxel mesh in low-risk areas is coarsened and merged. The spatial granularity of the voxels is adjusted by resolution scaling. At the same time, the total energy consumption after adjustment is verified to be compliant with the carrier energy consumption budget as a constraint, and finally a multi-resolution voxel mesh with accurate risk matching and controllable energy consumption is formed.

[0047] Please refer to Figure 3 , Figure 3A schematic diagram of voxel grid adaptive adjustment is provided. The left side shows the initial uniform voxel grid, and the right side shows the adjusted grid. After adjustment, high-risk voxels are refined into small cubes, and low-risk voxels are merged into large cubes. Voxel IDs, risk levels, and resolutions are labeled. Risk levels are distinguished by color: red represents high risk, yellow represents medium risk, and blue represents low risk. When a voxel is detected as a high-risk voxel with frequent water quality anomalies and historical risks, it is split into smaller sub-grids according to preset subdivision rules, such as reducing the side length to 1 / 2 or 1 / 4 of the original grid. This improves the monitoring resolution of the area, facilitating the capture of subtle changes and spatial evolution characteristics of water quality, and ensuring refined detection of anomaly areas. For low-risk voxels with stable water quality, no historical anomaly records, and mild hydrodynamic conditions, multiple adjacent small voxels are merged into a larger coarser voxel, reducing the number of invalid grids and monitoring points, thereby reducing computational power consumption and energy waste of inspection equipment, achieving efficient monitoring of stable areas. Through dynamic adjustment, the voxel grid can be adaptively optimized according to the actual monitoring needs of the area, which can significantly improve the monitoring efficiency of stable areas while ensuring the monitoring accuracy of abnormal areas and avoid the ineffective waste of resources.

[0048] Specifically, step S30 includes steps S31 to S34: Step S31: If the initial voxel is a high-risk voxel, then reduce the initial resolution by a preset factor to obtain the target resolution.

[0049] Step S32: If the initial voxel is a medium-risk voxel, then the initial resolution is determined as the target resolution.

[0050] Step S33: If the initial voxel is a low-risk voxel, the initial resolution is increased to a preset multiple to obtain the target resolution.

[0051] Step S34: Adjust the initial voxel grid based on the target resolution to obtain the target voxel grid.

[0052] In this embodiment, voxels are divided into three groups according to risk level: high-risk voxel set G1, medium-risk voxel set G2, and low-risk voxel set G3. The number of voxels (N1, N2, N3) and the total spatial coverage of each group are counted to ensure that the spatial range of voxels of the same risk level does not overlap or omit any, and that the total range of the three groups of voxels completely covers the original terrain area, thus obtaining a risk level and voxel grouping table.

[0053] The initial voxels are adjusted based on the risk level and voxel grouping table, as well as the preset resolution adjustment rules. Optionally, the resolution adjustment rules are as follows: for high-risk voxels, the resolution is refined to Δ1 = Δ0 / 2; a new ID is assigned to each refined voxel, the new ID being the original voxel ID + the split sequence number, and a new coordinate range is determined. After splitting, the X ranges of the two refined voxels are [X1, X1 + Δx1] and [X1 + Δx1, X2], respectively. The risk level of the original voxel is inherited, and the resolution Δ1 is recorded. For medium-risk voxels, the initial resolution Δ0 is maintained; the resolution, coordinate range, and voxel ID of all voxels are directly inherited from the original parameters, and the risk level label is retained. For low-risk voxels, the resolution Δ2 = 2 × Δ0; voxel grid rearrangement: the spatial range of the low-risk voxel set is re-divided into a uniform grid according to the resolution Δ2, forming a new merged voxel frame; multiple original low-risk voxels falling within the same new merged voxel frame are merged into one new voxel; a new ID is assigned to each merged voxel, the coordinate range after merging is determined, the risk level is inherited, and the resolution Δ2 is recorded. Through differentiated resolution adjustment, the voxel grid resolution is matched with the voxel risk level. High-risk areas receive a finer grid to ensure monitoring accuracy, while low-risk areas have a coarser grid to reduce resource consumption, balancing monitoring accuracy and resource utilization. This avoids the problem of insufficient coverage in high-risk areas and resource waste in low-risk areas under a uniform resolution, optimizes computing power allocation and carrier energy consumption, and improves overall inspection efficiency.

[0054] Optionally, the feasibility of the preliminary adjustment results is verified using an energy consumption budget as a constraint to avoid exceeding energy consumption limits due to unrestricted refinement of high-risk voxels, ensuring that the adjustment strategy conforms to actual operating conditions. Allocation of corresponding unit voxel coverage energy consumption is made for voxels of different resolutions: E_v1 = 4 × E_v0 for high-risk voxels, E_v0 = E_v0 for medium-risk voxels, and E_v3 = E_v0 / 4 for low-risk voxels. Total energy consumption is calculated by group. If the energy consumption budget ≤ total energy consumption: the preliminary adjustment results are compliant and no further optimization is needed; if the energy consumption budget > total energy consumption: the preliminary adjustment energy consumption exceeds the limit, and the merging ratio of low-risk voxels needs to be reduced.

