Acoustic Doppler flow velocity measurement method and system

Abstract

The invention discloses an acoustic Doppler flow velocity measurement method and system, and the method comprises the steps: generating an emission parameter combination adaptive to a current hydrological environment according to the depth characteristics and precision requirements of a target water area; original echo signals are collected based on the emission parameter combination, and a three-dimensional flow velocity vector of temperature compensation is output; inputting the three-dimensional flow velocity vector into the dynamic segmentation network to generate a layered flow velocity distribution model; calling a layered flow velocity distribution model, fusing fluxgate compass azimuth data to carry out flow direction compensation, triggering a deviation correction algorithm when equipment inclination exceeds a limit, and outputting absolute flow velocity distribution of pose calibration; and according to absolute flow velocity distribution and echo intensity dynamic range constraint, calculating an error interval of each depth unit, and outputting a final flow velocity profile with confidence identification. According to the embodiment of the invention, the environmental adaptability and data reliability of flow velocity measurement can be improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of flow velocity measurement, and particularly relates to an acoustic Doppler flow velocity measurement method and system. BACKGROUND

[0002] Traditional acoustic Doppler flow velocity measurement technology is limited by a fixed transmission mode, and it is difficult to balance the measurement requirements of high precision in shallow water and large range in deep water, resulting in insufficient adaptability of hydrological data. In addition, the Doppler frequency shift is easily affected by environmental temperature and device pose, and existing methods mostly rely on manual calibration, which is low in efficiency and poor in real-time performance. Although some schemes attempt to improve accuracy through hardware compensation, they lack dynamic parameter optimization and intelligent correction mechanisms, and the error is significant especially under complex hydrological conditions. SUMMARY

[0003] The purpose of the present application is to provide an acoustic Doppler flow velocity measurement method and system to solve the problems in the prior art and improve the environmental adaptability and data reliability of flow velocity measurement.

[0004] One embodiment of the present application provides an acoustic Doppler flow velocity measurement method, which comprises: According to the depth characteristics and precision requirements of the target water area, a preset high-resolution wideband mode or a large-range narrowband mode is automatically activated to generate a transmission parameter combination adapted to the current hydrological environment; Based on the transmission parameter combination, raw echo signals are collected, real-time temperature data of the transducer are combined, a pre-trained compensation model is used to correct the Doppler frequency shift, and a temperature-compensated three-dimensional flow velocity vector is output; The three-dimensional flow velocity vector is input into a dynamic segmentation network, the profile layer is segmented according to a preset depth unit size, echo intensity normalization processing is completed, and a layered flow velocity distribution model is generated; The layered flow velocity distribution model is called, magnetic fluxgate compass azimuth data is fused for flow direction compensation, a correction algorithm is triggered when the device inclination exceeds the limit, and an absolute flow velocity distribution after pose calibration is output; According to the absolute flow velocity distribution and echo intensity dynamic range constraints, the error interval of each depth unit is calculated, and the final flow velocity profile with a confidence identifier is output.

[0005] Optionally, the preset high-resolution wideband mode or the large-range narrowband mode is automatically activated according to the depth characteristics and precision requirements of the target water area to generate a transmission parameter combination adapted to the current hydrological environment, which comprises: According to the sonar depth detection data of the target water area and a preset precision threshold, a depth-precision characteristic matrix is calculated, and a mode selection decision factor is output; Based on the decision factor, a frequency configuration library of the high-resolution wideband mode or the large-range narrowband mode is activated, and a frequency parameter set containing a center frequency, a bandwidth and a pulse width is extracted. Combine the current water body salinity sensor data, dynamically adjust the pulse repetition period parameter by sound speed compensation, and generate a basic transmission parameter set; According to the basic transmission parameter set and the frequency parameter set, the sound beam opening angle and the transmission power combination are optimized, and the final transmission parameter combination is output.

[0006] Optionally, the original echo signal is collected based on the transmission parameter combination, combined with the real-time temperature data of the transducer, the Doppler frequency shift is corrected by the pre-trained compensation model, and the temperature-compensated three-dimensional flow velocity vector is output, including: According to the transmission parameter combination, drive the transducer array to emit frequency-modulated pulse signals, and collect the IQ quadrature components of the original echo signal; Synchronously acquire real-time data of transducer temperature sensor, input pre-trained frequency shift-temperature mapping model, and output temperature drift compensation coefficient; Complex signal demodulation is performed on the IQ components, the Doppler frequency shift spectrum is calculated by the window function weighted FFT, and the initial frequency shift matrix is generated; Apply the temperature drift compensation coefficient to the initial frequency shift matrix, and use the three-beam spatial decoupling algorithm to decompose the XYZ axial components; The decomposed axial components are subjected to motion interference filtering, and the temperature-compensated three-dimensional flow velocity vector containing direction components is output.

[0007] Optionally, the three-dimensional flow velocity vector is input into the dynamic segmentation network, the profile layer is segmented according to the preset depth unit size, the echo intensity normalization processing is completed, and the layered flow velocity distribution model is generated, including: The three-dimensional flow velocity vector is input into the dynamic segmentation network, the profile layer is divided according to the preset depth unit size, and the depth index vector is generated; Based on the depth index vector, the signal-to-noise ratio features of each layer of echo signals are extracted, and the outlier rejection is completed through the adaptive threshold; The layered signal after rejection is fitted with the energy weighted least squares method to output the unnormalized layered flow velocity; Combined with the transducer transmission energy attenuation model, the depth compensation coefficient is calculated and the flow velocity amplitude normalization is completed, and the layered flow velocity distribution model containing direction components is generated.

[0008] Optionally, the layered flow velocity distribution model is called, the flow direction compensation is performed by fusing the magnetic flux gate compass azimuth data, the deviation correction algorithm is triggered when the device inclination exceeds the limit, and the pose calibrated absolute flow velocity distribution is output, including: The depth unit data in the layered flow velocity distribution model is called, the three-axis azimuth angle of the magnetic flux gate compass is synchronously read, and the flow direction reference vector is generated; Combined with the flow direction reference vector and the layered flow velocity direction component, the flow direction compensation is completed through the spherical coordinate conversion, and the relative flow velocity distribution is output. Real-time monitoring of the tilt sensor data, when the device tilt angle exceeds the preset threshold, triggering the quaternion rotation correction, generating a rotation correction matrix; Apply the rotation correction matrix to the relative flow velocity distribution, convert the device coordinate system to the geodetic coordinate system; Eliminate the Coriolis force deviation caused by the earth's rotation through coordinate projection, output the absolute flow velocity distribution of pose calibration.

[0009] Optionally, according to the absolute flow velocity distribution and the echo intensity dynamic range constraint, the error interval of each depth unit is calculated, and the final flow velocity profile with confidence identification is output, including: According to the absolute flow velocity distribution data of each depth unit, the echo intensity dynamic range feature is extracted, and the signal-to-noise ratio decay curve is generated; Based on the signal-to-noise ratio decay curve, the flow velocity measurement variance is calculated through the Cramer-Rao lower bound theory, and the initial error interval is output; Fusion of acoustic scattering intensity and water turbidity data, construction of multi-physical field coupled error correction model, generation of correction error coefficient; Apply the correction error coefficient to the initial error interval, and verify the confidence probability by Monte Carlo simulation; According to the preset confidence level, the confidence identification of the flow velocity value of each depth unit is marked, and the final flow velocity profile is output.

[0010] Another embodiment of the application provides an acoustic Doppler flow velocity measurement system, the system comprising: The activation module is used to automatically activate the preset high-resolution broadband mode or large-range narrowband mode according to the depth characteristics and accuracy requirements of the target water area, and generate a transmission parameter combination suitable for the current hydrological environment; The correction module is used to collect original echo signals based on the transmission parameter combination, combine the real-time temperature data of the transducer, correct the Doppler frequency shift through the pre-trained compensation model, and output the temperature-compensated three-dimensional flow velocity vector; The processing module is used to input the three-dimensional flow velocity vector into a dynamic segmentation network, segment the profile layer according to a preset depth unit size, complete echo intensity normalization processing, and generate a layered flow velocity distribution model; The compensation module is used to call the layered flow velocity distribution model, fuse the fluxgate compass azimuth data for flow direction compensation, trigger the correction algorithm when the device tilt exceeds the limit, and output the absolute flow velocity distribution of pose calibration; The output module is used to calculate the error interval of each depth unit according to the absolute flow velocity distribution and the echo intensity dynamic range constraint, and output the final flow velocity profile with confidence identification.

