Water area flow measurement system and method based on amphibious unmanned aerial vehicle

Abstract

The invention discloses a water area flow measurement system and method based on an amphibious unmanned aerial vehicle, and aims to solve the problem that water surface waves caused by rotor wing airflow disturb flow measurement data when the unmanned aerial vehicle lands. The system comprises a wave sensing and triggering module, a directional influence area extraction module, an influence distance prediction module, an interference section cutting module and a compensation path planning module, a dynamic verification and correction module can be additionally arranged, and wave multi-dimensional characteristic parameters are extracted by triggering and collecting water surface images, intercepting a fan-shaped observation area and calculating an initial influence radius. Predicting an initial wave influence distance in combination with a scenarized attenuation coefficient and a hovering duration correction coefficient, cutting an interference data segment after dynamic verification and correction, and then screening a low-risk stable region to generate an equivalent compensation path; the method accurately eliminates invalid data, guarantees the accuracy and integrity of flow measurement data, avoids blind supplementary measurement, improves the operation efficiency, and is adaptive to a complex water area environment.

Description

Technical Field

[0001] This invention relates to the field of amphibious unmanned aerial vehicles (UAVs) and hydrological monitoring technology, and more specifically to a water flow measurement system and method based on amphibious UAVs. Background Technology

[0002] With the continuous development of hydrological monitoring technology, amphibious drones, with their dual characteristics of aerial flight and water surface drifting, have become the core equipment for measuring flow velocity and flow rate in complex river environments. They have effectively solved the problems of traditional flow measurement methods being unable to cover complex water areas and having low operational efficiency. In order to cope with problems such as path deviation, obstacle collision and data discontinuity in dynamic water flow environments, related technical fields have proposed a variety of adaptive planning and adjustment schemes. Through the collaborative work of multiple modules, dynamic adjustment of flow measurement paths and data integrity verification have been realized, improving the safety and automation level of flow measurement operations.

[0003] However, in actual flow measurement scenarios, it was found that the flow measurement data did not take into account the disturbance of the water surface state caused by the rotor airflow during the drone's landing. When an amphibious drone lands on the water, its high-speed rotor rotation generates strong airflow, which directly affects the water surface around the landing point, causing violent and irregular waves. Flow measurement data collection begins immediately after the drone lands on the water surface. At this time, the waves caused by the strong airflow have not yet decayed, resulting in severe interference to the flow measurement sensors carried by the floating vessel during data collection. On the one hand, the undulation of the water surface waves causes the sensor sampling attitude to be unstable, resulting in instantaneous deviations in flow velocity measurement. On the other hand, the water surface disturbance caused by the waves is superimposed on the natural water flow movement, making it difficult for the sensor to distinguish between the real water flow signal and the wave interference signal, thus leading to data distortion in the initial flow measurement stage. If the flow measurement results containing interference data are directly included in the final analysis, it will lead to systematic deviations in the overall flow velocity and flow rate calculation results. If the entire section is blindly supplemented with measurements, it will result in duplicated work and reduced flow measurement efficiency. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this invention is to solve the problem that the interference of water surface waves caused by rotor airflow disturbance on the initial flow measurement data is not considered in the existing amphibious UAV flow measurement operations. It solves the flow velocity measurement deviation caused by sensor attitude instability due to waves and the superposition of water flow signal and disturbance signal, and avoids the decrease in operation efficiency and waste of resources due to blind supplementary measurement.

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

[0006] A water flow measurement system based on an amphibious unmanned aerial vehicle (UAV) includes: The wave perception trigger module triggers the vision camera to capture images of the water surface facing the direction of the water flow after the drone rotor has completely stopped rotating. The directional influence area extraction module extracts a fan-shaped observation area in the direction of the drone's drift from the water surface image and calculates the actual distance from the outermost boundary of the wave to the center point of the drone's projection as the initial influence radius. The impact distance prediction module extracts multi-dimensional feature parameters representing the spatial periodicity, undulation amplitude, and texture energy of waves in the fan-shaped observation area. Based on the current water flow and wind speed scenario, it obtains the average value of the attenuation coefficient from the preset wave feature benchmark library. Then, it calculates the correction coefficient based on the actual hovering time of the UAV before landing and corrects the average value of the attenuation coefficient to obtain the wave attenuation coefficient. Finally, it predicts the initial wave impact distance through the multi-dimensional feature parameters, the wave attenuation coefficient, and the current water flow velocity. The interference section trimming module extracts and deletes the initial wave-affected distance segments from the flow measurement data. The compensation path planning module selects low-risk and stable adjacent areas based on the preset river hazard level map, and generates a new flow measurement sub-path that is consistent with the original effective path direction, matches the length, and avoids high disturbances, using the initial wave influence distance as the compensation length.