[0055] In this embodiment, the 3D terrain model is divided into a structured initial voxel grid according to a preset initial resolution. Then, multi-source data such as real-time water quality, flow velocity, historical anomalies, and carrier energy consumption are integrated. The risk value and risk level of each initial voxel grid are quantified through a risk field function. Finally, the initial voxel grid is adjusted according to the risk level and the differentiated target resolution. The dispersed multi-dimensional risk information is transformed into voxel-level accurate risk assessment results, making the risk distribution visible and quantifiable. The adaptive resolution adjustment of high-risk areas is refined and medium- and low-risk areas are coarsened. While ensuring the monitoring accuracy of high-risk areas, the computation and energy consumption costs of low-risk areas are reduced, achieving a balance between accuracy and efficiency.

[0056] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Furthermore, before step S10, steps S01 to S04 may also be included: Step S01: Obtain initial point cloud data of the target water body collected by underwater machinery and equipment, and preprocess the initial point cloud data to obtain target point cloud data.

[0057] In this embodiment, the original observation data of the terrain point cloud sensor is used as the basis. The point cloud coordinate deviation is corrected by the high-precision spatiotemporal reference provided by GNSS / IMU. Then, abnormal noise points are removed based on statistical characteristics and geometric features. Finally, a terrain point cloud dataset with both position accuracy and data validity is output.

[0058] The system controls underwater vehicles (ROVs / AUVs) to perform initial terrain scanning in the target water area, collecting terrain point cloud data. It acquires raw point cloud data output from terrain point cloud sensors and spatiotemporal data output from GNSS / IMU positioning systems in real time via hardware interfaces (such as Ethernet and serial ports). The spatiotemporal data includes the vehicle's real-time latitude, longitude, elevation, attitude angle, and velocity. Based on a timestamp alignment mechanism, using GNSS time as a reference, the sensor data is time-synchronized and calibrated, binding point cloud observations and spatiotemporal data at the same moment into a raw data set. Following standardized data formats, such as LAS point cloud format and JSON spatiotemporal data format, the raw data set is stored in a temporary buffer, resulting in a time-synchronized raw point cloud and spatiotemporal data binding set. Simultaneously, data acquisition logs are recorded. Timestamp alignment resolves the spatiotemporal asynchrony issue of multi-source sensor data, providing spatiotemporal reference data for subsequent coordinate calibration and ensuring that each point cloud observation is matched with the corresponding vehicle position and attitude information.

[0059] The algorithm converts the carrier coordinate system to the geodetic coordinate system, using the spatiotemporal parameters of the bound GNSS / IMU as input. First, it converts the carrier's real-time position and attitude to an absolute reference in the geodetic coordinate system. Then, combining the sensor installation and calibration parameters, it converts the relative coordinates of each observation point in the original point cloud (i.e., based on the sensor's own coordinate system) to absolute geographic coordinates (i.e., latitude, longitude, and elevation) in the geodetic coordinate system. Consistency checks are performed on the converted coordinate data, such as calculating the coordinate deviation between adjacent points in the same area. If the deviation exceeds a preset threshold, the time synchronization accuracy or installation calibration parameters are checked back. This results in a calibrated point cloud dataset, where each point contains geodetic coordinates, reflection intensity, and other information.

[0060] The initial point cloud dataset is divided into grids, with the grid size set according to the scanning resolution to ensure that each grid contains a certain number of neighboring points, providing a range for local statistical analysis. For the point cloud within each grid, the mean μ and standard deviation σ of its spatial coordinates are calculated. The coordinate value of each point is compared with the statistical mean of its grid. If the deviation value (|x-μ|) exceeds the preset denoising threshold, it is identified as a noise point and removed from the dataset. Connectivity analysis is performed on the noise-removed point cloud to remove isolated small clusters of points, such as discrete points with fewer than 3 neighboring points, to further purify the data. The target point cloud dataset after denoising by Deao is shown.

[0061] Step S02: Based on the spatial topology construction algorithm, construct spatial indexes for each point cloud in the target point cloud dataset to determine the spatial relationships of the point clouds.

[0062] In this embodiment, the KD-tree (K-Dimensional Tree) algorithm is used to construct a spatial index for all point clouds. All point clouds are organized into an array based on the structure of point ID + X coordinate + Y coordinate + Z coordinate. The dimensional division order is set to cycle from the X-axis to the Y-axis and then to the Z-axis. A threshold T is set for leaf nodes, which is the maximum number of point clouds allowed in each leaf node. The root node of the KD-tree is initialized using the entire point cloud array `Points` as its initial data. This results in a formatted point cloud array and an initialized root node.

