Amphibious unmanned aerial vehicle flow measurement navigation adaptive planning adjustment system and method

Abstract

This invention discloses an adaptive planning and adjustment system and method for amphibious unmanned aerial vehicle (UAV) flow measurement navigation, relating to the fields of amphibious UAVs and hydrological monitoring technology. The system includes modules for river environment perception, initial flow measurement route planning, real-time drift status detection, obstacle hazard level assessment, dynamic path adjustment execution, and flow measurement path splicing and verification. A calibration module is optional. It acquires and calibrates orthophotos of the entire river channel using a visual camera, identifies obstacles to generate a database for planning initial flow measurement segments and paths, monitors the floating vessel's status and flow parameters in real time, divides the river channel into grids and calculates a grid hazard level index, predicts the floating vessel's drift trajectory, and obtains a high-risk area entry index. When the index reaches a threshold, it triggers emergency takeoff and path replanning, and finally splices the trajectory and verifies data integrity. This invention achieves adaptive adjustment of flow measurement paths in complex dynamic hydrological environments, ensuring operational safety and continuous, accurate data, and improving flow measurement efficiency and automation.

Description

Technical Field

[0001] This invention relates to the field of amphibious unmanned aerial vehicles (UAVs) and hydrological monitoring technology, and more specifically to an adaptive planning and adjustment system and method for amphibious UAV current measurement navigation. Background Technology

[0002] With the development of hydrological monitoring and river flow measurement technology, amphibious unmanned aerial vehicles (UAVs), due to their combined aerial flight and water surface navigation capabilities, are gradually being applied to flow velocity, flow rate, and topographic measurement tasks in complex water areas. However, in actual flow measurement operations, river environments often contain obstacles such as tree branches, floating weeds, or protruding rocks, requiring segmented planning of the flow measurement path. At the same time, due to the fluidity of water flow, the drift trajectory of the UAV after landing on the water surface is difficult to control precisely, causing the preset landing point and takeoff point to deviate, which seriously affects the continuity and accuracy of flow measurement data. Existing technologies still have significant shortcomings in adapting flow measurement paths to dynamic water flow environments, as well as in real-time calibration and closed-loop control of landing and takeoff points.

[0003] A search revealed a hydrological monitoring amphibious drone and method disclosed in patent CN109178305B. This patent proposes an amphibious drone capable of taking off and landing on the water surface and autonomously navigating along a preset route. It integrates a liquid level sensor and a water flow velocity sensor through a float, and utilizes a winch to deploy and retrieve a sonar imager for underwater detection, effectively avoiding frequent takeoffs and landings and improving operational efficiency. However, this solution only supports autonomous navigation on the water surface based on a fixed route, does not consider the impact of river flow on the drone's actual position, and lacks the ability to dynamically correct deviations between the landing and takeoff points caused by water drift. Furthermore, its path planning is statically preset, unable to adjust the flow measurement segment in real time based on visually identified river obstacles (such as floating weeds and protruding rocks), making it difficult to adapt to complex and dynamic river flow measurement scenarios. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to construct an adaptive system that integrates river-wide environmental perception, dynamic calibration of take-off and landing points, risk prediction of drift trajectory, and multi-objective path replanning. This system aims to resolve the technical contradictions in existing technologies, such as the deviation of the floating vessel's drift trajectory from the preset path due to water flow disturbance, which in turn leads to the risk of obstacle collisions and discontinuity in flow measurement data. Through multi-level perception, quantitative evaluation, dynamic decision-making, and path reconstruction, the system ensures safe, continuous, and high-precision flow measurement operations in complex and dynamic hydrological environments.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An adaptive planning and adjustment system for amphibious unmanned aerial vehicle (UAV) navigation based on current measurement includes: The river environment perception module uses a visual camera mounted on a drone to collect orthophotos of the entire river area and perform geometric calibration before the flow measurement task begins. The initial flow measurement route planning module identifies obstacles based on the geometrically calibrated image and generates an obstacle attribute database containing obstacle types, locations, sizes, and distribution ranges. It then divides the continuous flow measurement into segments and plans the landing point, take-off point, and air flight path connecting the two segments for each segment. The real-time drifting status detection module continuously collects information on the floating vessel's current position, nearby obstacles, and water flow parameters after the drone lands. The obstacle hazard level assessment module divides the river into grids based on the obstacle attribute database and real-time environmental data, calculates the hazard level index for each grid cell, predicts the floating vessel's drift trajectory in the future, and calculates the proportion of the spatial intersection area between the trajectory and the high-risk area as the high-risk area entry index. The dynamic path adjustment execution module triggers a preset dynamic path adjustment strategy to execute an emergency takeoff control process when the high-risk area entry index reaches a preset threshold, and replans the flight path and the next stage drift path. The flow path splicing and verification module integrates the actual trajectory data of each flow measurement segment, performs path splicing after unifying the coordinate system, and reviews the overall safety based on the hazard level map. At the same time, it verifies the integrity of the flow measurement data and triggers local return-to-base remeasurement when missing segments are found.