[0007] Furthermore, the multidimensional feature parameters include wave wavelength, wave height, and wave energy value. The wave wavelength reflects the spatial periodicity of the wave, the wave height characterizes the physical amplitude of the wave, and the wave energy value reflects the intensity and uniformity of the wave texture.

[0008] Furthermore, the influence distance prediction module includes a multi-dimensional feature extraction strategy. The multi-dimensional feature extraction strategy includes performing a one-dimensional Fourier transform on the gray-level signal of the fan-shaped observation area using the Fourier transform method, taking the period corresponding to the peak of the frequency spectrum as the wave wavelength, locating the pixel position coordinates of the wave peak and trough through gray-level gradient detection, and calculating the actual wave height by combining the received visual camera intrinsic parameter matrix, constructing a gray-level co-occurrence matrix of the fan-shaped observation area, and calculating the matrix energy value as the wave energy value based on the gray-level co-occurrence matrix.

[0009] Furthermore, the influence distance prediction module includes a database construction submodule, which is used to construct a wave feature benchmark library containing wave attenuation coefficient samples corresponding to different water flow types and wind speed scenarios, and is also used to construct a mapping relationship model between rotor airflow intensity, hovering time and initial wave features.

[0010] Furthermore, the influence distance prediction module also includes a retrieval strategy, which includes: Based on the water flow type and wind speed parameters of the current flow measurement scenario, attenuation coefficient samples of the same scenario are matched in the benchmark library, and the benchmark mean of attenuation coefficient is obtained through statistical analysis. Based on the actual hovering time of the drone before landing, the hovering time correction coefficient is calculated using the mapping model. The average value of the attenuation coefficient and the hovering time correction coefficient are then weighted and fused according to a preset weight ratio to obtain the wave attenuation coefficient in the current scenario.

[0011] Furthermore, the influence distance prediction module also includes a prediction distance calculation strategy, which includes: Based on the current wave height, wave attenuation coefficient, and initial influence radius, the wave attenuation degree at the influence radius is quantified using a preset attenuation model to obtain the wave attenuation value. The initial wave influence distance is calculated by integrating the initial influence radius, multidimensional characteristic parameters, wave attenuation value, and current water flow velocity.

[0012] Furthermore, it also includes a dynamic verification and correction module, which sets the monitoring starting point with the initial wave impact distance. When the real-time position of the UAV coincides with the monitoring starting point, a data acquisition command is triggered, and water surface images are acquired at a certain acquisition frequency. Then, the wave change rate is analyzed by the water surface images acquired at adjacent times to determine whether the wave disturbance tends to stabilize, thereby dynamically correcting the predicted initial wave impact distance to obtain the actual interference distance.

[0013] Furthermore, the dynamic verification correction module includes a determination strategy, which includes: Multidimensional feature parameters of waves in the sector observation area at adjacent time points are extracted, and the change amplitude of each parameter in the multidimensional feature parameters is calculated. When the variation amplitude of each parameter in multiple consecutive adjacent time periods is less than the preset amplitude, or when the comprehensive wave variation rate calculated by combining the variation amplitudes of all parameters is less than the preset rate, it is determined that the wave interference of the rotor airflow on the water surface has been eliminated, and the corresponding target time and the current position of the UAV are recorded. The actual interference distance of the waves is determined based on the landing position and the current position of the UAV.

[0014] Furthermore, the interference segment trimming module includes a timestamp alignment unit and a data truncation unit. The timestamp alignment unit synchronizes the visual trigger time with the sampling sequence of the flow sensor. The data truncation unit removes the flow data segments affected by the disturbance based on the drift time window corresponding to the corrected actual interference distance.