[0063] The root node is first divided along the first dimension (X-axis, d=0); its children are divided along the second dimension (Y-axis, d=1); and its children are divided along the third dimension (Z-axis, d=2). The next level node reverts to the first dimension (X-axis), and this process is repeated, with the dimension index d = (current tree depth) mod 3. The coordinate values ​​corresponding to dimension d are extracted from the current node's point cloud array, forming a subset of coordinates. This subset is sorted in ascending order, and the median position is found. The coordinate value corresponding to the median is used as the partition threshold, and the point corresponding to the median is used as the partition point. The current node's point cloud array is divided into two groups based on the relationship between the coordinate values ​​of dimension d and the partition threshold: the left child node's point cloud contains points whose coordinate values ​​are less than the partition threshold; the right child node's point cloud contains points whose coordinate values ​​are greater than or equal to the partition threshold. The partition points are stored in the current node as boundary markers for the subspace. If the number of points in the left child node after the current node is divided is less than or equal to the threshold T for leaf nodes, then the left child node is a leaf node, and all points in the left child node's point cloud are directly stored without further division; the processing logic for the right child node is the same as that for the left child node, and will not be elaborated here.

[0064] For intermediate nodes, the partitioning dimension d, partitioning threshold, partitioning point, left child node pointer, and the 3D boundary of the subspace corresponding to that node are stored; for leaf nodes, the node type identifier, 2. all point cloud data contained in that node, and the 3D boundary of the subspace corresponding to that leaf node are stored, resulting in a complete 3D KD tree structure, including the hierarchical index of the root node, intermediate nodes, and leaf nodes, with each node associated with a specific subspace range and point cloud data.

[0065] For each target point, its neighborhood is located using a tree. With P as the center, a neighborhood radius R is set, and the neighborhood radius is 1.5 to 2 times the scanning resolution. The leaf nodes of the KD tree are traversed, and all points that fall within the spherical space with P as the center and R as the radius are selected to form the neighborhood point set N(P) of P, which represents the spatial relationship between point clouds.

[0066] Step S03: Perform local plane fitting based on the spatial correlation, and determine the spatial coordinates and normal vector of the point cloud based on the fitting result.

[0067] In this embodiment, the mean coordinate value (μ) of the neighborhood point set N(P) is calculated. x μ (μ_z), and Standard deviation of the coordinates of the neighborhood point set (σ) x , σ , σ_z); Calculate the deviation between the initial coordinates of point P and the mean of its neighborhood: ΔX=|X0-μ x |,ΔY=|Y0-μ |, ΔZ = |Z0 - μ_z|; Set the deviation threshold, such as 2σ, that is, exceeding 2 times the standard deviation is abnormal, if ΔX ≤ 2σ x ΔY≤2σ If ΔZ ≤ 2σ_z, it means the coordinates of point P conform to neighborhood consistency, and the initial coordinates are retained. If any dimension deviates beyond the standard, it is determined to be a coordinate outlier, and the coordinates of point P are corrected using the neighborhood mean: X = μ x Y=μ Z = μ_z; or, a weighted average is used, with the weight of neighboring points being greater the closer they are to P, thus improving the accuracy of the correction; the precise spatial coordinates (X, Y, Z) of each point P are obtained, and by correcting abnormal offsets, the spatial distribution of all points is ensured to conform to the terrain continuity.

[0068] Using point P as the origin, transform the coordinates of the neighborhood point set N(P) into local coordinates (x, y). i =X i -X, y i =Y i -Y, z i =Z i-Z); Assume the local terrain surface conforms to the plane equation: ax + by + cz = 0, where (a, b, c) are the normal vectors of the plane; solve by least squares optimization to minimize the sum of squared distances from all neighboring points to the plane, i.e., minΣ(ax i + by i +cz i ) 2 By taking the partial derivatives with respect to a, b, and c and setting them to 0, the problem is transformed into solving for the eigenvalues ​​and eigenvectors of the covariance matrix. The covariance matrix C of the local coordinates of the neighborhood points is calculated, and eigenvalue decomposition is performed on the covariance matrix C to obtain the eigenvalues ​​(λ1≥λ2≥λ3) and the corresponding eigenvectors (v1, v2, v3), thus obtaining the normal vector of each point P.

[0069] Step S04: Input the spatial coordinates and the normal vector into a preset surface reconstruction algorithm to obtain the three-dimensional terrain model.

[0070] In this embodiment, the Poisson surface reconstruction algorithm is invoked, using the spatial coordinates and normal vectors of the point cloud as input, and a continuous three-dimensional terrain surface model is generated through implicit function fitting. The Poisson algorithm treats the point cloud as sampling points in three-dimensional space, and constructs a smooth implicit function by solving the Poisson equation. The zero isosurface of the implicit function is the terrain surface, ensuring the continuity and smoothness of the model. Based on the characteristics of water body terrain, model boundary constraints are set, such as limiting the model boundary and elevation constraints according to the range of the collection area. The average distance deviation between the reconstructed model and the original point cloud is calculated. If the deviation exceeds the standard, the neighborhood radius or algorithm parameters are adjusted, and the model is reconstructed to obtain the three-dimensional terrain model.

[0071] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Furthermore, after step S30, steps S40 to S60 may also be included: Step S40: Path planning is performed based on the preset global planning algorithm and the target voxel mesh to obtain the global path.