[0007] Furthermore, it also includes a calibration module, which is activated when the UAV reaches a preset altitude or a preset distance from the planned landing point or takeoff point. The module uses a feature extraction algorithm to identify shore markers in the shore image that match a preset shore marker library. Based on the shore markers and visual camera parameters, it calculates the relative distance and azimuth between the markers in the image and the UAV. Then, based on the relative distance and azimuth, it obtains the UAV's real-time geodetic coordinates through spatial geometric conversion. The real-time geodetic coordinates are compared with the planned landing and takeoff points. Based on the comparison results, it selects whether to use a correction strategy for real-time flight correction.

[0008] Furthermore, the obstacle hazard level assessment module includes a grid division strategy, which includes a region segment division step and a grid filling division step. The section division step defines the river channel length direction as longitudinal and the river channel width direction as transverse. According to the preset length unit and the distribution of obstacles, the river channel is divided into several continuous longitudinal segments. The types and distribution densities of obstacles in each segment are marked. From the bank to the center of the river channel, the river channel is divided into near-bank zone, mid-bank zone and mid-channel zone according to the preset distance interval. The risk type of obstacles in each zone is marked. The grid filling and division step involves adapting the grid size according to the actual distribution area of ​​the obstacles, with the principle of being able to completely cover single or continuously distributed obstacles and minimizing the number of grids, to fill the entire river channel with grids. At the same time, the type, location and distribution range information of each obstacle are filled into the corresponding grids to form a river channel grid distribution map containing obstacle information.

[0009] Furthermore, the obstacle hazard level assessment module includes a hazard level calculation strategy, which includes steps for establishing an assessment index system and calculating a hazard level index. The steps for establishing the evaluation index system are as follows: obstacle type, distance between obstacle and floating vessel, water flow velocity, and floating vessel drifting stability are selected as evaluation indicators. The obstacle type is assigned a basic weight according to the degree of risk, the distance is set with attenuation weight according to near and far distance, the water flow velocity is set with influence weight according to the fluctuation amplitude, and the drifting stability is set with safety weight based on attitude parameters. The hazard level index calculation step uses the analytic hierarchy process (AHP) to weight and sum the various assessment indicators to obtain a hazard level index with a preset score range, and then divides the area into high-risk, medium-risk, and low-risk zones according to the index score.

[0010] Furthermore, the obstacle hazard level assessment module includes a floating vessel drift trajectory prediction strategy, which includes a trajectory prediction step and a high-risk area entry index calculation step. The trajectory prediction step uses a Kalman filter algorithm to predict the drift trajectory coordinate sequence of the floating vessel within a preset time period based on the floating vessel's current real-time coordinates and real-time flow direction data. The high-risk area entry index calculation step involves extracting the overlapping area between the predicted trajectory and the high-risk area, and calculating the proportion of the overlapping area to the total coverage area of ​​the predicted trajectory. This proportion represents the pontoon entry index to the high-risk area.

[0011] Furthermore, the dynamic path adjustment strategy includes an emergency takeoff control step, a flight path replanning step, and a temporary landing point planning step. The emergency takeoff control procedure involves adjusting the drone's attitude according to its real-time attitude to maintain stability when the high-risk area entry index reaches a preset threshold, and then taking off to a safe altitude at a preset angle. The flight path replanning step, based on real-time river environment data collected after the UAV takes off and hazard level data generated by the obstacle hazard level assessment module, plans the air flight path from the current emergency takeoff position to the next temporary landing point through a path planning algorithm that introduces hazard level weights.

[0012] The temporary landing point planning step involves selecting temporary landing points that meet the flow measurement conditions based on real-time water flow data, obstacle distribution, and hazard level distribution.

[0013] Furthermore, the path planning algorithm for the hazard level weighting aims to generate a path with the optimal path length and the lowest weighted sum of hazard level indices, and marks obstacle avoidance control points on the path.