[0015] A method for water flow measurement based on amphibious unmanned aerial vehicles includes the following steps: After the drone rotors have completely stopped rotating, the vision camera is triggered to capture an image of the water surface facing the direction of the water flow. In the water surface image, a fan-shaped observation area in the direction of the UAV's drift is extracted, and the actual distance from the outermost boundary of the wave to the center point of the fuselage projection is calculated as the initial influence radius. Multidimensional feature parameters characterizing wave spatial periodicity, undulation amplitude, and texture energy are extracted from the fan-shaped observation area. Based on the current water flow and wind speed scenario, the average value of the attenuation coefficient is obtained from the preset wave feature benchmark library. Then, the correction coefficient is calculated based on the actual hovering time before the UAV lands. The average value of the attenuation coefficient is corrected according to the correction coefficient to obtain the wave attenuation coefficient. Finally, the initial wave influence distance is predicted by multidimensional feature parameters, wave attenuation coefficient, and current water flow velocity. Extract and delete the initial wave-affected distance segment from the current measurement dataset; Based on the preset river hazard level map, low-risk and stable adjacent areas are selected. The initial wave influence distance is used as the compensation length to generate a new flow measurement sub-path that is consistent with the original effective path direction, matches the length, and avoids high disturbances.

[0016] The beneficial effects of this invention are as follows: 1. This invention solves the problem of water surface wave interference caused by rotor airflow disturbance in existing amphibious UAV flow measurement operations. Specifically, it achieves interference identification and data purification through multi-module collaboration. The wave perception trigger module accurately captures the initial state of waves after the rotor stops rotating; the directional influence area extraction module calculates the initial influence radius using advanced edge detection technology; the influence distance prediction module predicts the interference range based on multi-dimensional feature parameters and a scenario-based correction mechanism; and the interference segment trimming module completely eliminates distorted data through nanosecond-level time synchronization and precise truncation, fundamentally solving the measurement deviation caused by sensor attitude instability and signal aliasing, and significantly improving the accuracy and reliability of flow velocity and flow rate calculations. The compensation path planning module selects low-risk stable areas based on the river channel hazard level map and generates a compensation sub-path equivalent to the original path, which can make up for data loss without blindly supplementing measurements. This ensures the integrity and spatial representativeness of flow measurement data, avoids resource waste caused by repeated operations, and significantly improves the operational efficiency of flow measurement in complex waters.

[0017] 2. It can compensate for the unevenness of the initial disturbance and the problem that images acquired in a short period of time cannot fully reflect the true propagation characteristics of waves. Specifically, the dynamic verification and correction module collects the changes in wave state at adjacent times after the monitoring starting point and performs real-time calibration to ensure that the identification of the interference area is without deviation. This allows for the adjustment of the predicted initial wave impact distance, effectively compensating for the possible deviations that may exist if only the initial prediction is relied upon. This makes the judgment of the actual interference distance more accurate and further improves the accuracy of interference section clipping, ensuring the authenticity and reliability of the flow measurement data. Attached Figure Description

[0018] Figure 1 This is the overall system framework diagram of the present invention; Figure 2 This is a schematic diagram of the current measurement interference data removal in this invention; Figure 3 This is a flowchart of the dynamic verification and correction process in this invention; Figure 4 This is a simplified test diagram of the amphibious unmanned aerial vehicle in this invention on a calm, static water surface. 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] Example 1: Because existing amphibious drones do not consider the interference of water surface waves caused by rotor airflow disturbance on the initial flow measurement data in flow measurement operations, and to solve the flow velocity measurement deviation caused by sensor attitude instability and superposition of water flow signal and disturbance signal due to waves, this invention designs a water flow measurement system and method based on amphibious drones. Through multimodal perception and dynamic path compensation, it accurately identifies and eliminates invalid flow measurement sections caused by airflow disturbance under the rotor, and automatically generates an equivalent compensation path while maintaining the total flow measurement length and spatial representativeness.

[0021] Specifically, such as Figure 1 As shown, the system as a whole includes a wave sensing trigger module, a directional impact area extraction module, an impact distance prediction module, an interference segment trimming module, and a compensation path planning module. The system can be set up independently in the controller carried by the amphibious UAV for real-time automatic control, or it can be set up in an external central collaborative control terminal for remote command control.