[0011] Still another embodiment of the present application provides a storage medium having a computer program stored therein, wherein the computer program is configured to execute the method described in any of the above when running.

[0012] Still another embodiment of the present application provides an electronic device comprising a memory having a computer program stored therein and a processor configured to execute the computer program to perform the method described in any of the above.

[0013] Compared with the prior art, the acoustic Doppler flow velocity measurement method provided by the application generates a combination of transmission parameters suitable for the current hydrological environment according to the depth characteristics and accuracy requirements of the target water area, acquires original echo signals based on the combination of transmission parameters, outputs a temperature-compensated three-dimensional flow velocity vector, inputs the three-dimensional flow velocity vector into a dynamic segmentation network to generate a layered flow velocity distribution model, calls the layered flow velocity distribution model, fuses magnetic fluxgate compass azimuth data to perform flow direction compensation, triggers a correction algorithm when the device inclination exceeds the limit, and outputs an absolute flow velocity distribution after pose calibration, calculates the error interval of each depth unit according to the absolute flow velocity distribution and echo intensity dynamic range constraints, and outputs a final flow velocity profile with a confidence identifier, thereby improving the environmental adaptability and data reliability of flow velocity measurement. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 A hardware structure block diagram of a computer terminal of an acoustic Doppler flow velocity measurement method provided by an embodiment of the application is shown in the figure. Figure 2 A flowchart of an acoustic Doppler flow velocity measurement method provided by an embodiment of the application is shown in the figure. Figure 3 A structure diagram of an acoustic Doppler flow velocity measurement system provided by an embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0015] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the application and cannot be explained as a limitation of the application.

[0016] An embodiment of the application first provides an acoustic Doppler flow velocity measurement method, which can be applied to an electronic device such as a computer terminal, specifically a general computer, etc.

[0017] The following will be described in detail taking a computer terminal as an example. Figure 1 A hardware structure block diagram of a computer terminal of an acoustic Doppler flow velocity measurement method provided by an embodiment of the application is shown in the figure. Figure 1 As shown in the figure, the computer device comprises a processor, a memory and a network interface connected through a system bus, wherein the memory can comprise a non-volatile storage medium and an internal memory.

[0018] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions which, when executed, can cause the processor to perform any acoustic Doppler flow velocity measurement method.

[0019] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0020] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, can cause the processor to perform any acoustic Doppler flow velocity measurement method.

[0021] The network interface is configured to perform network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that, Figure 1 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0022] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0023] Referring to Figure 2 The embodiments of the present application provide an acoustic Doppler flow velocity measurement method, which can include the following steps: S201, automatically activating a preset high-resolution wideband mode or a large-range narrowband mode according to the depth characteristics and accuracy requirements of the target water area, to generate a combination of transmission parameters adapted to the current hydrological environment; Specifically, the depth-accuracy characteristic matrix can be calculated according to the sonar depth detection data of the target water area and the preset accuracy threshold, and a mode selection decision factor is outputted; Sonar depth detection and accuracy threshold matching System startup, first through the integrated high-frequency sonar probe (frequency range 500 kHz-1 MHz) to the water bottom short pulse signal. Receiver capture time difference (Time of Flight, TOF), combined with the pre-stored sound velocity profile (such as 1500 m / s reference value), calculate the real-time depth value (Depth Value, DV) of the target water area. At the same time, the user's preset precision threshold (Precision Threshold, PT) is loaded into the decision system - for example, the coastal hydrological survey requires flow velocity error ≤0.01 m / s (high precision), while the marine monitoring can accept ≤0.1 m / s (regular precision). The control core will input DV and PT into the depth-precision feature matrix (Depth-Precision Feature Matrix, DPFM). The matrix is a pre-trained two-dimensional lookup table, with the horizontal axis as the depth classification (such as 0-20m shallow water area, 20-100m mid-depth area, >100m deep water area), and the vertical axis as the precision level (such as 0.001m / s, 0.01m / s, 0.1m / s). By matching the depth range where DV is located and the precision level to which PT belongs, DPFM outputs an environmental adaptation index (Environmental Adaptation Index, EAI) with a value range of 0 to 100. For example, when DV=35m (mid-depth area) and PT=0.02m / s (high precision), EAI=85.

[0024] Decision factor generation logic: EAI is input into the mode decision engine (Mode Decision Engine, MDE). The engine has a built-in fuzzy logic rule base: if EAI≥80 (high environmental adaptation requirement), activate the high-resolution broadband mode (High-Resolution Broadband Mode, HRBM); if EAI≤60 (low adaptation requirement), activate the long-range narrowband mode (Long-Range Narrowband Mode, LRNM); intermediate values trigger mixed mode. The final output is the mode selection decision factor (Mode Selection Decision Factor, MSDF), which is an enumerated variable (HRBM or LRNM). For example, in the complex topography of estuary area, DV=15m and PT=0.005m / s, EAI=92→MSDF=HRBM; while in the ocean observation, DV=300m and PT=0.1m / s, EAI=45→MSDF=LRNM.

[0025] Anti-interference fault-tolerant mechanism: To prevent abnormal data from triggering false alarms, the system is equipped with a confidence verification layer (CVL). If the standard deviation (SD) of the sonar depth measurement results exceeds 5% of the depth value (e.g., SD > 2m when DV = 40m) for three consecutive times, the data is deemed unreliable. At this time, the system automatically switches to the redundant differential pressure depth sensor (accuracy ±0.1% FS) and recalculates the DPFM. At the same time, an error code is recorded in the log to ensure the reliability of the decision factor.

[0026] Based on the decision factor, the frequency configuration library of high-resolution wideband mode or large-range narrowband mode is activated, and the frequency parameter set containing center frequency, bandwidth, and pulse width is extracted; Mode-specific frequency configuration library call: According to the enumeration value of MSDF, the system activates the corresponding frequency configuration library (FCL). The HRBM library is for shallow water high precision scenarios, and its core parameters are: Center frequency (CF): higher frequency band (e.g., 1.2MHz), to improve spatial resolution; Bandwidth (BW): wide frequency band (e.g., 100kHz), to enhance Doppler shift detection sensitivity; Pulse width (PW): short pulse (e.g., 0.2ms), to reduce depth unit size.

[0027] The LRNM library is suitable for deep water large range: CF: lower frequency band (e.g., 75kHz), to reduce sound wave attenuation; BW: narrow frequency band (e.g., 10kHz), to increase penetration distance; PW: long pulse (e.g., 5ms), to improve signal-to-noise ratio; The library stores multiple preset parameter combinations (e.g., 10 configurations for HRBM and 8 configurations for LRNM), arranged according to water depth gradient.

[0028] Dynamic parameter extraction algorithm: Taking HRBM as an example, the system performs nearest neighbor interpolation (NNI) according to the current DV in the library. If DV = 25m, the weighted average value of the two configurations of 20m and 30m in the library is selected: CF = (1.5MHz x 0.7 + 1.0MHz x 0.3) = 1.35MHz; BW = (120kHz x 0.7 + 80kHz x 0.3) = 108kHz; PW = (0.15ms x 0.7 + 0.25ms x 0.3) = 0.18ms; Thus, the Frequency Parameter Set (FPS) is generated, containing the three core parameters: CF, BW, and PW.

[0029] Environmental Adaptation Fine-tuning: Further Fusion of Water Type Data: If the Salinity Sensor (SS) reports that the water body is fresh water (salinity < 0.5‰), the CF of the HRBM is automatically increased by 5% to compensate for the decrease in sound speed; if it is a high-turbidity water area (turbidity > 100 NTU), the PW of the LRNM is increased by 20% to combat scattering loss. This process is achieved through the Type-Parameter Mapping Table (TPMT), ensuring optimal adaptation of the parameter set.

[0030] Combining current water salinity sensor data, dynamically adjusting the pulse repetition period parameter through sound speed compensation, generating the basic transmission parameter set; Sound Speed Compensation Principle: The propagation speed of sound waves in seawater (Sound Velocity, SV) is affected by salinity (Salinity, S), temperature (T), and pressure (P). The empirical relationship is: *SV≈1449.2 + 4.6T - 0.055T 2 + 0.0003T 3 + (1.34-0.01T)(S-35) + 0.016P*.

[0031] The system pre-stores the Sound Velocity Compensation Lookup Table (SVCLT). Input the real-time data of the salinity sensor (e.g., S = 32‰), the temperature sensor data (e.g., T = 15℃), and the pressure value converted from DV (every 10m water depth ≈ 1 atmosphere), and the current SV is output (example: 1502m / s).