[0014] Furthermore, the correction strategy includes a deviation threshold comparison step and a coordinate dynamic compensation step; The deviation threshold comparison step calculates the lateral and longitudinal deviation values ​​between the real-time geodetic coordinates and the planned landing or takeoff point coordinates, and compares the two deviation values ​​with the preset precise takeoff and landing threshold to determine whether the deviation is within the allowable range. The coordinate dynamic compensation step involves using the original planned coordinates when the deviation is within the allowable range, and adjusting the original planned coordinates according to the deviation value when the deviation exceeds the allowable range, based on the real-time geodetic coordinates and combined with the obstacle attribute database.

[0015] Furthermore, the flow path splicing and verification module includes a verification strategy, which includes a path splicing step and a data integrity verification step. The path splicing step involves connecting the actual take-off and landing points of each segment according to the longitudinal coordinate sequence, and supplementing the connection trajectory using corresponding interpolation or curve generation method based on the distance between the two points. After splicing, the path hazard level index is checked and found to be low-risk. The data integrity verification step involves standard verification based on a set of valid flow measurement data corresponding to a preset path length. Missing segments trigger local return-to-home retesting, and the retesting path is based on the overall spliced ​​path planning.

[0016] An adaptive planning and adjustment method for amphibious unmanned aerial vehicle (UAV) navigation based on current measurement includes the following steps: The river environment perception step involves acquiring orthophotos of the entire river area and performing geometric calibration using a visual camera mounted on a drone before the flow measurement task begins. The initial flow measurement route planning step involves identifying obstacles based on the geometrically calibrated image and generating an obstacle attribute database containing obstacle types, locations, sizes, and distribution ranges. This database is then used to divide the continuous flow measurement into segments and plan the landing point, takeoff point, and air flight path connecting each segment. The real-time drifting status detection process continuously collects information on the floating vessel's current position, nearby obstacles, and water flow parameters after the drone lands. The obstacle hazard level assessment steps involve dividing the river channel into grids based on the obstacle attribute database and real-time environmental data, calculating the hazard level index for each grid cell, predicting the floating vessel's drift trajectory in the future, and calculating the proportion of the spatial intersection area between this trajectory and the high-risk area as the high-risk area entry index. The dynamic path adjustment execution steps involve triggering a preset dynamic path adjustment strategy to execute an emergency takeoff control process when the high-risk area entry index reaches a preset threshold, and replanning the flight path and the next stage drift path. The flow measurement path splicing and verification steps integrate the actual trajectory data of each flow measurement segment, unify the coordinate system, and then splice the path. The overall safety is reviewed based on the hazard level map. At the same time, the integrity of the flow measurement data is verified, and a local return-to-base measurement is triggered when a missing segment is found.

[0017] The beneficial effects of this invention are as follows: 1. By constructing an adaptive system that integrates river-wide environmental perception, dynamic calibration of take-off and landing points, drift trajectory risk prediction, and multi-objective path replanning, the core pain points of existing technologies, such as static paths being unable to cope with dynamic water flow disturbances, take-off and landing points being prone to deviations due to drift, high risk of obstacle collisions, and easy discontinuity in flow measurement data, are successfully solved. Furthermore, by dynamically correcting take-off and landing point deviations through the SIFT+EPnP algorithm, and combining multi-dimensional indicators to quantitatively assess the risk level, along with Kalman filter trajectory prediction and risk level weighted path planning technology, the invention can not only predict the drift risk of floating vessels in advance and quickly trigger emergency adjustments, but also effectively avoid obstacle collision accidents, ensure equipment and operational safety, and adapt to various complex dynamic hydrological scenarios. 2. By using a partial return-to-the-shore measurement and trajectory splicing verification mechanism, the continuity and integrity of flow measurement data are ensured, significantly improving the measurement accuracy of hydrological data such as flow velocity and flow rate. At the same time, the system realizes fully automated operation from initial path planning, dynamic calibration, risk assessment to path replanning and trajectory verification, reducing manual intervention and ineffective operations, reducing the labor intensity of operators, and improving the overall flow measurement efficiency. Attached Figure Description

[0018] Figure 1 This is the overall flowchart of flow path planning in this invention; Figure 2 This is a schematic diagram of the flow measurement path planning in this invention; Figure 3 This is a flowchart of the dynamic calibration and correction process for take-off and landing points in this invention. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0020] This invention provides an adaptive planning and adjustment system and method for amphibious unmanned aerial vehicle (UAV) flow measurement navigation, which realizes a full-cycle control process covering pre-mission environmental modeling, initial path generation, dynamic calibration of take-off and landing points, risk assessment of drifting process, emergency take-off and path replanning, to post-mission trajectory stitching and data integrity verification. It effectively solves problems such as poor adaptability of flow measurement paths, large deviation of take-off and landing points, and insufficient data continuity in dynamic water flow environments, and significantly improves the safety and accuracy of flow measurement operations.