[0072] In this embodiment, a planning voxel table is constructed based on the information of the target voxel mesh, including voxel ID, three-dimensional coordinate range, risk level, resolution, center point coordinates, etc.; the coordinate range of the preset no-navigation surface is converted into a no-navigation voxel set, and all voxels overlapping with the no-navigation surface are marked as impassable, while the coordinate range of the no-navigation voxels is recorded to avoid path crossing; real-time flow velocity vectors (u, v, w) are added to each voxel according to the voxel coordinate range association for subsequent local path attitude stability calculation; the current position of the carrier, maximum sailing speed, turning radius, and energy consumption characteristics are recorded.

[0073] The set of center points of all high-risk voxels is defined as the target domain, ensuring that the path covers this set. All voxels except those in no-fly zones are defined as feasible domains, while no-fly zones are defined as infeasible domains. The preset path planning constraints are: the path cannot cross no-fly zones; the path must maximize coverage of high-risk voxels; total energy consumption ≤ remaining energy consumption of the vehicle; vehicle attitude angles ≤ safety thresholds, including roll, pitch, and yaw.

[0074] In one feasible implementation, the A algorithm is used to prioritize the search for the optimal path direction by evaluating the actual cost from the starting point to the current node and the estimated cost from the current node to the target node. Combined with multi-objective optimization weights, a balance between coverage and energy consumption is achieved.

[0075] Specifically, the voxel center point (x0, y0, z0) is used as the path node, and each node is associated with the voxel's risk level, energy consumption parameters, and flow velocity data; the cost function is f(n) = g(n) + h(n), where g(n) is the actual cost from the current position of the carrier to node n: g(n) = ω1 × E(n) + ω2 × (1 - C(n)), E(n) is the energy consumption from S to n, C(n) is the high-risk coverage contribution corresponding to n, if n is a high-risk voxel, C(n) = 1; otherwise C(n) = 0, ω1 is the energy consumption weight, ω2 is the coverage weight; h(n) is the estimated cost from node n to the target domain.

[0076] Create an open list to store nodes to be searched and a closed list to store nodes already searched. Add the starting point S to the open list, g(S) = 0, h(S) = the estimated minimum distance to T, and f(S) = g(S) + h(S). Select the node with the smallest f(n) from the open list as the current node n and move it to the closed list. Generate the neighboring nodes of the current node n. The neighboring nodes are the center points of adjacent voxels, including the six directions of up, down, left, right, front, and back, excluding nodes corresponding to prohibited voxels. For each neighboring node m, if m is in the closed list, skip it; if m is not in the open list: calculate g(m) = g(n) + S(n, m), h(m), and f(m), and add m to the open list; where S(n, m) is the cost from n to m, i.e., energy consumption × ω1 + coverage contribution × ω2; if m is already in the open list, compare the newly calculated g(m) with the original g(m). If the new g(m) is smaller, update m's g(m) and f(m). (m), and set the parent node of m to n; when the last uncovered high-risk voxel node of the target domain appears in the open list, or the open list is empty, backtrack the parent node chain to generate a global path from S to cover all high-risk voxels. Perform Bézier curve fitting on the global path to eliminate sharp angles in the path and ensure the stability of the carrier's navigation attitude.

[0077] Optionally, calculate the total energy consumption E_global of the global path. If E_global > the remaining energy consumption of the carrier, adjust the cost function weights and search for a new path. Confirm that the global path does not cross any restricted areas. If it does, adjust the path nodes to bypass the restricted areas.

[0078] Step S50: Based on the preset local path planning algorithm and real-time flow velocity gradient, the global path is adjusted to obtain the target path.

[0079] In this embodiment, a local planning range is set with the current position of the carrier as the center, such as a spherical area with a radius of 50m, or a range of 10 nodes before and after the current node on the global path; voxel data in the local area, such as flow velocity, risk level, whether navigation is prohibited, etc., are extracted as the spatial basis for local planning.

[0080] In one feasible implementation, step S50 may include steps S51 to S53: Step S51: Determine the area within the preset range as the local planning range, and perform local path planning on the local planning range based on the RRT algorithm to obtain the initial local path.

[0081] Step S52: When the velocity gradient corresponding to the path node in the initial local path is less than or equal to a preset velocity gradient threshold, the deviation between the initial local path and the global path is less than or equal to a preset deviation threshold, and the local energy consumption increment is less than or equal to a preset local energy consumption increment threshold, then the initial local path is determined to be the target local path.

[0082] Step S53: Adjust the global path based on the target local path to obtain the target path.

[0083] In this embodiment, local path planning is performed based on the RRT algorithm. The sampling space is the feasible region within the local area, and the target point is the next key node in the local area on the global path. The constraints are: the velocity gradient corresponding to the path node is ≤ a preset velocity gradient threshold to avoid attitude loss of control; the deviation between the local path and the global path is ≤ a preset deviation threshold to ensure that it does not deviate from the global target; and the local energy consumption increment is ≤ a preset local energy consumption increment threshold to avoid total energy consumption exceeding the limit.