[0032] Dynamic Adjustment of Pulse Repetition Period: The Pulse Repetition Period (PRP) needs to meet two constraints: Maximum Unambiguous Range: PRP ≥ 2 x DV / SV; Maximum Measurable Flow Rate: PRP ≤ Wavelength / (4 x Maximum Flow Rate); For example, when DV=100m, SV=1500m / s, PRP needs at least 133ms; while in HRBM mode with central wavelength λ=1.25mm (CF=1.2MHz), if the maximum flow velocity is 2m / s, PRP needs ≤0.156ms —— at this time, the system triggers the Constraint Conflict Resolution (CCR) algorithm: priority is given to the distance constraint, PRP is set to 135ms, and the flow velocity range is expanded through multi-pulse coherent processing.

[0033] Basic Emission Parameter Set generation: Integrate all parameters before: mode decision result (MSDF); frequency parameter set (CF, BW, PW in FPS); dynamically calculated PRP; packaged as Basic Emission Parameter Set (BEPS). For example: {MSDF: HRBM, CF: 1.35MHz, BW: 108kHz, PW: 0.18ms, PRP: 0.5ms}. This set has met the basic detection needs of the hydrological environment.

[0034] According to the basic emission parameter set and the frequency parameter set, the beam opening angle and the transmit power combination are optimized, and the final transmit parameter combination is output.

[0035] Beam Opening Angle adaptive optimization: Beam Opening Angle (BOA) directly affects spatial resolution and signal-to-noise ratio. The system calculates the theoretical opening angle based on PW and CF in BEPS: theoretical BOA ∝ SV × PW / transducer aperture.

[0036] But need to adjust according to the actual scene: In the turbulent area (such as estuary), reduce BOA to 15° to improve the flow direction resolution; In the uniform flow field (such as deep sea), expand BOA to 30° to increase the sampling volume; Through the Beam Opening Angle-Environment Correlation Model (BECM), the best value is dynamically selected, and the phase controller of the transducer array is driven to realize electronic deflection.

[0037] Transmit Power closed-loop control: The setting of Transmit Power (TP) needs to balance the detection distance and the power consumption of the equipment: The initial value is determined by the empirical formula: TP_init = k × DV 2 × attenuation coefficient (k is a device constant).

[0038] Real-time feedback regulation: the receiver monitors the echo intensity index (EII), if the EII is lower than the threshold (such as-60dB), the TP is increased by 5% steps until it meets the standard; if the EII is too high (such as >-40dB) for a long time, the TP is reduced to prevent saturation distortion.

[0039] This process adopts proportional-integral-derivative control (PID control) to avoid power oscillation.

[0040] Final parameter combination output: Integrate the optimized BOA and TP into BEPS to form the final emission parameter combination (FEPC), for example: { Mode: HRBM, Center_Freq: 1.35MHz, Bandwidth: 108kHz, Pulse_Width: 0.18ms, PRP: 0.5ms, Beam_Angle: 20°, Power: 200W }。 The group is written into the FPGA emission controller synchronously, completing the entire adaptive parameter configuration process.

[0041] This step selects the most suitable acoustic emission mode by intelligently analyzing the water depth and measurement accuracy requirements. The high-resolution wideband mode is suitable for fine measurement in shallow water, and the time domain resolution is improved by using wideband signals. The large-range narrowband mode is optimized for deep water environments, and narrow-band pulses are used to ensure signal penetration. The system dynamically adjusts key parameters such as center frequency and bandwidth based on real-time hydrological data to achieve the optimal balance between measurement performance and energy consumption, solving the problem of relying on manual experience in mode selection in traditional flow measurement. Through adaptive parameter configuration, the applicability of the device in different water environments is significantly improved. It can not only ensure millimeter-level flow rate resolution in shallow water, but also achieve hundreds of meters of effective detection in deep water, greatly expanding the operating range of a single device.

[0042] S202, based on the emission parameter combination, collect the original echo signal, combine the real-time temperature data of the transducer, correct the Doppler shift through the pre-trained compensation model, and output the temperature-compensated three-dimensional flow rate vector; Specifically, the transducer array can be driven to emit a frequency-modulated pulse signal according to the combination of emission parameters, and the IQ quadrature components of the original echo signal can be collected; Transmit control and signal generation: The system drives the transducer array according to the generated combination of emission parameters (including parameters such as center frequency, bandwidth, pulse width, etc.). The transducer array is usually composed of multiple piezoelectric ceramic units arranged in a specific geometry (such as a four-element cross array), and each unit is independently controlled by a high-voltage pulsing circuit (HVPC). Taking a chirp signal with a center frequency of 150 kilohertz (kHz) and a bandwidth of 30 kilohertz (kHz) as an example, the HVPC generates a high-voltage pulse (typical value 300 volts) to excite the ceramic unit to generate a sound wave of a specific frequency. The chirp signal uses linear frequency modulation (LFM) with a starting frequency of 135 kilohertz (kHz) and an ending frequency of 165 kilohertz (kHz), and a pulse width of 10 milliseconds (ms). The sound wave is radiated in a conical beam (beam opening angle 7 degrees) towards the water body, penetrating the target water area.

[0043] Echo signal collection and IQ demodulation: When the sound wave encounters scatterers in the water flow (such as plankton, bubbles, suspended particles), it generates an echo. The transducer array switches to receive mode, and the weak echo signal is captured through a preamplifier (gain 60 decibels) and a band-pass filter (bandwidth matching the emission signal). The received signal is input into a quadrature demodulator (QD) module, which includes a local oscillator (LO) and a mixer. The LO generates a reference signal with the same origin as the center frequency of the emission, which is divided into two paths: one path with a phase of 0 degrees (I reference), and the other path with a phase offset of 90 degrees (Q reference). After mixing the echo signal with the I and Q reference signals, the high-frequency components are filtered out through a low-pass filter (cutoff frequency 5 kilohertz), and the in-phase (I) and quadrature (Q) components of the baseband signal are output. The I / Q components together form the raw echo signal (RES) in complex form, which fully retains the echo amplitude and phase information.

[0044] Signal synchronization and timing control: The acquisition process strictly follows the pulse repetition interval (PRI). The PRI is determined by the sound speed in water (calculated by the salinity-temperature-pressure model) and the maximum detection depth. For example, at a sound speed of 1500 meters per second (m / s) and a detection depth of 100 meters, the PRI is at least 133 milliseconds (ms). The timing controller (TC) synchronizes the transmission pulse, the reception window, and the data sampling clock (typical sampling rate of 100 kilohertz). Each depth cell (DC) corresponds to a fixed time window (e.g., 0.5 milliseconds), ensuring that the I / Q signals are stored by depth layer. Finally, the raw echo signal is temporarily stored in the cache area as a complex sequence (I+jQ) for subsequent processing.

[0045] Synchronize the real-time data of the transducer temperature sensor, input the pre-trained frequency shift-temperature mapping model, and output the temperature drift compensation coefficient; Temperature monitoring and data synchronization: The transducer base is embedded with a high-precision platinum resistance thermometer (PRT) to monitor the transducer operating temperature in real time (typical range -5°C to +40°C) at a sampling rate of 10 times per second. After digitizing the temperature data by an analog-to-digital converter (ADC) with a resolution of 0.01°C, the data synchronization module (DSM) associates the temperature data with the raw echo signal. The DSM aligns the temperature data with the echo signal based on the timestamp, ensuring that the temperature value corresponding to each transmission pulse is accurately recorded.

[0046] Frequency shift-temperature mapping model construction: This model is pre-trained through calibration experiments before the device is shipped. The specific process is as follows: Place the transducer in a temperature-controlled water tank and change the temperature at intervals of 5°C in the range of 5°C to 40°C; at each temperature point, transmit a fixed frequency sound wave (e.g., 150 kilohertz) and receive the echo of the static water body; calculate the frequency shift (i.e., temperature drift frequency shift Δf_temp) of the echo center frequency relative to the transmission frequency through spectral analysis. Finally, establish a mapping dataset of temperature (T) and frequency shift (Δf_temp), and generate a function using a cubic polynomial fitting: Δf_temp = a × T 3 + b × T 2 + c × T + d.

[0047] Where a, b, c, d are fitting coefficients (example values: a=0.0003, b=-0.02, c=0.5, d=-10). The model accuracy requires that the frequency shift prediction error be less than 0.1 hertz.