[0021] The hardware consists of a drone and a floating platform. The drone is permanently mounted on the platform, which is equipped with various sensors for monitoring water flow parameters (including flow velocity, flow direction, and water depth sensors). The drone carries a visual camera with optical image stabilization to suppress image blurring caused by airflow disturbances during flight. Figure 1 As shown, after the mission is launched, the UAV flies autonomously above the target water area, performing a full-coverage aerial photography operation on the target flow measurement channel at a constant height of 3-5 meters. The river environment perception module can split the video captured by the visual camera frame by frame. For each original image after splitting, a radiometric correction instrument is used to eliminate the influence of high intensity light, and geometric distortion correction is performed to compensate for lens distortion. SIFT feature matching and RANSAC are used to remove mismatched point pairs. Finally, a unified coordinate system orthophoto map of the entire river area is generated through multi-view stereo matching and bundle adjustment. The spatial reference system of this orthophoto map adopts the geodetic coordinate system. Through image preprocessing and modeling operations, the actual situation of river topography and obstacle distribution can be accurately restored, providing high-precision basic data support for subsequent path planning, avoiding misjudgment or omission of obstacles due to image distortion, and ensuring the reliability of the initial planned path.

[0022] Based on orthophotos, the initial flow route planning module performs obstacle identification and flow segmentation planning. Obstacle identification employs an improved YOLOv8 target detection model. During the training phase, this model uses a labeled dataset containing typical river obstacles such as floating weeds, protruding rocks, and shallows for transfer learning. The backbone network uses a CSPDarknet53 structure, and the detection head incorporates deformable convolutional layers to enhance its ability to perceive irregular edge contours. The model input size is fixed at 640×640 pixels, and the output includes the bounding box coordinates, class confidence score, and pixel-level segmentation mask for each detected target. Compared to traditional detection models, this improved model can more accurately identify obstacles of various shapes in the river channel and reduce the false negative rate of small-sized and low-contrast obstacles. To map the obstacle locations in the image to real geographic coordinates, the system establishes image pixel coordinates. with the three-dimensional geographic coordinates of the river channel Affine transformation relationship between them: ,in A 3×3 homography matrix is ​​obtained by solving a direct linear transformation algorithm using at least four pairs of evenly distributed control points of the same name. This transformation relationship enables accurate mapping between pixel coordinates and geographic coordinates, ensuring the authenticity and usability of obstacle location information. The physical dimensions of obstacles are determined as follows: the target area is divided into obstacle regions, and then the outline points of the obstacles are marked using an edge contour algorithm. The major axis L, minor axis W, and height H are then inverted using an image segmentation mask. Finally, all obstacle information is structured and stored in an obstacle attribute database, with each record containing the obstacle type and center coordinates. , The distribution range of the polygon vertices is arranged in clockwise order, with a number of vertices greater than or equal to 4.

[0023] After the obstacle database is built, such as Figure 2 As shown, the system measures the flow along the longitudinal direction of the river at a preset single-segment measurement length. The continuous flow measurement is divided into segments based on a baseline. Segmented planning can avoid the accumulation of path deviations caused by excessively long single flow measurement distances, and at the same time, it facilitates flexible responses to changes in obstacle distribution in local areas. Each segment needs to determine a pair of landing points. and takeoff point Both must simultaneously meet the following constraints: First, they must be located within the same longitudinal segment, and the longitudinal spacing must meet the continuity requirements for flow measurement data acquisition, i.e., the longitudinal spacing... Second, the horizontal coordinate. ,in and These are the coordinates of the left and right bank boundaries of the river channel, ensuring that the lateral distance from obstacles on the bank is less than [the specified value]. Third, priority should be given to the mid-channel zone, where the water flow conditions are relatively stable, which can improve the stability of the floating vessel during flow measurement and ensure the accuracy of the flow measurement data. ,in The coordinates of the river centerline were determined by cluster analysis of water flow texture in the orthophoto image. The flight path was fitted with a cubic B-spline curve. The control point sequence consisted of the takeoff point, several obstacle avoidance intermediate points, and the landing point. The curve equation was: ,in, For the i-th control point, Using 4th-order B-spline basis functions, during path planning, all obstacles are projected outward by 5 meters on the horizontal plane to form an obstacle expansion zone, which is included as a no-fly zone in the path search constraint to ensure that the flight path is in a safe airspace throughout.