[0022] The purpose of the wave perception trigger module is to ensure strict synchronization between the acquisition of water surface images and the contact of the floating vessel with the water surface. This module uses the rotor motor stop signal as a hardware interrupt source. When the UAV lands according to the pre-planned landing point in the flow measurement route, the UAV will calibrate the landing point and hover above the landing point until the position is accurate and correct. Then, the UAV controls the rotor speed to slowly decrease to the landing point and shuts off the rotor output power. After detecting that all rotors have stopped rotating for more than a set time, the high-resolution vision camera installed on the UAV is immediately triggered to acquire images. The lens of the vision camera is aligned with the direction of the main water flow. Its purpose is to accurately acquire images of waves in the direction of the UAV's flow measurement drift. After the acquired water surface images are denoised and contrast enhanced, they are transmitted to the directional influence area extraction module for further processing. The trigger condition setting of the wave perception trigger module can accurately capture the initial state of the water surface waves after the rotor stops rotating, avoiding misjudgment of wave characteristics due to improper acquisition timing. This lays an accurate data foundation for subsequent interference area identification and effectively improves the timeliness and accuracy of wave perception.

[0023] The moment when the UAV rotor stops and the wave perception trigger module triggers the first water surface image acquisition command is defined as... At any given time, the directional influence area extraction module receives the output from the real-time drift state detection unit. By analyzing the real-time water flow direction data and combining it with the fuselage heading angle provided by the inertial measurement unit, the current drift direction of the UAV can be calculated. The drift direction is defined as the projection direction of the water flow velocity vector onto the horizontal plane. Subsequently, the orientation influence area extraction module uses the pixel coordinates of the fuselage center in the image. Taking the origin as the starting point, along the current drift direction of the drone Establish a polar coordinate system and extract the polar angle range. The fan-shaped observation area covers the main influence zone of the rotor airflow disturbance propagating along the water flow direction. The angle range can be manually adjusted to effectively capture the outermost boundary of the disturbance wave.

[0024] Within this fan-shaped observation area, the directional influence region extraction module extracts the contours of water waves using the Canny adaptive threshold edge detection algorithm. By precisely truncating the fan-shaped observation area and employing an advanced edge detection algorithm, it can quickly and accurately locate the outermost boundary of the waves. Combined with the camera intrinsic parameter matrix, it achieves accurate conversion from pixel coordinates to actual geographic coordinates, making the initial influence radius calculation results more reliable and providing accurate basic parameters for subsequent influence distance prediction. The algorithm first applies Gaussian filtering to the grayscale image to suppress high-frequency noise, then calculates the image gradient magnitude and direction, retains local gradient maxima points through non-maximum suppression, and finally uses a dual-threshold hysteresis connection strategy to generate continuous edges. The adaptive threshold mechanism dynamically adjusts the high and low thresholds based on local image contrast to ensure stable extraction of wave crest contours under different lighting and water reflection conditions. The extracted wave crest pixel coordinate set... Pre-calibrated camera intrinsic parameter matrix Convert to actual geographic coordinates, camera intrinsic matrix Including focal length , Principal point offset , and radial and tangential distortion coefficients The calibration was completed using a calibration method, and the Euclidean distance between the outermost edge of the wave crest and the projection point of the fuselage center was calculated and denoted as the initial influence radius. .

[0025] The impact distance prediction module further extracts multi-dimensional features from the fan-shaped observation area to quantify the spatial characteristics of wave disturbances. The extraction of multi-dimensional feature parameters comprehensively characterizes wave characteristics from three key dimensions: spatial periodicity, physical amplitude, and texture energy. Compared to single feature parameters, this provides a more comprehensive and accurate reflection of the actual situation of wave disturbances, offering rich and reliable evidence for subsequent wave attenuation coefficient calculation and impact distance prediction. The multi-dimensional feature extraction strategy includes three sub-steps: First, along... A band is selected within the sector area for analysis. A one-dimensional Fourier transform is performed on the grayscale signal within this band to obtain the frequency spectrum. The period corresponding to the peak with the largest amplitude in the frequency spectrum is the wave wavelength. This parameter reflects the spatial periodicity of the disturbing waves. Secondly, the Sobel operator is used to calculate the image gradient, locating the pixel positions of gradient maxima (peaks) and minima (troughs). Combining the camera intrinsic parameter matrix and water surface elevation information, the pixel spacing is converted into the actual physical height difference to obtain the wave height. This parameter characterizes the magnitude of the disturbance; finally, a gray-level co-occurrence matrix of the sector observation area is constructed, with the direction taken as the average of four directions: 0°, 45°, 90°, and 135°, and the matrix energy value... Defined as the sum of squares of all elements in the gray-level co-occurrence matrix, i.e. ,in grayscale and The normalized probability value of simultaneous occurrence under a specified spatial relationship; this parameter comprehensively reflects the intensity and uniformity of the wave texture.