[0048] Compensation coefficient real-time calculation: At runtime, the pre-trained model receives the real-time temperature value (T_current) and calculates the theoretical frequency shift Δf_temp_pred. Since the frequency shift is directly related to the flow rate calculation (Doppler formula: v = c x Δf / (2 x f0)), it is necessary to convert the frequency shift into an equivalent velocity compensation amount. Define the temperature drift compensation coefficient (K_temp) as: K_temp = -Δf_temp_pred / f0. Where f0 is the center frequency of transmission (such as 150 kHz). This coefficient is stored in vector form (including size and sign), and is directly applied to the frequency shift matrix in the future. For example, when T_current = 25°C, the model outputs Δf_temp_pred = 15 Hz, then K_temp = -0.0001 (dimensionless).

[0049] Complex signal demodulation, Doppler frequency shift spectrum calculation through window function weighted FFT, generate initial frequency shift matrix; Complex signal preprocessing: Read the I / Q component sequence (length corresponding to the sampling point number of a single pulse, such as 2000 points) from the buffer area. First, perform DC offset correction (remove I / Q mean), and then enhance signal resolution through Hilbert transform. Next, add a window function (Window Function) to each depth unit signal segment (such as 50 sampling points) to suppress spectral leakage. The commonly used Hanning window (Hanning Window): W(n) = 0.5 - 0.5 x cos(2πn / (N-1)). Where n is the window index (0 to N-1), and N is the window length. The window function makes the signal smoothly transition to zero at both ends, reducing the truncation effect.

[0050] Fast Fourier transform and spectral analysis: Perform Fast Fourier Transform (FFT) on the windowed I / Q complex signal segment (real part I, imaginary part Q). The FFT point number is usually a power of 2 (such as 64 points), and the output is a complex spectrum. Calculate the power spectral density (PSD) by squaring the spectral amplitude: PSD(k) = |I_fft(k)| 2 + |Q_fft(k)| 2where k is the frequency index (0 to N_fft-1). The PSD main lobe is located by a peak detection algorithm (e.g. local maximum search), and its corresponding frequency index k_peak is converted to physical frequency: f_peak = (k_peak x Fs) / N_fft. Fs is the sampling rate (e.g. 100 kHz), and N_fft is the FFT point number (e.g. 64). This f_peak is the Doppler shift Δf_doppler.

[0051] Doppler shift matrix generation: The above process is repeated for each depth cell to obtain the Doppler shift sequence along the depth direction. Since the transducer array contains multiple beams (e.g. beams 1, 2, 3), the Doppler shift sequence of each beam needs to be calculated separately. Finally, the initial Doppler shift matrix (DSM) is constructed, with dimensions [number of beams x number of depth cells]. For example, when a three-beam system detects 50 depth cells, the DSM is a 3-row x 50-column matrix, and each element stores the Δf_doppler value (unit: Hz) of the corresponding beam and depth.

[0052] Apply the temperature drift compensation coefficient to the initial Doppler shift matrix, and use the three-beam spatial decoupling algorithm to decompose the XYZ axis components; Temperature drift compensation: Read the initial Doppler shift matrix DSM and the temperature compensation coefficient K_temp. For each element in the matrix (i.e. the Doppler shift value Δf_doppler of each beam-depth cell), compensation is performed: Δf_comp = Δf_doppler - (K_temp x f0).

[0053] where f0 is the transmit center frequency. The compensated Doppler shift Δf_comp eliminates the influence of transducer temperature drift and only retains the Doppler shift caused by water flow motion. The compensated data overwrite the original DSM matrix.

[0054] Three-beam geometric decoupling principle: The transducer array adopts a three-beam spatial symmetric layout (e.g. beam 1 points straight ahead, beams 2 and 3 have an elevation angle θ = 20° and a 120° azimuth angle separation). The Doppler shift Δf_comp measured by each beam satisfies: Δf_comp = (2 x f0 x v_beam) / c, where v_beam is the component of the water flow velocity in the beam direction. c is the sound speed (calculated in real time through a salinity-temperature-pressure model). Convert the beam velocity component to the device coordinate system (XYZ axes): X-axis (device forward direction): v_x ∝ Δf_beam1; Y-axis (lateral direction): v_y ∝ (Δf_beam3 - Δf_beam2); Z-axis (vertical): v_z ∝ (Δf_beam2 + Δf_beam3)×cosθ - Δf_beam1×sinθ.

[0055] Decoupling with matrix operation: Define a transformation matrix T (3×3) with elements determined by beam geometry. For each depth cell, construct a beam frequency shift vector F = [Δf_beam1, Δf_beam2, Δf_beam3]^T. Perform matrix multiplication: V_xyz = T × F. Where V_xyz = [v_x, v_y, v_z]^T, represents the XYZ-axis flow velocity components (unit: meter / second). An example of transformation matrix T: T = [ k, 0, 0 ] [ 0, -k / √3, k / √3 ] [ -k sinθ, k cosθ / 2, k cosθ / 2 ] k is a proportionality coefficient (k = c / (2f0 cosθ)). After decoupling by depth cell, a three-dimensional flow velocity vector matrix (dimension 3×number of depth cells) is obtained.

[0056] Motion interference filtering on the decomposed axial components, output temperature-compensated three-dimensional flow velocity vector with direction components.

[0057] Motion interference source analysis: The motion of the device-carrying platform (such as the hull of a ship, a buoy) itself (roll, pitch, heave) will introduce additional frequency shifts, polluting the water flow signal. Such interference is characterized by low-frequency vibration (0.1-2 Hz) and strong correlation in each beam. It is necessary to separate the water flow signal (broadband random) from the platform motion signal (narrowband periodic).

[0058] Adaptive notch filter implementation: Apply an adaptive notch filter (ANF) independently to each axial component (v_x, v_y, v_z). Take the vertical velocity v_z as an example: Reference signal generation: read the vertical acceleration data a_z of the platform attitude sensor (such as the inertial measurement unit IMU), and integrate to obtain the velocity estimate v_platform.

[0059] Filter structure: use a second-order infinite impulse response (IIR) filter, with transfer function: H(z) = (1 -2cosω0 z -1 + z -2 ) / (1 - 2r cosω0 z -1 + r 2 z-2 )。

[0060] where ω0is the interference frequency (determined by FFT analysis of the v_zspectrum main peak), r is the bandwidth control parameter (0.95-0.99).

[0061] Parameter update: use the least mean square algorithm (LMS) to dynamically adjust the filter coefficients, so that the mean square error of the output signal (v_z_filtered) and v_platform is minimized.

[0062] Output three-dimensional flow velocity vector: After filtering, the water flow signal retains the frequency band of 0.01 Hz to the Nyquist frequency (such as 5 Hz). For each depth unit, the filtered three-axis components (v_x, v_y, v_z) are combined into a vector form: V = [v_x, v_y, v_z]^T.

[0063] The vector contains the size (flow velocity amplitude) and direction (three-dimensional direction cosine in the device coordinate system). Finally, output the temperature-compensated three-dimensional flow velocity vector matrix (dimension 3 x depth unit number) as the input of subsequent layering processing.

[0064] After the system transmits the customized acoustic wave signal, it synchronously collects the orthogonal IQ components in the echo and fuses high-precision temperature sensor data. Through a pre-trained neural network model, the acoustic characteristic drift of the transducer material caused by temperature changes is compensated in real time, and the Doppler frequency shift error caused by this is eliminated. Using three-beam spatial decoupling technology, the corrected frequency shift signal is converted into accurate three-dimensional flow velocity components, breaking through the technical bottleneck of traditional Doppler flowmeters that are greatly affected by environmental temperature, and reducing temperature drift error. Three-dimensional vector measurement can simultaneously obtain water flow velocity size and direction information, providing complete data basis for complex flow field analysis.