[0084] The current position is taken as the root node of the RRT search tree, and the target point is taken as the local target. A sampling point is randomly generated in the local sampling space. The nearest first node is found in the search tree. A second node is extended from the first node towards the sampling point, where the extension step size is equal to the minimum turning radius of the carrier, such as 1m. The second node is checked to see if it is in the feasible region, if the velocity gradient is compliant, and if the deviation from the global path exceeds the limit. If it is compliant, the second node is added to the search tree, and the parent node is recorded as the first node. When the search tree extends to the local target point, or the distance from the target point is ≤ a preset distance, the local path is backtracked and generated. The latest velocity data and carrier attitude data are obtained once at a preset period. If a velocity change is detected, such as the velocity gradient > the preset velocity change threshold, the above steps are repeated to update the local path.

[0085] In one feasible implementation, the global path planning algorithm is the DLite algorithm, with the addition of dynamic obstacle states, where 0 = feasible, 1 = static no-navigation, 2 = dynamic obstacle, and risk coverage weights. The cost function is: f(n) = g(n) + h(n), where: g(n) is the actual cost from the starting point to node n = energy cost (distance × unit energy consumption) + obstacle penalty (dynamic obstacle node penalty = 1000, static no-navigation = ∞) + coverage contribution (high-risk node contribution = -0.5, encouraging coverage); h(n) is the estimated cost from node n to the target domain = Euclidean distance × risk weight. Starting from the current location of the carrier as S and the target domain as the set of endpoints T, a global cost map is constructed and a priority queue is initialized. An initial global path covering all high-risk voxels and avoiding static no-navigation surfaces is generated through a reversible search algorithm. Dynamic obstacle data is received in real time, and the corresponding voxels are marked as dynamic obstacle nodes, and the global cost map is updated. If a dynamic obstacle affects a key node of the global path, such as a path node becoming infeasible, a global replanning is triggered: only the node cost in the area affected by the obstacle is updated, the path is quickly adjusted through the priority queue, and the final global path is generated.

[0086] Optionally, the local path planning algorithm is an MPC (Model Predictive Control) algorithm. Using the continuous reference trajectory provided by the global path as the optimization objective, a state equation is first constructed based on the underwater vehicle's dynamic characteristics. A disturbance prediction model is then established by integrating real-time flow velocity data, clarifying multi-dimensional constraints such as attitude angle, thruster thrust, path deviation, and energy consumption. Subsequently, the rolling optimization logic of Model Predictive Control (MPC) is adopted, setting reasonable prediction and control time domains. By solving a constrained quadratic programming problem, control commands such as rudder angle and thrust at multiple future moments are optimized in real time, executing only the currently optimal command to drive the vehicle's movement. At fixed intervals, based on the latest vehicle state feedback and updated flow velocity data, the prediction and optimization process is iterated again, dynamically correcting the path to adapt to flow velocity disturbances, ultimately generating a continuous local path that conforms to the global main line, has stable attitude, and controllable energy consumption.

[0087] In one feasible implementation, after step A20, steps A30-A40 are further included: Step A30: When a new high-risk voxel is detected, and / or the current energy consumption is less than or equal to the preset remaining energy threshold, the global path update process is triggered to obtain the updated global path.

[0088] Step A40: When an unmarked obstacle is detected, the local path update process is triggered to obtain the updated local path.

[0089] In this implementation, new risk voxels, remaining energy consumption, and sudden obstacles are monitored in real time, and replanning trigger conditions are preset. When a critical change occurs that affects the effectiveness of the path, the corresponding replanning process is triggered to ensure that the path always meets the requirements of covering high risks, controlling energy consumption, and ensuring safety and feasibility.

[0090] Real-time monitoring and triggering conditions include: the appearance of new high-risk voxels; when real-time risk field calculations detect uncovered high-risk voxels, global replanning is triggered to re-cover the new risk area; when the remaining energy consumption is less than or equal to the preset remaining energy consumption threshold, global replanning is triggered to optimize the path to reduce energy consumption and prioritize the coverage of core high-risk areas; when the carrier sensor detects unknown obstacles in real time, such as sudden underwater floating objects whose coordinates are not in the preset no-navigation voxel set, confirming the existence of unpreset no-navigation obstacles, local replanning is triggered to bypass the sudden obstacles; if the local path cannot meet the attitude constraints after a preset number of consecutive adjustments, such as if the flow velocity continues to exceed the standard, global replanning is triggered to replan the global path and avoid high-interference areas.

[0091] After triggering global replanning, based on the A* algorithm, new high-risk voxels are added to update the target domain, the cost function is adjusted, and the optimal path is searched again to generate the updated global path. Local paths are also updated synchronously to ensure a smooth transition. After triggering local replanning, based on the RRT algorithm's planning range, sudden obstacles are marked as temporary no-fly voxels, and the local path is searched again. After bypassing the obstacle, the original global path is returned to. The final path is converted into a carrier-executable command format, such as the coordinates of each node, navigation speed, turning angle, and dwell time. The commands are sent to the carrier's execution layer, and the carrier's execution status feedback is received. If an execution anomaly occurs, local replanning is triggered again.

[0092] Based on the first embodiment of this application, in the fourth embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Furthermore, after step S30, steps B10 to B30 may also be included: Step B10: Based on the real-time collected water quality index data and the preset water quality anomaly coefficient quantification formula, determine the water quality anomaly coefficient of each target voxel grid.