[0026] The database construction submodule pre-builds two data models. The first is a wave feature benchmark library, which covers attenuation coefficient samples under various complex scenarios. This library includes wave attenuation coefficient sample sets under combinations of various water flow types (such as slow current, rapid current, backflow, and turbulent flow) and wind speed levels (calm wind, light wind, gentle wind, and strong wind), such as... Figure 2 and Figure 4 As shown, this invention selects a slow-flowing water type and a calm wind level as examples for subsequent explanations. Slow-flowing water type and calm wind level have minimal impact on UAV flow measurement. Each sample records the attenuation coefficient for the corresponding scenario. The coefficient is obtained by fitting the slope of the exponential decay of disturbance wave height with distance in historical measurements; secondly, there is the mapping relationship model, which accurately quantifies the correlation between rotor airflow intensity, hovering time, and initial wave characteristics. This provides solid data support and a model foundation for the rapid and accurate acquisition of wave attenuation coefficients in different scenarios, improving the system's adaptability to complex environments and describing rotor airflow intensity. Hovering duration With initial wave characteristics The quantitative relationship between them is established by using a three-layer feedforward neural network structure, with the input layer being... and The output layer is the normalized initial feature vector, the hidden layer has twelve nodes, the activation function is ReLU, and the loss function is mean squared error.

[0027] When the strategy is executed, it first retrieves all attenuation coefficient samples matching the scenario from the wave feature benchmark library based on the water flow type identifier in the current flow measurement task metadata and the real-time wind speed sensor readings, and calculates their arithmetic mean to obtain the benchmark mean of the attenuation coefficient. Subsequently, based on the actual hovering time of the drone before landing... Substitute these values ​​into the mapping model to output the corresponding initial feature correction factor, and then calculate the hover duration correction coefficient accordingly. The correction factor is defined as follows: ,in Standard hovering duration, The empirical scaling factor, determined through sensitivity analysis of the mapping relationship model, ultimately determines the wave attenuation coefficient. Through weighted fusion formula Obtain, among which By using preset weights, the baseline mean of the attenuation coefficient is obtained through scene matching and then corrected by combining it with the actual hovering time. This enables personalized calculation of the wave attenuation coefficient, fully considering the impact of different flow measurement scenarios and UAV operation parameters on wave attenuation. This makes the calculated wave attenuation coefficient more in line with the actual situation and improves the accuracy of subsequent impact distance prediction.

[0028] The predicted distance calculation strategy is based on the aforementioned multidimensional feature parameters and wave attenuation coefficient. It quantifies the degree of disturbance attenuation through a pre-defined exponential attenuation model. Existing attenuation models include linear attenuation models, fuzzy function attenuation models, and exponential attenuation models. This invention adopts the exponential attenuation model, first calculating the wave attenuation value. Its expression is ,in To eliminate the net disturbance wave height after removing natural wind and wave interference, by starting from the original... The initial wave influence distance is obtained by subtracting the background wave height when no drones are operating during the same period. Calculated using the following formula: , in, This is an empirical correction factor. For the present By combining the real-time water flow velocity with key factors such as multidimensional characteristic parameters, wave attenuation coefficient, and water flow velocity, the initial wave impact distance can be calculated through a mathematical model. This enables accurate prediction of the impact range of wave disturbances, providing an accurate basis for subsequent interference section trimming and effectively avoiding the loss of effective data or the retention of interference data due to inaccurate judgment of the interference range.