[0065] S203, input the three-dimensional flow velocity vector into the dynamic segmentation network, segment the profile layer according to the preset depth unit size, complete the echo intensity normalization processing, and generate a layered flow velocity distribution model; Specifically, the three-dimensional flow velocity vector can be input into the dynamic segmentation network, the profile layer is divided according to the preset depth unit size, and a depth index vector is generated. The system receives temperature-compensated three-dimensional flow velocity vector data (including velocity components in the eastward U, northward V, and vertical W directions, as well as corresponding timestamps and depth information). The core of the Dynamic Segmentation Network (DSN) is an intelligent layering module based on a Depthwise Convolutional Neural Network (DCNN). This module first reads the user's preset Depth Cell Size (DCS) parameter (for example, 0.5 meters, 1 meter, or 2 meters, set according to the measurement accuracy requirements). Assuming that the total depth of the current water area is 50 meters and the DCS is set to 1 meter, the system will automatically divide the entire water depth range into 50 consecutive depth layers (Depth Layer). Each depth layer has a unique Depth Index (DI), numbered sequentially from the water surface (DI=0) to the bottom (DI=49). The DSN scans and cuts the three-dimensional flow velocity vector data in the depth dimension through a sliding window mechanism with a step size of DCS. For example, the depth layer corresponding to DI=0 contains all flow velocity data points within 0-1 meters of water depth; DI=1 corresponds to 1-2 meters, and so on. During the cutting process, the DSN automatically identifies and excludes invalid data regions (such as the bubble interference layer near the water surface or the reverberation region near the bottom), retaining only the valid measurement profiles. Finally, the DSN outputs a structured Depth Index Vector (DIV). This vector is an ordered array, with each element storing the starting depth, ending depth, number of original flow velocity data points in the corresponding depth layer (DI), and their position pointers in the original data stream, providing an accurate spatial positioning framework for subsequent layering processing.

[0066] The segmentation process of DSN is dynamically adaptive. When special hydrological conditions (such as strong thermocline, salinity jump layer) are detected, DSN triggers an adaptive subdivision mechanism (ASM). ASM automatically reduces the local DCS (for example, temporarily switches from 1 meter to 0.25 meters) near the jump layer by analyzing the flow gradient and echo intensity mutation characteristics (for example, the flow rate difference between adjacent data points exceeds 0.2 meters / second or the echo intensity changes more than 5 decibels) in real time, generating higher resolution sub-layers. At the same time, in the area where the flow rate distribution is flat (gradient less than 0.05 meters / second / meter), DSN will start an intelligent merging mechanism (IMM) to merge multiple small size units into one large unit (such as merging 4 0.25 meter layers into 1 1 meter layer), reducing redundant calculations. This "well-distributed" segmentation strategy not only ensures high resolution in key areas, but also significantly improves overall processing efficiency. After segmentation, DIV not only contains static layering information, but also marks the location and reason of dynamic adjustment area (such as "DI=12.5 to DI=13.5 is the thermocline subdivision area"), providing context for subsequent analysis.

[0067] To ensure the robustness of segmentation, DSN integrates a boundary verification module (BVM). BVM uses echo intensity profile (EIP) to assist depth positioning. For example, when sound waves encounter a strongly reflective river bed or sea floor, EIP will show a steep rising edge. BVM compares the bottom boundary predicted by DIV (such as DI=49 corresponding to 49-50 meters) with the actual bottom boundary identified by EIP. If the deviation exceeds the preset threshold (such as 0.1 meters), the range of the bottommost DI is automatically adjusted (such as DI=49 is corrected to 48.5-49.5 meters), ensuring that each depth unit corresponds accurately to the actual physical layer. At the same time, BVM will eliminate "false boundary" signals caused by suspended matter shielding or instrument jitter. The final generated DIV is a reliable spatial segmentation blueprint that has been dynamically optimized and boundary calibrated.

[0068] Based on the depth index vector, the signal-to-noise ratio features of each layer echo signal are extracted, and the adaptive threshold is used to complete the outlier rejection; The system extracts all raw echo signals (Raw Echo Signal, RES) in each depth layer (DI) according to the guidance of DIV. For each DI, the signal-to-noise ratio feature (Signal-to-Noise Ratio Feature, SNRF) is calculated. SNRF includes three core indicators: Mean SNR (MSNR): The ratio of the average of all echo signal power to the average of noise power in this layer (unit: decibel dB).

[0069] SNR Standard Deviation (SDSNR): Reflects the fluctuation degree of SNR in the layer.

[0070] Valid Signal Proportion (VSP): The percentage of data points with SNR higher than a basic threshold (e.g., 3dB) in this layer.

[0071] For example, in the DI=10 layer (10-11 meters water depth), MSNR=15dB, SDSNR=1.2dB, and VSP=98%, indicating that the signal quality is excellent and stable.

[0072] Based on SNRF, the system starts the Adaptive Threshold-based Outlier Rejection (ATOR) process. The core of ATOR is a two-layer decision model: First layer: Global threshold filtering. Set the basic outlier criterion: if the Instant SNR (ISNR) of a data point is lower than the Absolute Noise Threshold (ANT, e.g., 0dB), or higher than the Saturation Distortion Threshold (SDT, e.g., 40dB), it is directly marked as an outlier and rejected.

[0073] Second layer: Local dynamic threshold filtering. For points that do not trigger the global threshold, calculate the deviation from the SNRF of the layer: If the difference between the ISNR of the point and the MSNR exceeds K × SDSNR (K is a dynamic coefficient, default 2.5, automatically increased to 3.5 when VSP<90%), it is considered a local outlier.

[0074] If the deviation of the velocity vector amplitude of the point from the average flow velocity in the layer exceeds 3 × the standard deviation of the velocity in the layer, and its ISNR is in the "low confidence interval" of [ANT+1dB, MSNR-3dB], it is marked as a suspected interference point.

[0075] For example, in the DI=25 layer (MSNR=8dB, SDSNR=2.5dB), a point with ISNR=2dB and velocity deviation of 0.8 meters / second (layer standard deviation is 0.2 meters / second) will be rejected. ATOR records the reason for rejecting each outlier (such as "low signal-to-noise ratio" or "velocity outlier") for quality reporting.

[0076] For special scenarios (e.g. high sediment-laden flow, dense bubble zone), ATOR introduces a Multi-feature Joint Decision Mechanism (MJDM). MJDM synchronously analyzes auxiliary features such as Spectral Width (SW), Autocorrelation Time (ACT), etc. For example: A point with ISNR = 6dB (higher than ANT), but its SW abnormally increases (>0.5m / s, normal layer average is 0.1m / s), and ACT significantly shortens (<10ms, normal is 20ms), indicating that there may be turbulent scattering or incoherent interference, even if the velocity is not significantly out of line.

[0077] For DI close to the bottom layer, additional check Echo Intensity Gradient (EIG). If a point EIG mutates (e.g. >3dB / 0.1m) and the velocity vector direction is abnormal (e.g. vertical velocity increases sharply), it may indicate bottom bed reflection interference, triggering removal.

[0078] After removal, the system calculates the Valid Data Retention Rate (VDRR) of each layer, and if VDRR < 70%, a data quality warning is issued.

[0079] For the layered signal after removal, the energy-weighted least squares method is used to fit the flow velocity distribution curve, and the unnormalized layered flow velocity is output. After removal of outliers, each depth layer (DI) retains a set of valid three-dimensional flow velocity vector points. The system uses the Energy-Weighted Least Squares (EWLS) method to fit the flow velocity distribution curve along the depth of the layer. The core idea of EWLS is to give higher weight to high-quality data points. The weight W_i is determined by two factors: SNR Weight (W_snr): proportional to the ISNR of the data point. For example, W_snr = ISNR / MSNR (normalized to the layer).

[0080] Temporal Correlation Weight (W_tc): calculated based on the velocity correlation of the point and its adjacent points in the time series. Points with high correlation (W_tc > 0.8) indicate stable measurement, and the weight is increased.

[0081] W_i = a x W_snr + b x W_tc (a + b = 1, default a = 0.7, b = 0.3). For example, a point with ISNR = 12 dB (MSNR = 10 dB) and W_tc = 0.9, its W_i = 0.7 x (12 / 10) + 0.3 x 0.9 = 1.14.

[0082] Using W_i weighting, curve fitting of three-dimensional velocity components is independently performed for each DI: Horizontal flow velocity fitting: taking depth z as independent variable, eastward velocity U(z) and northward velocity V(z) as dependent variables, respectively fitting quadratic polynomial: U(z) = a_u x z 2 + b_u x z + c_u; V(z) = a_v x z 2 + b_v x z + c_v; solving coefficients a_u, b_u, c_u, a_v, b_v, c_v by weighted least squares method, so that the residual sum of squares Σ[W_i x (U_i - U(z_i)) 2 ] is minimized.

[0083] Vertical flow velocity fitting: considering that the vertical velocity W(z) is usually small and changes gently, linear fitting is adopted: W(z) = k_w x z + b_w.

[0084] For example, in DI = 15 layer (15-16 meters), the fitting obtains U(z) = -0.002z 2 + 0.05z + 0.8 (unit: m / s), representing that the velocity of this layer weakly presents parabolic distribution with depth.