[0024] During the approaching takeoff and landing phase, such as Figure 3 As shown, the calibration module is automatically activated to correct planning coordinate deviations. This module effectively compensates for takeoff and landing point offsets caused by water flow drift, ensuring precise takeoff and landing. When the horizontal distance between the UAV and the planned landing point is ≤10 meters or the horizontal distance between the UAV and the planned takeoff point is ≤50 meters, the system calls the pre-set physical marker point library on both sides of the river. Each marker point has known geodetic coordinates. Unique visual feature encoding (such as high-contrast concentric circle patterns) facilitates rapid and accurate identification, reducing recognition errors caused by environmental interference. The RGB camera on the drone captures images of the shoreline scene in real time, extracts key points from the image using the SIFT algorithm, and generates a 128-dimensional descriptor vector. This vector is then matched with template descriptors in the marker point library using Euclidean distance. The matching criterion is: the minimum distance is less than a threshold. =0.6 and the second smallest distance ratio is greater than 0.8. When the number of valid matching points is not less than 3, the system calls the PnP pose calculation module and uses the EPnP algorithm to solve the camera pose. The coordinates of m marker points in the world coordinate system are known. Corresponding image pixel coordinates By solving: ,in, For the pre-calibrated camera intrinsic parameter matrix, Using the scale factor, we obtain the rotation matrix. With translation vector From this, the current geodetic coordinates of the drone can be deduced. System calculation With the planned take-off and landing points Deviation: Lateral deviation Longitudinal deviation ,like Miqie If the distance is within meters, the original planned coordinates will be maintained; otherwise, under the constraints of the obstacle attribute database, the coordinates will be adjusted accordingly. Based on the reference point, fine-tune the coordinates along the water flow direction to generate the corrected take-off and landing points, ensuring their distance from obstacles. Safety conditions for meters.

[0025] After the drone lands, it enters the floating survey phase. During this time, the obstacle hazard assessment module continues to operate. This module can assess the river environment risk in real time, providing a basis for dynamic path adjustment and preventing the floating vessel from entering high-risk areas and causing safety accidents. This module first performs grid division on the entire river area, including two steps: segment division and grid filling. Grid division can break down the complex river environment into several simple grid units, which facilitates risk quantification assessment. In the segment division, the river is divided into sections with the river length direction as the X-axis (longitudinal) and the width direction as the Y-axis (lateral), with a certain distance as the unit. Each longitudinal segment The frequency and density of obstacle types within each segment were statistically analyzed. ,in, For the number of obstacles, For each segment, density statistics can intuitively reflect the density of obstacle distribution. Simultaneously, from the left bank to the right bank, the area is divided into three transverse zones according to the Y-coordinate: the nearshore zone... Mid-coast Zhonghong Belt The dominant obstacle types within each zone are recorded. Differentiating risk types across different zones improves the targeting of risk assessments. The mesh filling uses an adaptive quadtree mesh, recursively subdividing the mesh containing obstacles until a single obstacle is completely enclosed within a leaf node mesh, minimizing the overall mesh count. This adaptive subdivision method accurately covers obstacles while reducing the mesh count and lowering the system's computational load. Each mesh unit... It stores the vertical segment, horizontal band, list of obstacles it contains, and their attributes. The structured grid data facilitates quick querying and risk calculation.

[0026] Based on this grid, the hazard level index calculation unit performs... The calculation and evaluation index system includes four dimensions: obstacle type. Relative distance Water flow velocity and floating vessel attitude stability A multi-dimensional indicator system can comprehensively reflect the risk status of grid units, avoiding the one-sidedness caused by a single indicator assessment. Obstacle types Assign basic weights Floating grass pile =0.3, protruding rocks =0.8, shallow water =0.6, bridge pier =0.9; Weights are assigned based on the varying degrees of collision risk posed by different obstacles to the floating vessel, with higher risks resulting in greater weights, and relative distances... Using an exponential decay function , The Euclidean distance from the center of the floating vessel to the nearest point of the obstacle; the influence weight of water flow velocity. ,in The standard deviation of flow rate over the past 10 seconds. Average flow velocity; attitude stability weights , The pitch angle, The roll angle. The weight vectors of each indicator are determined using the analytic hierarchy process (AHP). Then the grid Danger Level Index for: ,in This represents the normalization of the i-th indicator, with the normalization interval determined based on historical river data statistics. Classified by threshold: This is a high-risk area. It is a medium-risk area. It is a low-risk area.