[0093] Step B10: If the water quality anomaly coefficient is greater than the preset first water quality anomaly coefficient threshold, an anomaly warning is triggered, and the target voxel grid is marked as a highly abnormal voxel.

[0094] In this embodiment, monitoring data uploaded by the receiving carrier is associated with a voxel grid according to voxel ID and timestamp, and the risk level of the voxel is supplemented. The water quality anomaly coefficient quantification formula in the risk value calculation model is called to calculate the anomaly coefficient of each monitoring voxel. If the water quality anomaly coefficient is greater than a preset first water quality anomaly coefficient threshold, for example, water quality anomaly coefficient ≥ 0.9, an anomaly warning is triggered, and warning information such as voxel ID, risk level, real-time water quality anomaly coefficient, and monitoring timestamp is recorded, marking the voxel as a highly anomalous voxel. If the water quality anomaly coefficient is less than the first water quality anomaly coefficient threshold, only the monitoring data is stored. A high-severity anomaly voxel warning list is generated, which includes voxel ID, risk level, water quality anomaly coefficient, and warning timestamp.

[0095] In this embodiment, the carrier executes path instructions to complete the inspection of the designated area; during the inspection process, the sampling interface module controls the sampling device to perform sampling. When the water quality anomaly coefficient determined by a voxel based on the real-time data of the sampling is greater than the preset first water quality anomaly coefficient threshold, an anomaly warning is triggered. According to the risk level of the anomaly voxel, a revisit cycle is matched from the revisit cycle mapping table, and the revisit task is added to the priority queue, wherein the priority of high-risk revisit tasks is higher than that of medium / low-risk tasks.

[0096] The system receives monitoring data uploaded from the carrier, associates it with the voxel grid by voxel ID and timestamp, and supplements the risk level of the voxel. It then calls the water quality anomaly coefficient quantification formula in the risk value calculation model to calculate the anomaly coefficient for each monitored voxel based on real-time water quality indicators. If the water quality anomaly coefficient is greater than a preset first water quality anomaly coefficient threshold (e.g., ≥0.9), an anomaly warning is triggered, and warning information such as voxel ID, risk level, real-time water quality anomaly coefficient, and monitoring timestamp is recorded, marking the voxel as a high-severity anomaly voxel. If the water quality anomaly coefficient is less than the first water quality anomaly coefficient threshold, only the monitoring data is stored. A high-severity anomaly voxel warning list is generated, containing voxel ID, risk level, water quality anomaly coefficient, and warning timestamp.

[0097] Step B30: Based on the risk level of the highly anomalous voxels, match the revisit cycle from the revisit cycle mapping table, generate and add the revisit tasks to the priority queue to obtain the revisit task list.

[0098] For each voxel in the anomaly alert list, a revisit cycle is matched from the mapping table according to its risk level. For example, high-risk voxels are matched with a 4-hour revisit cycle, medium-risk with an 18-hour cycle, and low-risk with a 5-day cycle. A min-heap is used to store the revisit tasks. Elements in the heap are priority indicators and task information. The task information includes the voxel ID, risk level, revisit cycle, and last revisit time. The priority indicator is calculated as (current time - last revisit time) / revisit cycle; a higher ratio indicates a higher degree of expiration and thus a higher priority. The revisit task for each anomalous voxel is added to the min-heap according to the above rules, with the highest priority task always at the top of the heap, resulting in the revisit task list.

[0099] Please refer to Figure 4 , Figure 4 A revisit scheduling sequence diagram is provided, with time on the horizontal axis and voxel risk level on the vertical axis, showing the revisit time points for voxels with different risk levels. High-risk voxels have short revisit intervals, while low-risk voxels have long revisit intervals. The insertion and execution nodes of priority queue tasks are marked.

[0100] Optionally, if the voxel anomaly coefficient is still greater than or equal to a preset second water quality anomaly coefficient threshold after a revisit, a sampling command is triggered to control the sampling device of the carrier execution layer to sample at the voxel location, wherein the first water quality anomaly coefficient threshold is greater than the second water quality anomaly coefficient threshold. For example, if the water quality anomaly coefficient is ≥0.8, it is determined to be a persistent anomaly, and a resampling command is generated to control the carrier to sample again.

[0101] In this embodiment, basic information such as voxel ID, voxel coordinate range, and risk level are acquired; task information such as task ID, path node coordinates, and revisit cycle is acquired; sampling information such as monitoring timestamp, sampling timestamp, sample number, water quality index data, and water quality anomaly coefficient is acquired; and spatiotemporal information such as the carrier's positioning coordinates and attitude data during task execution is acquired. A full-link data pool is constructed based on the above information. A unique identifier (UID) is generated for each data entry, with the binding rule: UID = Hash(Voxel ID + Monitoring Timestamp + Task ID + Sample Number). The completeness of the four-tuple information is checked; if any is missing, it is marked as "incomplete data," stored separately, and an anomaly log is recorded. By establishing a unique association between data and space, time, task, and physical sample, a core index can be provided for subsequent auditing and traceability.