[0085] After fitting, the system outputs the unnormalized stratified velocity (Unnormalized Stratified Velocity, USV). USV contains three key outputs: Layer representative velocity (Layer Representative Velocity, LRV): taking the fitted velocity value (U, V, W) at the midpoint depth of the layer (such as DI = 15 corresponds to 15.5 meters). For example, LRV = (0.82, 0.15, -0.01) m / s.

[0086] Intra-layer velocity distribution function (Intra-layer Velocity Distribution Function, IVDF): that is, the expression and coefficients of U(z), V(z), W(z) functions obtained by fitting.

[0087] Goodness-of-Fit Metrics (GFM): including Weighted R 2 , WR 2 , Max Residual (MR), Mean Absolute Error (MAE). For example, WR 2 = 0.92 means the model explains 92% of the variation.

[0088] The velocity value at this time has not been compensated for depth attenuation, so it is called "unnormalized".

[0089] Combined with the transducer energy attenuation model, the depth compensation coefficient is calculated and the flow velocity amplitude is normalized to generate a layered flow velocity distribution model containing direction components.

[0090] The system calls the preset transducer energy attenuation model (Transducer Energy Attenuation Model, TEAM). TEAM is a semi-empirical model, and its inputs include: Basic Attenuation (BA): the attenuation per unit distance (dB / m) determined by the sound wave frequency (e.g. 300 kHz) and the water absorption coefficient (calculated from temperature and salinity).

[0091] Geometric Spreading Correction (GSC): energy attenuation inversely proportional to the square of the distance due to spherical wave spreading.

[0092] Beam Pattern Function (BPF): describes the radiation intensity distribution of the main lobe and side lobes of the sound beam.

[0093] The model output is the energy attenuation factor (Energy Attenuation Factor, EAF) at depth z: EAF(z) = 10^{[-BA×z - 20×log10(z) + BPF(θ_z)] / 10}, where θ_z is the sound beam incidence angle at depth z.

[0094] Use EAF to calculate the depth compensation coefficient (Depth Compensation Coefficient, DCC) for each depth layer (DI). DCC is defined as: DCC(DI) = 1 / √[EAF(z_mid)].

[0095] where z mid is the midpoint depth of the layer. The physical meaning of this formula is that the attenuation of the flow velocity amplitude (proportional to the square root of the echo intensity) is proportional to EAF, so the compensation needs to take its inverse. For example, DI = 20 layers (z mid = 20.5 meters), if EAF(20.5) = 0.25, then DCC = 1 / V0.25 = 2.0, indicating that the original flow velocity amplitude of this layer is underestimated by half, and needs to be enlarged by 2 times.

[0096] Velocity Amplitude Normalization (VAN) is performed: For the layer representative velocity LRV: multiply the original amplitude by DCC, and the direction remains unchanged. That is: U norm = U raw x DCC; V norm = V raw x DCC; W norm = W raw x DCC.

[0097] For the layer internal velocity distribution function IVDF: the amplitude coefficients (such as c u, c v, b w) in the function are scaled by DCC times synchronously. For example, if the original function is U(z) = 0.1z + 0.5 and DCC = 1.5, then after normalization it is U(z) = 0.15z + 0.75.

[0098] The final generated Stratified Velocity Distribution Model (SVDM) includes: normalized layer representative velocity (Normalized LRV); normalized layer internal velocity distribution function (Normalized IVDF); DCC value and compensation record of each layer; complete depth index vector DIV and signal-to-noise ratio characteristics.

[0099] This model completely retains the three-dimensional directional information (east / north / vertical), and the velocity amplitude has eliminated the depth-caused energy attenuation bias, which can be directly used for subsequent flow direction compensation and error analysis.

[0100] The depth learning network is used for intelligent stratification processing of the water profile, and the measurement unit is dynamically divided according to the preset vertical resolution requirement. The feature extraction network automatically identifies the echo signal quality of each layer, and uses the energy attenuation model to compensate the signal intensity difference at different depths, realizes the standardization processing of the full-profile flow velocity data, realizes the adaptive fine stratification of the water profile, and overcomes the measurement distortion problem of the fixed stratification mode in the jump layer water area. The normalization processing eliminates the influence of energy attenuation in the sound wave transmission process, so that the flow velocity data at different depths has direct comparability.

[0101] S204, calling the layered velocity distribution model, fusing the fluxgate compass azimuth data for flow direction compensation, triggering the deviation correction algorithm when the device inclination exceeds the limit, and outputting the absolute flow velocity distribution after pose calibration; Specifically, the depth unit data in the layered velocity distribution model can be called, and the three-axis azimuth angle of the fluxgate compass can be read synchronously to generate a flow direction reference vector. Depth unit data calling and processing: The system first extracts structured data from the constructed layered velocity distribution model (Layered Velocity Distribution Model, LVDM). The model divides the flow profile according to the preset depth unit size (e.g. 0.5 meters or 1 meter), and each unit contains the flow velocity amplitude (unit: meters / second) and direction component (unit: angle, 0-360°). For example, depth unit D5 (corresponding to water depth 10-10.5 meters) stores a flow velocity vector of 0.3 meters / second and a direction angle of 120°. The central processing unit (Central Processing Unit, CPU) traverses all depth unit indexes (Depth Index Vector, DIV), reads the flow velocity direction data of each unit in order, and forms an initial direction data set (Initial Direction Dataset, IDD).

[0102] Synchronous acquisition of fluxgate compass data: At the same time, the built-in fluxgate compass (Fluxgate Compass, FC) in the device outputs real-time three-axis azimuth angle data: Heading Angle (HA): the angle between the device longitudinal axis and magnetic north (range 0-360°), reflecting the device orientation; Pitch Angle (PA): the front-to-back tilt angle of the device (range ±90°), positive value indicating forward tilt; Roll Angle (RA): the left-to-right tilt angle of the device (range ±90°), positive value indicating right tilt.

[0103] The data is transmitted to the main control board at a frequency of 100 Hertz (Hz) through the RS485 serial port protocol. To ensure time synchronization, the system uses a hardware trigger signal (Hardware Trigger Signal, HTS) to align the timestamps (Timestamp, TS) of the flow velocity data and the compass data, with a time error controlled within 10 milliseconds (ms).

[0104] Generation of flow direction reference vector: Based on the current heading angle HA (e.g. HA = 45°), the system constructs a Flow Reference Vector (FRV). This vector, with 0° reference to magnetic north, converts the flow direction in the device coordinate system to the absolute flow direction in the geodetic coordinate system. For example, if a certain flow direction is measured as 120° (relative to the device) when the device heading is HA = 45°, the absolute flow direction is HA + 120° = 165° (relative to magnetic north). The Initial Direction Data (IDD) of all depth cells are converted according to this rule, and the standardized Flow Reference Vector FRV is generated, with data structure [depth cell number, absolute flow direction angle].

[0105] The flow direction compensation is performed by combining the Flow Reference Vector and the layered flow direction component, and the relative flow velocity distribution is output through spherical coordinate conversion; Necessity of flow direction compensation: Due to the possible swing of the device in the water or the deflection caused by the cable, the flow direction compensated by the compass heading angle HA alone has systematic errors. For example, when the device roll angle RA reaches 20°, the transducer beam pointing deviation will cause the measured direction angle to be distorted. Flow direction compensation (FDC) aims to eliminate the influence of the device pose on the direction measurement.

[0106] Spherical coordinate conversion algorithm execution: The system converts the flow velocity vector of each depth cell from the Cartesian coordinate system (Cartesian Coordinates, including eastward U, northward V, and vertical W components) to the spherical coordinate system (Spherical Coordinates): Horizontal Magnitude (HM): calculated as √(U 2 +V 2 ).

[0107] Horizontal Direction Angle (HDA): calculated as arctan(U / V), and corrected according to the quadrant (e.g. +180° when V is negative).

[0108] The HDA is calibrated in combination with the absolute flow direction angle in the Flow Reference Vector FRV: If the FRV indicates that the theoretical flow direction of a certain depth cell should be 165°, but the measured HDA is 155°, a compensation offset value (Offset Value, OV) of +10° is applied.

[0109] This process traverses all depth cells to ensure that the horizontal flow velocity direction is aligned with the geodetic coordinate system.

[0110] Relative Velocity Distribution (RVD) generation: The compensated data is reconstructed into a Relative Velocity Distribution (RVD). This distribution preserves the magnitude of the flow velocity (HM) and the calibrated heading angle for each depth cell, and adds a vertical component W (unit: m / s). For example, the output data for depth cell D5 is [HM=0.3 m / s, HDA=165°, W=0.02 m / s]. The RVD has eliminated the device orientation error, but the beam geometry distortion caused by device tilt is still not resolved.