[0027] Meanwhile, the Kalman filter trajectory prediction module performs predictions based on the current state of the floating vessel. This module can predict the floating vessel's drift trajectory in advance, allowing sufficient time for risk warning. The system dynamic model is as follows: ,in, , For zero-mean Gaussian process noise, the covariance matrix is... The observation vector contains the key data acquired by the sensor, i.e., the observation vector. Observation matrix Observation noise covariance The system is based on Using seconds as the step size, iteratively predict the future. Trajectory point sequence within seconds Then, the intersection area of ​​the convex hull polygon covered by the predicted trajectory and all high-risk areas is calculated. Total coverage area Then high-risk areas enter the index ,when At that time, high-risk areas entered the index Once the dynamic path adjustment execution module is triggered, risk response measures should be initiated promptly.

[0028] The dynamic path adjustment execution module is in The emergency takeoff control procedure is initiated, and the flight control system reads IMU data. or Then the attitude stabilization control unit executes the stabilization program: adjusting the speed of the four rotors through the PID controller, so that... and The drone converges to within ±1° within 2 seconds. Then, it climbs vertically. After takeoff, the wide-angle camera initiates a new round of environmental scanning to update the local obstacle distribution. The newly acquired local obstacle image data is aligned with the coordinate system of the original obstacle attribute database to perform incremental updates of obstacle information. For newly added obstacles, an improved YOLOv8 target detection model, consistent with the initial obstacle identification, is used to complete type determination, bounding box extraction, and geographic coordinate mapping, supplementing the corresponding grid cells in the database. For obstacles that have been recorded in the original database but whose positions have shifted (such as floating obstacles), the target correlation is confirmed through feature matching algorithms, and their center coordinates and distribution range polygon vertex sequences are updated. For obstacles that have disappeared, their status is marked as "invalid" and historical records are retained to avoid subsequent evaluation biases caused by accidental data deletion.

[0029] Based on the updated obstacle distribution data, real-time water flow parameters, and hazard level distribution, a second verification is performed on the initially selected temporary landing points. The verification includes: confirming that there are no new high-risk obstacles within the preset range around the temporary landing point, and that the fluctuation range of water flow velocity in the area meets the requirements for stable floating of the floating vessel; and ensuring the connectivity between the temporary landing point and the preset flow measurement section in the next stage, in conjunction with the subsequent flow measurement segment planning requirements, to avoid data gaps caused by improper landing point selection. Temporary landing points that pass the verification are marked as "available," while those that fail are restarted in the screening process until a landing point that meets all constraints is determined.

[0030] When the UAV flies to a preset distance from the temporary landing point, the calibration module is activated. Using the same SIFT feature matching and EPnP pose calculation process as the initial take-off and landing point calibration, the temporary landing point is accurately calibrated. The lateral and longitudinal deviations between the UAV's real-time geodetic coordinates and the planned coordinates of the temporary landing point are calculated. If the deviation exceeds the allowable range, the flight trajectory is adjusted through a coordinate dynamic compensation step to ensure that the deviation converges to within the accurate take-off and landing threshold. During the landing process, the relative positional relationship between the floating vessel and the temporary landing point is continuously monitored. The UAV's landing attitude is adjusted in combination with the real-time water flow direction to suppress the landing deviation caused by water flow impact and achieve a smooth landing.

[0031] After the drone successfully lands, the system automatically archives the dynamically adjusted flight path data (including emergency takeoff position, coordinates of the replanned path control point, and actual coordinates of the temporary landing point), the updated obstacle attribute data, and the hazard level assessment results to the global database. At the same time, it links the trajectory information of the current flow measurement segment with that of the previous flow measurement segment, providing complete segmented data support for the subsequent flow measurement path splicing and verification module, and ensuring the traceability of the trajectory and data throughout the entire process.

[0032] After the mission is completed, the flow path splicing and verification module integrates all segmented trajectory data. First, it splices the flight trajectories of each UAV segment. drifting trajectory of the floating vessel (Drone drift trajectory) is uniformly converted to the initially defined river coordinate system to eliminate coordinate drift introduced by multiple calibrations. The path is spliced ​​and the segments are arranged in ascending order of X coordinate, with the previous segment's actual takeoff point as the reference. and the actual landing point of the latter part To connect endpoints, if the distance between two points is less than 5 meters, linear interpolation is used to generate the centerline point. If the distance between two points is greater than 5 meters, a cubic B-spline curve is constructed in the low-risk area using the two points as endpoints. Control points are determined by searching for the shortest connection path in the safety grid using the Djkstra algorithm. After splicing, the path splicing verification unit performs a hazard check on the entire path: if any path segment crosses the low-risk area, a detour trajectory is regenerated within 5 meters on both sides of that segment.