[0102] Extract key operation records for each step in ascending order of timestamp. Each log entry contains basic fields, related fields, and status fields. The basic fields include: log ID, operation timestamp, operation stage, etc.; the related fields include: related task ID, related data UID, and data summary; the status fields include: operation result and failure reason. Sort all logs by operation timestamp to form a full-link log chain. The preceding log ID of each log entry points to the next adjacent operation log, forming a chain relationship, resulting in a full-link log chain file.

[0103] Optionally, an overall SHA-256 check hash is generated for the bound dataset. The dataset is converted into a byte stream using UTF-8 encoding, and the hash value is calculated using the SHA-256 algorithm to generate a check file containing the hash value, calculation timestamp, and dataset UID range. A separate hash value is generated for each piece of data and stored in the dataset for easy verification of individual data.

[0104] Select the export format according to user needs and perform targeted conversion; for example, export the voxel risk heatmap as GeoTIFF format, retaining spatial attributes such as voxel coordinate range, risk value, and risk level; The sampled numerical data is exported as a CSV file, which includes UID, quaternion binding information, water quality indicators, etc. The task data and log chain are exported in JSON format, preserving the hierarchical structure. The standardized export file, SHA-256 checksum file, and log chain file are packaged into a compressed file named: dataexport_timestamp_hashprefix.zip. Export operations are recorded and added to the end of the log chain. Data integrity is ensured through hash verification, and data reusability is guaranteed through standard formatting, ultimately outputting a reliable and complete dataset.

[0105] Please refer to Figure 5 , Figure 5A data binding and export diagram is provided. Data from sensors such as methane sensors, CTDs, and flow meters, as well as ROV / AUV equipment information and timestamps, are input into the processing module. After unified format conversion and association, a binding relationship table of "voxel ID - timestamp - task ID - sample number" is formed. Subsequently, task detail reports, risk assessments, event handling results, and other data are generated. Finally, an output list containing GeoTIFF, CSV, and hash files is exported. A summary and signature are generated through operation records to form an auditable log chain, realizing full-link binding and traceability of data from acquisition to export.

[0106] This application provides an adaptive voxel grid planning device for natural water bodies, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the adaptive voxel grid planning method for natural water bodies in Embodiment 1 above.

[0107] The following is for reference. Figure 6 The diagram illustrates a structural schematic of an adaptive voxel grid planning device suitable for implementing embodiments of this application for natural water bodies. The adaptive voxel grid planning device for natural water bodies in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, personal digital assistants (PDAs), tablet computers (PADs), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The adaptive voxel grid planning device for natural water bodies shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0108] like Figure 6As shown, the adaptive voxel grid planning device for natural water bodies may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The random access memory 1004 also stores various programs and data required for the operation of the adaptive voxel grid planning device for natural water bodies. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the adaptive voxel grid planning device for natural water bodies to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows an adaptive voxel grid planning device for natural water bodies with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0109] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0110] The adaptive voxel grid planning device for natural water bodies provided in this application employs the adaptive voxel grid planning method for natural water bodies described in the above embodiments. This method solves the technical problem of how to construct differentiated voxel grids, thereby improving the efficiency and accuracy of water body inspection. Compared with the prior art, the beneficial effects of the adaptive voxel grid planning device for natural water bodies provided in this application are the same as those of the adaptive voxel grid planning method for natural water bodies provided in the above embodiments. Furthermore, other technical features of this adaptive voxel grid planning device for natural water bodies are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0111] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0113] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the adaptive voxel grid planning method for natural water bodies in the above embodiments.

[0114] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0115] The aforementioned computer-readable storage medium may be included in an adaptive voxel grid planning device for natural water bodies; or it may exist independently and not assembled into an adaptive voxel grid planning device for natural water bodies. The aforementioned computer-readable storage medium carries one or more programs that, when executed by the adaptive voxel grid planning device for natural water bodies, cause the device to: divide the three-dimensional terrain model of the target water body based on a preset initial resolution to obtain an initial voxel grid; input real-time water quality data, flow velocity data, historical anomaly data, and carrier energy consumption parameters of each initial voxel grid into a preset risk field function to obtain the risk level of each initial voxel grid; and adjust the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain the target voxel grid.

[0116] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the client computer, partially on the client computer, as a standalone software package, partially on the client computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the client computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0117] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0118] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0119] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the adaptive voxel grid planning method for natural water bodies described above. This method can solve the technical problem of how to construct differentiated voxel grids, thereby improving the efficiency and accuracy of water body inspection. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the adaptive voxel grid planning method for natural water bodies provided in the above embodiments, and will not be elaborated upon here.

[0120] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. An adaptive voxel grid planning method for natural water bodies, characterized in that, The adaptive voxel grid planning method for the natural water body includes: The three-dimensional terrain model of the target water body is divided based on the preset initial resolution to obtain an initial voxel mesh; The real-time water quality data, flow velocity data, historical anomaly data, and carrier energy consumption parameters of each initial voxel grid are input into a preset risk field function to obtain the risk level of each initial voxel grid. Based on the target resolution corresponding to the risk level, the size of the initial voxel grid is adjusted to obtain the target voxel grid.