[0111] Real-time monitor the tilt sensor data, when the device tilt angle exceeds the preset threshold, trigger the quaternion rotation correction, generate a rotation correction matrix; Tilt threshold monitoring mechanism: The system continuously reads the pitch angle PA and roll angle RA (from the fluxgate compass). The preset tilt threshold (TiltThreshold, TT) is set according to the device model (for example, TT=15°). When the absolute value of PA or RA exceeds TT (such as PA=20°), the correction algorithm (Tilt Correction Algorithm, TCA) is triggered. When not exceeding the limit, the RVD is directly output to the next link.

[0112] Quaternion rotation correction principle: Quaternion (Q) is a four-dimensional complex number (structure: [w, x, y, z]) used to efficiently describe three-dimensional rotation. The system calculates the rotation relationship from the device coordinate system to the geodetic coordinate system according to the current PA and RA: Synthesize the tilt vector (Tilt Vector, TV) from PA and RA, for example, TV=[PA=20°, RA=10°].

[0113] Convert TV to Quaternion Q (call open source math library such as Eigen when implemented).

[0114] For example, the Q corresponding to the tilt angle TV may be [0.96, 0.17, 0.08, 0.01], which physically represents a rotation of about 22.3° around a specific axis.

[0115] Rotation correction matrix construction: Convert the quaternion Q to a 3x3 rotation correction matrix (Rotation Correction Matrix, RCM). This matrix can be directly applied to the flow velocity vector: For example, a certain behavior of RCM is [0.94, -0.03, 0.34], which represents the new projection direction of the original X-axis (device east direction) in the geodetic coordinate system.

[0116] The matrix is stored in cache for all depth units to call uniformly, ensuring the calculation efficiency.

[0117] Apply the rotation correction matrix to the relative velocity distribution to convert the device coordinate system to the earth coordinate system; eliminate the Coriolis force deviation caused by the earth's rotation through coordinate projection, and output the absolute velocity distribution of pose calibration.

[0118] Coordinate system conversion execution: Multiply the rotation correction matrix RCM by each flow velocity vector (including U, V, and W components) in RVD: Input vector: [U_device, V_device, W_device] in the device coordinate system; Output vector: [U_earth, V_earth, W_earth] in the earth coordinate system.

[0119] For example, [U=0.2, V=0.1, W=0.02] in the original device coordinate system may output [U_earth=0.18, V_earth=0.12, W_earth=0.02] after RCM transformation. This step corrects the beam pointing deviation caused by device tilt.

[0120] Coriolis force compensation mechanism: The Coriolis force generated by the earth's rotation will interfere with the measurement of horizontal flow velocity (water flow in the northern hemisphere will deflect to the right). The system dynamically calculates the Coriolis parameter (CP) based on the device latitude (Latitude, LAT, from the GPS module): CP = 2 × Earth rotation angular velocity (7.292 × 10 -5 radians / second) × sin(LAT). For example, when the latitude is 30° north, CP ≈ 7.292 × 10 -5 .

[0121] Apply reverse compensation to the converted horizontal flow velocity [U_earth, V_earth]: east component U_corrected = U_earth + (CP × measurement integration time × V_earth); north component V_corrected = V_earth - (CP × measurement integration time × U_earth).

[0122] Absolute velocity distribution output: Finally, generate the absolute velocity distribution of pose calibration (Pose-Calibrated Absolute Velocity Distribution, PAVD): Data structure: [Depth Cell, Eastward Velocity U_corrected, Northward Velocity V_corrected, Vertical Velocity W_earth]. Unit is meter per second, direction is implied by component sign (U positive = eastward, V positive = northward).

[0123] For example, depth cell D5 outputs [-0.15, 0.25, -0.01], which means 0.15 m / s westward, 0.25 m / s northward, and 0.01 m / s downward. This result has fused the compass azimuth compensation, tilt correction, and Coriolis force correction, and can be directly used for hydrological analysis.

[0124] The system integrates the azimuth reference provided by the high-precision fluxgate compass, converts the relative flow velocity into absolute flow velocity in the geographic coordinate system through spherical coordinate conversion. When the device tilt exceeds the safety threshold, the quaternion rotation algorithm is automatically started for attitude compensation, eliminating the measurement error caused by the installation angle, solving the measurement deviation problem caused by the change of the underwater device due to ocean current impact, and improving the flow direction measurement accuracy. The geodetic coordinate system output can be directly used for hydrological model construction without subsequent data conversion.

[0125] S205, according to the absolute flow velocity distribution and the echo intensity dynamic range constraint, calculating the error interval of each depth cell, outputting the final flow velocity profile with confidence identification.

[0126] Specifically, according to the data of each depth cell in the absolute flow velocity distribution, the echo intensity dynamic range feature can be extracted to generate a signal-to-noise ratio attenuation curve; The system first parses the absolute flow velocity distribution (AVD) data, which takes depth cell (DC) as the basic unit, and each unit contains flow velocity value (VV) and three-dimensional direction information. For each depth cell, its corresponding echo intensity (EI) original data is extracted synchronously. The extraction of echo intensity dynamic range (DR) is based on the propagation characteristics of sound waves in water: the energy of sound waves decays exponentially with depth. The system calculates the ratio of echo intensity to reference intensity (usually the echo intensity value near the transducer surface) for each depth cell, generating an intensity attenuation rate (IAR). For example, at depth cell index DC=10 (corresponding to a water depth of 5 meters), if the echo intensity is -60 decibels (dB), and the reference intensity is -40 decibels, then the IAR is -20 decibels. By traversing all depth cells (such as from DC=1 to DC=100, covering a water depth of 0-50 meters), the system draws a scatter plot of IAR versus depth.

[0127] To quantify the signal quality decay trend, the system adopts a Sliding Window Polynomial Fitting Algorithm (SWPFA). Taking every 5 adjacent depth cells as a window (e.g. DC1-5, DC2-6), a 2nd order polynomial is fitted to the IAR values within each window, generating a local fitting curve. The window slides by one depth cell each time, and finally a smooth, full-profile covering Signal-to-Noise Ratio Attenuation Curve (SNAC) is obtained. Key features of this curve include: Initial Attenuation Slope (IAS): Reflects the near-field signal decay rate, influenced by the water body suspended particle concentration; Inflection Depth (ID): Critical point where the curve slope changes significantly, usually corresponding to the thermocline or high turbidity layer; Limit SNR (LSNR): Stable value at the end of the curve, representing the lowest signal-to-noise ratio (e.g. -15 decibels) that the system can detect.

[0128] For example, in an estuary area, the SNAC may have an inflection point at a depth of 10 meters, with an initial slope of -3 decibels per meter and a limit signal-to-noise ratio of -12 decibels.

[0129] The generation of the signal-to-noise ratio attenuation curve needs to exclude interference factors. The system synchronously accesses the real-time data (unit: turbidity unit NTU) of the water body turbidity sensor (Turbidity Sensor, TS) and the sound velocity gradient data of the sound velocity profiler (Sound Velocity Profiler, SVP). If the turbidity of a certain depth unit suddenly changes (e.g. from 5 NTU to 50 NTU) and the abnormal fluctuation of SNAC (e.g. the decay rate increases by 30%) matches in time and space, mark this area as "strong scattering interference area", its data does not participate in curve fitting. The final generated SNAC serves as the input reference for subsequent error calculation, which is mathematically expressed as a function form: SNAC(z)=a⋅z 2 +b⋅z+c. Where z is the depth (unit: meters), coefficients a, b, c are determined by fitting, and their physical meanings correspond to the attenuation acceleration, the initial attenuation rate and the background noise level respectively.

[0130] Based on the signal-to-noise ratio attenuation curve, the flow velocity measurement variance is calculated by the Cramer-Rao lower bound theory, and the initial error interval is output; After the system generates the signal-to-noise ratio attenuation curve (SAC), the next goal is to estimate the statistical uncertainty, i.e. measurement error, for the flow velocity measurement value of each depth unit in the absolute flow velocity distribution. The core principle of flow velocity measurement is Doppler frequency shift, and the accuracy of frequency shift estimation is strictly limited by signal quality. Here, a theoretical tool is introduced: Cramér-Rao Lower Bound (CRLB). CRLB is a statistical concept that theoretically defines the lower limit of the variance that can be achieved for an unbiased estimate of a parameter (here, the Doppler frequency shift, and then the flow velocity) under given signal model and noise conditions. That is, the variance of any actual estimation method cannot be lower than CRLB. Therefore, CRLB is the best theoretical benchmark for evaluating the potential accuracy of flow velocity measurement.