[0033] Finally, the system automatically integrates all core data to form a complete task archive package. The archive includes: a full-area orthophoto map of the river channel, an obstacle attribute database (including update records), coordinates of take-off and landing points for each segment (including calibration correction records), complete flow measurement trajectory data (including flight and drift paths), hazard level assessment results, dynamic path adjustment logs, original and calibrated flow measurement data, and descriptions of supplementary measurement areas. Based on the archived data, a standardized hydrological flow measurement report is generated. The report covers core content such as basic information of the flow measurement task (time, location, river overview), system operation status analysis, obstacle distribution and risk assessment summary, flow measurement data statistics (velocity, flow rate, cross-sectional average velocity, etc.), and data accuracy assessment. The report supports data export and visualization.

[0034] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An adaptive planning and adjustment system for amphibious unmanned aerial vehicle (UAV) navigation based on current measurement, characterized in that: include: The river environment perception module uses a visual camera mounted on a drone to collect orthophotos of the entire river area and perform geometric calibration before the flow measurement task begins. The initial flow measurement route planning module identifies obstacles based on the geometrically calibrated image and generates an obstacle attribute database containing obstacle types, locations, sizes, and distribution ranges. It then divides the continuous flow measurement into segments and plans the landing point, take-off point, and air flight path connecting the two segments for each segment. The real-time drifting status detection module continuously collects information on the floating vessel's current position, nearby obstacles, and water flow parameters after the drone lands. The obstacle hazard level assessment module divides the river into grids based on the obstacle attribute database and real-time environmental data, calculates the hazard level index for each grid cell, predicts the floating vessel's drift trajectory in the future, and calculates the proportion of the spatial intersection area between the trajectory and the high-risk area as the high-risk area entry index. The dynamic path adjustment execution module triggers a preset dynamic path adjustment strategy to execute an emergency takeoff control process when the high-risk area entry index reaches a preset threshold, and replans the flight path and the next stage drift path. The flow path splicing and verification module integrates the actual trajectory data of each flow measurement segment, performs path splicing after unifying the coordinate system, and reviews the overall safety based on the hazard level map. At the same time, it verifies the integrity of the flow measurement data and triggers local return-to-base remeasurement when missing segments are found.

2. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 1, characterized in that: It also includes a calibration module, which is activated when the UAV reaches a preset altitude or a preset distance from the planned landing point or takeoff point. The module uses a feature extraction algorithm to identify shore markers in the shore image that match a preset shore marker library. It calculates the relative distance and azimuth between the markers in the image and the UAV based on the shore markers and visual camera parameters. Then, it obtains the UAV's real-time geodetic coordinates through spatial geometric conversion based on the relative distance and azimuth. The module compares the real-time geodetic coordinates with the planned landing and takeoff points to determine whether to use a correction strategy for real-time flight correction based on the comparison results.

3. The amphibious unmanned aerial vehicle (UAV) current measurement navigation adaptive planning and adjustment system according to claim 1 or 2, characterized in that: The obstacle hazard level assessment module includes a grid division strategy, which includes a region segment division step and a grid filling division step. The section division step defines the river channel length direction as longitudinal and the river channel width direction as transverse. According to the preset length unit and the distribution of obstacles, the river channel is divided into several continuous longitudinal segments. The types and distribution densities of obstacles in each segment are marked. From the bank to the center of the river channel, the river channel is divided into near-bank zone, mid-bank zone and mid-channel zone according to the preset distance interval. The risk type of obstacles in each zone is marked. The grid filling and division step involves adapting the grid size according to the actual distribution area of ​​the obstacles, with the principle of being able to completely cover single or continuously distributed obstacles and minimizing the number of grids, to fill the entire river channel with grids. At the same time, the type, location and distribution range information of each obstacle are filled into the corresponding grids to form a river channel grid distribution map containing obstacle information.

4. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 3, characterized in that: The obstacle hazard level assessment module includes a hazard level calculation strategy, which includes steps for establishing an assessment index system and calculating a hazard level index. The steps for establishing the evaluation index system are as follows: obstacle type, distance between obstacle and floating vessel, water flow velocity, and floating vessel drifting stability are selected as evaluation indicators. The obstacle type is assigned a basic weight according to the degree of risk, the distance is set with attenuation weight according to near and far distance, the water flow velocity is set with influence weight according to the fluctuation amplitude, and the drifting stability is set with safety weight based on attitude parameters. The hazard level index calculation step uses the analytic hierarchy process (AHP) to weight and sum the various assessment indicators to obtain a hazard level index with a preset score range, and then divides the area into high-risk, medium-risk, and low-risk zones according to the index score.

5. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 4, characterized in that: The obstacle hazard level assessment module includes a floating vessel drift trajectory prediction strategy, which includes a trajectory prediction step and a high-risk area entry index calculation step. The trajectory prediction step uses a Kalman filter algorithm to predict the drift trajectory coordinate sequence of the floating vessel within a preset time period based on the floating vessel's current real-time coordinates and real-time flow direction data. The high-risk area entry index calculation step involves extracting the overlapping area between the predicted trajectory and the high-risk area, and calculating the proportion of the overlapping area to the total coverage area of ​​the predicted trajectory. This proportion represents the pontoon entry index to the high-risk area.

6. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 5, characterized in that: The dynamic path adjustment strategy includes emergency takeoff control steps, flight path replanning steps, and temporary landing point planning steps. The emergency takeoff control procedure involves adjusting the drone's attitude according to its real-time attitude to maintain stability when the high-risk area entry index reaches a preset threshold, and then taking off to a safe altitude at a preset angle. The flight path replanning step, based on real-time river environment data collected after the UAV takes off and hazard level data generated by the obstacle hazard level assessment module, plans the air flight path from the current emergency take-off position to the next temporary landing point through a path planning algorithm that introduces hazard level weights. The temporary landing point planning step involves selecting temporary landing points that meet the flow measurement conditions based on real-time water flow data, obstacle distribution, and hazard level distribution.

7. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 6, characterized in that: The path planning algorithm for the hazard level weighting aims to generate a path with the optimal path length and the lowest weighted sum of hazard level indices, and marks obstacle avoidance control points on the path.

8. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 2, characterized in that: The correction strategy includes a deviation threshold comparison step and a coordinate dynamic compensation step; The deviation threshold comparison step calculates the lateral and longitudinal deviation values ​​between the real-time geodetic coordinates and the planned landing or takeoff point coordinates, and compares the two deviation values ​​with the preset precise takeoff and landing threshold to determine whether the deviation is within the allowable range. The coordinate dynamic compensation step involves using the original planned coordinates when the deviation is within the allowable range, and adjusting the original planned coordinates according to the deviation value when the deviation exceeds the allowable range, based on the real-time geodetic coordinates and combined with the obstacle attribute database.

9. The amphibious unmanned aerial vehicle current measurement navigation adaptive planning and adjustment system according to claim 3, characterized in that: The flow path splicing and verification module includes a verification strategy, which includes a path splicing step and a data integrity verification step. The path splicing step involves connecting the actual take-off and landing points of each segment according to the longitudinal coordinate sequence, and supplementing the connection trajectory using corresponding interpolation or curve generation method based on the distance between the two points. After splicing, the path hazard level index is checked and found to be low-risk. The data integrity verification step involves standard verification based on a set of valid flow measurement data corresponding to a preset path length. Missing segments trigger local return-to-home retesting, and the retesting path is based on the overall spliced ​​path planning.

10. An adaptive planning and adjustment method for current measurement navigation of an amphibious unmanned aerial vehicle, characterized in that: Includes the following steps: The river environment perception step involves acquiring orthophotos of the entire river area and performing geometric calibration using a visual camera mounted on a drone before the flow measurement task begins. The initial flow measurement route planning step involves identifying obstacles based on the geometrically calibrated image and generating an obstacle attribute database containing obstacle types, locations, sizes, and distribution ranges. This database is then used to divide the continuous flow measurement into segments and plan the landing point, takeoff point, and air flight path connecting each segment. The real-time drifting status detection process continuously collects information on the floating vessel's current position, nearby obstacles, and water flow parameters after the drone lands. The obstacle hazard level assessment steps involve dividing the river channel into grids based on the obstacle attribute database and real-time environmental data, calculating the hazard level index for each grid cell, predicting the floating vessel's drift trajectory in the future, and calculating the proportion of the spatial intersection area between this trajectory and the high-risk area as the high-risk area entry index. The dynamic path adjustment execution steps involve triggering a preset dynamic path adjustment strategy to execute an emergency takeoff control process when the high-risk area entry index reaches a preset threshold, and replanning the flight path and the next stage drift path. The flow measurement path splicing and verification steps integrate the actual trajectory data of each flow measurement segment, unify the coordinate system, perform path splicing, and review the overall safety based on the hazard level map. At the same time, the integrity of the flow measurement data is verified, and a local return-to-base supplementary measurement is triggered when a missing segment is found.

Images