2. The adaptive voxel grid planning method for natural water bodies as described in claim 1, characterized in that, The step of inputting real-time water quality data, flow velocity data, historical anomaly data, and carrier energy consumption parameters of each initial voxel grid into a preset risk field function to obtain the risk level of each initial voxel grid includes: Based on the voxel grid identifier of the initial voxel grid, the real-time water quality anomaly coefficient corresponding to the real-time water quality data, the historical anomaly probability corresponding to the historical anomaly data, the flow velocity shear coefficient corresponding to the flow velocity data, and the energy consumption coefficient corresponding to the carrier energy consumption parameter are obtained. The real-time water quality anomaly coefficient, the historical anomaly probability, the flow velocity shear coefficient, and the energy consumption coefficient are input into the risk field function and weighted and summed to obtain the risk value of the initial voxel grid. The risk value is compared with the preset risk level classification rules to obtain the risk level.

3. The adaptive voxel grid planning method for natural water bodies as described in claim 2, characterized in that, The steps of obtaining the real-time water quality anomaly coefficient corresponding to the real-time water quality data, the historical anomaly probability corresponding to the historical anomaly data, the flow velocity shear coefficient corresponding to the flow velocity data, and the energy consumption coefficient corresponding to the carrier energy consumption parameter include: The deviation rate between the real-time water quality data and the preset water quality standard data is determined as the real-time water quality anomaly coefficient; Obtain the historical anomaly count and total monitoring count associated with the initial voxel grid, and determine the historical anomaly probability based on the historical anomaly count and total monitoring count; Obtain the three-dimensional velocity vector associated with the initial voxel mesh, and determine the maximum absolute value of the partial derivative of the three-dimensional velocity vector as the velocity shear coefficient; The distance between the initial voxel grid and the preset voxel center point is obtained, and the product of the distance and the energy consumption per unit distance is determined as the energy consumption coefficient.

4. The adaptive voxel grid planning method for natural water bodies as described in claim 1, characterized in that, The step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain the target voxel grid includes: If the initial voxel is a high-risk voxel, then the initial resolution is reduced by a preset factor to obtain the target resolution; If the initial voxel is a medium-risk voxel, then the initial resolution is determined as the target resolution; If the initial voxel is a low-risk voxel, the initial resolution is increased to a preset multiple to obtain the target resolution; The initial voxel grid is adjusted based on the target resolution to obtain the target voxel grid.

5. The adaptive voxel grid planning method for natural water bodies as described in claim 1, characterized in that, After the step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain the target voxel grid, the adaptive voxel grid planning method for natural water bodies further includes: A global path is obtained by performing path planning based on a preset global planning algorithm and the target voxel mesh; Based on a preset local path planning algorithm and real-time flow velocity gradient, the global path is adjusted to obtain the target path.

6. The adaptive voxel grid planning method for natural water bodies as described in claim 5, characterized in that, The step of adjusting the global path based on a preset local path planning algorithm and real-time flow velocity gradient to obtain the target path includes: The area within the preset range is defined as the local planning range. Based on the RRT algorithm, local path planning is performed on the local planning range to obtain the initial local path. If the velocity gradient corresponding to the path node in the initial local path is less than or equal to a preset velocity gradient threshold, the deviation between the initial local path and the global path is less than or equal to a preset deviation threshold, and the local energy consumption increment is less than or equal to a preset local energy consumption increment threshold, then the initial local path is determined to be the target local path. The global path is adjusted based on the target local path to obtain the target path.

7. The adaptive voxel grid planning method for natural water bodies as described in claim 5, characterized in that, After the step of adjusting the global path based on a preset local path planning algorithm and real-time flow velocity gradient to obtain the target path, the adaptive voxel grid planning method for natural water bodies further includes: When a new high-risk voxel is detected, and / or the current energy consumption is less than or equal to the preset remaining energy threshold, the global path update process is triggered to obtain the updated global path. When an unmarked obstacle is detected, the local path update process is triggered to obtain the updated local path.

8. The adaptive voxel grid planning method for natural water bodies as described in claim 1, characterized in that, After the step of adjusting the size of the initial voxel grid based on the target resolution corresponding to the risk level to obtain the target voxel grid, the adaptive voxel grid planning method for natural water bodies further includes: Based on real-time collected water quality index data and a preset water quality anomaly coefficient quantification formula, the water quality anomaly coefficient of each target voxel grid is determined. If the water quality anomaly coefficient is greater than the preset first water quality anomaly coefficient threshold, an anomaly warning is triggered, and the target voxel grid is marked as a highly anomalous voxel. Based on the risk level of the highly anomalous voxels, the revisit cycle is matched from the revisit cycle mapping table, and the revisit tasks are generated and added to the priority queue to obtain the revisit task list.

9. An adaptive voxel grid planning device for natural water bodies, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the adaptive voxel grid planning method for natural water bodies as claimed in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the adaptive voxel grid planning method for natural water bodies as described in any one of claims 1 to 8.

Images