[0029] The interference section trimming module is responsible for cleaning up the flow measurement data affected by disturbances, while the timestamp alignment unit receives the image acquisition timestamp from the wave sensing trigger module. The data is synchronized with the sampling sequences of flow measurement sensors, such as the Doppler current meter and inertial measurement unit on the UAV, at the nanosecond level. Hardware timestamps are used to mark each sensor data packet, ensuring precise alignment of image events and flow velocity data on the timeline. The data truncation unit centralizes the UAV drift flow measurement data... The current position is recorded, and the drift distance is recorded until the initial wave influences the distance. The data segment is retained as the valid data segment, which includes flow velocity, flow direction, and water depth data, thereby ensuring the authenticity and accuracy of the measurement data. Time synchronization enables precise alignment between image events and flow measurement data. Data truncation is performed based on the accurately predicted initial wave influence distance, which can efficiently and accurately eliminate invalid data affected by disturbances, retain true and valid flow measurement data, significantly improve the quality of flow measurement data, and provide a guarantee for the accuracy of subsequent flow velocity and flow rate calculations.

[0030] The task of the compensation path planning module is to generate equivalent alternative paths to compensate for the data loss caused by the clipping. This module receives a river hazard level map output from the obstacle hazard level assessment module. The evaluation logic of the river hazard level map is based on the preliminary analysis and location of obstacles during UAV patrols, and the obstacles are marked according to parameters such as size and type. The map divides the flow measurement area into three categories: high-risk, medium-risk, and low-risk. The levels are stored in grid form with the same resolution as the flow measurement grid. The selection of compensation areas must meet three conditions: first, the area hazard level is low-risk; second, the area is spatially adjacent to the end or beginning of the original effective path, that is, sharing at least one grid boundary; third, areas located in the middle channel of the river are preferred, as their flow is usually the most stable and representative.

[0031] In candidate regions that meet the above conditions, the module adopts an improved A... The path planning algorithm generates compensated paths, and its objective function is to minimize the weighted cost of the paths. Defined as: , in, To plan the path length, Let i be the danger level index of the i-th grid cell on the path. The historical wave influence index of this grid, As a weighting coefficient, the algorithm constrains the starting point of the path to coincide with the coordinates of the end point of the original valid path, and the endpoint to extend in the direction of water flow. The overall direction of the path and the angle between the original path and the path are less than 15 degrees to ensure consistent hydrological representativeness. The finally generated compensation path is sent to the UAV flight control system, which drives it to perform supplementary flow measurement tasks along the new path. The obtained data is seamlessly stitched with the original valid data to form a complete flow profile. This not only ensures the integrity of the flow measurement data, but also ensures that the hydrological representativeness of the compensation data is consistent with the original valid data. It avoids the decrease in operational efficiency and waste of resources caused by blind supplementary measurements, and at the same time improves the safety of flow measurement operations.

[0032] The measurement data trimming method of this invention can be applied to complete flow measurement routes with no obstacles on the water surface, and can also be applied to segmented flow measurement routes with obstacles on the water surface as previously applied by the applicant. In the latter case, in actual flow measurement, if the obstacles are densely distributed, that is, the segmented flow measurement route is too short, then it is necessary to use the initial wave influence distance obtained from the analysis of the first segment of flow measurement. Based on this, the subsequent segmented flow measurement routes will be compared with the initial wave influence distance. The comparison shows that if the length of the segmented flow measurement route is less than the initial wave influence distance... If necessary, the segmented flow measurement route can be cancelled directly, or the drone hovering time can be reduced so that the length of the segmented flow measurement route is greater than the initial wave influence distance. .

[0033] Example 2: Based on Example 1, a dynamic verification and correction module has been added to the system, such as... Figure 3 As shown, since only the initial wave influence distance is predicted, in actual flow measurement scenarios, the water flow velocity may vary due to changes in river topography and wind speed, thus affecting the initial wave influence distance. The image data collected in a short period of time deviates from the actual wave impact range, and the waves generated by the UAV rotor airflow may exhibit uneven initial disturbances. Therefore, the dynamic verification and correction module in this invention... At any given moment, the drone's position drifts to the distance affected by the initial wave. A monitoring start point is set between the points, which can be arbitrarily selected as needed. When the real-time position of the drone deviates from the geographical coordinates of the monitoring start point by less than a preset tolerance, a continuous image acquisition command is triggered, with adjacent moments... and The acquired water surface images are processed by the directional influence region extraction module to obtain their respective multidimensional feature parameter sets. and The module calculates the magnitude of change for each parameter: , , The decision-making strategy stipulates that: when all values ​​in three consecutive adjacent time periods... All values ​​are less than their respective preset thresholds, or the overall wave change rate. Less than the comprehensive threshold When it is determined that the rotor airflow disturbance has been eliminated, the target time is recorded. and corresponding drone locations Actual interference distance That is, the landing location and By dynamically monitoring the rate of change of wave characteristic parameters, the projected distance between the two sides along the water flow direction can be verified and corrected in real time for the initial wave influence distance. This effectively compensates for the possible deviations that may exist if only the initial prediction is relied upon, making the judgment of the actual interference distance more accurate. This further improves the accuracy of interference section cutting and ensures the authenticity and reliability of the flow measurement data.

[0034] The interference section trimming module is responsible for cleaning up the current measurement data affected by disturbances. The timestamp alignment unit receives the image acquisition timestamp from the wave sensing trigger module. The data is synchronized with the sampling sequences of flow sensors such as Doppler current meters and inertial measurement units at the nanosecond level. Hardware timestamps are used to mark each sensor data packet, ensuring precise alignment of image events and flow velocity data on the time axis. The data truncation unit is based on... Calculate the corresponding drift time window ,in From to The module removes the average flow velocity from the flow measurement dataset during the period. All flow velocity, flow direction, and water depth data within the time period are retained only. The subsequent valid data segments ensure that subsequent analyses are based on undisturbed, real water flow information.

[0035] In addition, the mapping relationship model can adopt an online learning mechanism. After each task, the system will conduct actual tests. With prediction The residuals are fed back to the model, and the network weights are fine-tuned through incremental learning algorithms, so that the model is continuously optimized as the number of uses increases, adapting to the personalized disturbance patterns of specific water areas.

[0036] 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. A water flow measurement system based on an amphibious unmanned aerial vehicle (UAV), characterized in that: include: The wave perception trigger module triggers the vision camera to capture images of the water surface facing the direction of the water flow after the drone rotor has completely stopped rotating. The directional influence area extraction module extracts a fan-shaped observation area in the direction of the drone's drift from the water surface image and calculates the actual distance from the outermost boundary of the wave to the center point of the drone's projection as the initial influence radius. The impact distance prediction module extracts multi-dimensional feature parameters representing the spatial periodicity, undulation amplitude, and texture energy of waves in the fan-shaped observation area. Based on the current water flow and wind speed scenario, it obtains the average value of the attenuation coefficient from the preset wave feature benchmark library. Then, it calculates the correction coefficient based on the actual hovering time of the UAV before landing and corrects the average value of the attenuation coefficient to obtain the wave attenuation coefficient. Finally, it predicts the initial wave impact distance through the multi-dimensional feature parameters, the wave attenuation coefficient, and the current water flow velocity. The interference section trimming module extracts and deletes the initial wave-affected distance segments from the flow measurement data. The compensation path planning module selects low-risk and stable adjacent areas based on the preset river hazard level map, and generates a new flow measurement sub-path that is consistent with the original effective path direction, matches the length, and avoids high disturbances, using the initial wave influence distance as the compensation length.

2. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 1, characterized in that: The multidimensional characteristic parameters include wave wavelength, wave height, and wave energy value. The wave wavelength reflects the spatial periodicity of the wave, the wave height characterizes the physical amplitude of the wave, and the wave energy value reflects the intensity and uniformity of the wave texture.

3. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 2, characterized in that: The influence distance prediction module includes a multi-dimensional feature extraction strategy. The multi-dimensional feature extraction strategy includes performing a one-dimensional Fourier transform on the gray-level signal of the fan-shaped observation area using the Fourier transform method, taking the period corresponding to the peak of the frequency spectrum as the wave wavelength, locating the pixel position coordinates of the wave peak and trough through gray-level gradient detection, and calculating the actual wave height by combining the received visual camera intrinsic parameter matrix, constructing a gray-level co-occurrence matrix of the fan-shaped observation area, and calculating the matrix energy value as the wave energy value based on the gray-level co-occurrence matrix.

4. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 1, characterized in that: The influence distance prediction module includes a database construction submodule, which is used to construct a wave feature benchmark library containing wave attenuation coefficient samples corresponding to different water flow types and wind speed scenarios, and is also used to construct a mapping relationship model between rotor airflow intensity, hovering time and initial wave characteristics.

5. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 4, characterized in that: The influence distance prediction module also includes a retrieval strategy, which includes: Based on the water flow type and wind speed parameters of the current flow measurement scenario, attenuation coefficient samples of the same scenario are matched in the benchmark library, and the benchmark mean of attenuation coefficient is obtained through statistical analysis. Based on the actual hovering time of the drone before landing, the hovering time correction coefficient is calculated using the mapping model. The average value of the attenuation coefficient and the hovering time correction coefficient are then weighted and fused according to a preset weight ratio to obtain the wave attenuation coefficient in the current scenario.

6. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 5, characterized in that: The influence distance prediction module also includes a prediction distance calculation strategy, which includes: Based on the current wave height, wave attenuation coefficient, and initial influence radius, the wave attenuation degree at the influence radius is quantified using a preset attenuation model to obtain the wave attenuation value. The initial wave influence distance is calculated by integrating the initial influence radius, multidimensional characteristic parameters, wave attenuation value, and current water flow velocity.

7. A water flow measurement system based on an amphibious unmanned aerial vehicle according to any one of claims 1-6, characterized in that: It also includes a dynamic verification and correction module, which sets the monitoring starting point with the initial wave impact distance. When the real-time position of the UAV coincides with the monitoring starting point, a data acquisition command is triggered, and water surface images are acquired at a certain acquisition frequency. The wave change rate is then analyzed by the water surface images acquired at adjacent times to determine whether the wave disturbance tends to stabilize, thereby dynamically correcting the predicted initial wave impact distance to obtain the actual interference distance.

8. The water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 7, characterized in that: The dynamic verification correction module includes a determination strategy, which includes: Multidimensional feature parameters of waves in the sector observation area at adjacent time points are extracted, and the change amplitude of each parameter in the multidimensional feature parameters is calculated. When the variation amplitude of each parameter in multiple consecutive adjacent time periods is less than the preset amplitude, or when the comprehensive wave variation rate calculated by combining the variation amplitudes of all parameters is less than the preset rate, it is determined that the wave interference of the rotor airflow on the water surface has been eliminated, and the corresponding target time and the current position of the UAV are recorded. The actual interference distance of the waves is determined based on the landing position and the current position of the UAV.

9. A water flow measurement system based on an amphibious unmanned aerial vehicle according to claim 8, characterized in that: The interference segment trimming module includes a timestamp alignment unit and a data truncation unit. The timestamp alignment unit synchronizes the visual trigger time with the sampling sequence of the flow sensor. The data truncation unit removes the flow data segments affected by the disturbance based on the drift time window corresponding to the corrected actual interference distance.

10. A method for measuring water flow based on amphibious unmanned aerial vehicles, characterized in that: Includes the following steps: After the drone rotors have completely stopped rotating, the vision camera is triggered to capture an image of the water surface facing the direction of the water flow. In the water surface image, a fan-shaped observation area in the direction of the UAV's drift is extracted, and the actual distance from the outermost boundary of the wave to the center point of the fuselage projection is calculated as the initial influence radius. Multidimensional feature parameters characterizing wave spatial periodicity, undulation amplitude, and texture energy are extracted from the fan-shaped observation area. Based on the current water flow and wind speed scenario, the average value of the attenuation coefficient is obtained from the preset wave feature benchmark library. Then, the correction coefficient is calculated based on the actual hovering time before the UAV lands. The average value of the attenuation coefficient is corrected according to the correction coefficient to obtain the wave attenuation coefficient. Finally, the initial wave influence distance is predicted by multidimensional feature parameters, wave attenuation coefficient, and current water flow velocity. Extract and delete the initial wave-affected distance segment from the current measurement dataset; Based on the preset river hazard level map, low-risk and stable adjacent areas are selected. The initial wave influence distance is used as the compensation length to generate a new flow measurement sub-path that is consistent with the original effective path direction, matches the length, and avoids high disturbances.

Images