[0131] The system calculates the flow velocity measurement variance based on the signal-to-noise ratio attenuation curve (SAC) as follows: for each depth unit in the absolute flow velocity distribution, the system reads the signal-to-noise ratio SNR value corresponding to the unit (directly from the SAC curve). Then, the system applies the Cramér-Rao Lower Bound (CRLB) formula corresponding to the acoustic Doppler current profiler (ADCP) signal processing algorithm to calculate the theoretical minimum variance of Doppler frequency shift estimation (Variance_fd_min) for the depth unit. The specific form of this CRLB formula depends on the signal type used by the ADCP (such as pulse type, coding method), signal processing algorithm (such as FFT point number, window function type), and the most important parameter - the signal-to-noise ratio SNR of the unit. Generally speaking, CRLB is inversely proportional to SNR, the higher the SNR, the smaller the CRLB (i.e. more accurate frequency shift estimation); the lower the SNR, the larger the CRLB (less accurate estimation). After calculating the CRLB of Doppler frequency shift (Variance_fd_min), it needs to be converted to the variance of flow velocity (Variance_V). There is a fixed conversion coefficient K between flow velocity V and Doppler frequency shift fd (usually called Doppler scaling factor, K = C / (2 * F0 * cosθ), where C is the sound speed, F0 is the transmission frequency, and θ is the beam angle). Therefore, the variance of flow velocity Variance_V_min is theoretically equal to (K 2 )*Variance_fd_min. This Variance_V_min is the theoretical minimum variance of flow velocity measurement for the depth unit (determined by CRLB).

[0132] After obtaining the theoretical minimum variance Variance_V_min of each depth cell flow rate, the system derives the initial error interval (IEI) of the flow rate value of the cell based on this. The error interval is usually expressed in the form of a confidence interval, and the most common is the 95% confidence interval. Assuming that the flow rate estimate value obeys Gaussian (normal) distribution (which is usually a reasonable approximation in signal processing), then the 95% confidence interval means that the true flow rate value has a 95% probability of falling within this interval. The relationship between the half-width (HW) of the interval and the standard deviation (SD) is: HW = 1.96 * SD. And the standard deviation SD is the square root of the variance, that is, SD_min = sqrt(Variance_V_min). Therefore, for each depth cell, the initial 95% confidence error interval of its flow rate components (U, V, W) can be expressed as: Measurement U ± (1.96 * sqrt(Variance_U_min)); Measurement V ± (1.96 * sqrt(Variance_V_min)); Measurement W ± (1.96 * sqrt(Variance_W_min)).

[0133] The system performs the above calculation for each flow rate component of each depth cell, and outputs a set of preliminary initial error intervals (IEI) based on the theoretical optimal (CRLB) and the current signal quality (SNR). This IEI reflects the best accuracy estimate that can be achieved under ideal conditions, considering only the signal noise level.

[0134] Fusion of acoustic scattering intensity and water turbidity data, construct a multi-physical field coupling error correction model, generate a correction error coefficient; The system recognizes that the initial error interval (IEI) calculated based on the signal-to-noise ratio attenuation curve (SAC) and the Cramer-Rao lower bound (CRLB) alone can underestimate the uncertainty in actual measurements. This is because the actual water environment is complex, and there are many factors that can affect the accuracy of flow velocity measurement, which are not fully covered by the CRLB model. In order to obtain a more realistic and reliable error estimate, the system needs to introduce more environmental information for error correction. Here, two key physical quantities are particularly emphasized: the acoustic scattering intensity (Scattering Intensity, SI) itself (i.e. echo intensity) and the turbidity of the water body (Turbidity, TUR). Although the echo intensity EI (or mean intensity MEI) has been used to calculate the SNR, its absolute level and distribution characteristics also contain information. Turbidity TUR is measured in real time by an integrated optical turbidity sensor (e.g. using 90-degree scattered light or transmission light principle), which reflects the concentration of suspended particles in the water body.

[0135] The system performs the operation of fusing acoustic scattering intensity and water turbidity data. For each depth unit, in addition to the existing mean echo intensity MEI and signal-to-noise ratio SNR, the system also reads the turbidity measurement value TUR corresponding to the unit (if the turbidity sensor is point type, it is assigned to the depth unit by spatial interpolation or nearest neighbor method). The system analyzes the relationship between these physical quantities: Scattering intensity MEI and turbidity TUR: Ideally, echo intensity MEI should be positively correlated with turbidity TUR, because more particles mean stronger scattering. In practice, particle types (such as mineral particles, organic organisms), particle size distribution, acoustic characteristics (such as density, compressibility) will affect this relationship. The system establishes an MEI-TUR empirical relationship through historical data or physical models.

[0136] Signal-to-noise ratio SNR and turbidity TUR: High turbidity usually means stronger scattering signals (which is beneficial to improve SNR), but too high turbidity can cause excessive attenuation or multiple scattering of acoustic waves (which may otherwise reduce the SNR in deep water).

[0137] Flow velocity variance and scattering characteristics: The motion characteristics of scatterers (such as settling velocity, consistency of motion with the flow) will affect the broadening of the Doppler spectrum. For example, a large number of slowly settling fine particles can cause the spectrum to widen, making the variance of the flow velocity estimate increase, even if the SNR may not be low. MEI and TUR can help infer the type and state of the scatterers.

[0138] Based on the above analysis, a Multi-physics Coupled Error Correction Model (MCECM) is constructed. This model is an empirical or semi-empirical mathematical relationship (e.g., using machine learning models such as Support Vector Machine SVR or Random Forest RF), whose input features (IF) usually include: the initial flow velocity theoretical standard deviation SD_min (sqrt(Variance_V_min)) of the depth cell; the mean echo intensity MEI of the depth cell; the echo intensity standard deviation SDEI (reflecting the scattering uniformity) of the depth cell; the signal-to-noise ratio SNR of the depth cell; the water turbidity TUR of the depth cell; possibly also the depth D itself, the flow velocity value V (as high-speed flow can cause additional spectral broadening), etc.

[0139] The output of the model is the correction error coefficient (CEC). CEC is usually a coefficient greater than or equal to 1. The training goal of the model is to learn the proportional relationship between the actual flow velocity measurement error (the standard deviation of the difference between the actual flow velocity and the reference flow velocity) and the theoretical minimum standard deviation SD_min calculated only by CRLB+SNR, using a large amount of ADCP measured data collected under known flow velocities (such as laboratory calibration flume or placed reference flow velocity instrument) and different environmental conditions (different turbidity, different scatterers). After training, for new measurement data, the model inputs the above features IF, and outputs a correction coefficient CEC. This CEC represents the amplification multiple of the actual flow velocity standard deviation caused by actual environmental factors (such as scatterer type, spectral broadening effect) relative to the theoretical minimum standard deviation SD_min. That is, the actual estimated standard deviation SD_actual ≈ CEC * SD_min. After generating the correction error coefficient CEC, a more accurate precision evaluation basis that conforms to the actual environmental impact is provided for each flow velocity component of each depth cell.

[0140] Apply the correction error coefficient to the initial error interval, and use Monte Carlo simulation to verify the confidence probability; After the system obtains the corrected error coefficient (CEC) for each depth unit and each velocity component, it needs to be applied to the previously calculated initial error interval (IEI) to obtain a more realistic corrected error interval (AEI). As mentioned earlier, the half-width of the initial error interval HW_initial is calculated based on the theoretical minimum standard deviation SD_min (for example, for a 95% confidence interval, HW_initial = 1.96 * SD_min). The actual corrected standard deviation is estimated to be SD_actual ≈ CEC * SD_min. Therefore, the half-width of the corrected error interval HW_adjusted = 1.96 * SD_actual ≈ 1.96 * (CEC * SD_min) = CEC * HW_initial. In other words, the corrected half-width can be obtained by simply multiplying the half-width of the initial error interval by the corrected error coefficient CEC. The corrected 95% confidence interval becomes: Measured value U ± (CEC_U * HW_initial_U); Measured value V ± (CEC_V * HW_initial_V); Measured value W ± (CEC_W * HW_initial_W).

[0141] The system performs this multiplication operation on all velocity components for all depth cells and generates a set of adjusted error intervals (AEIs).

[0142] Although the MCECM model is trained based on data, to more rigorously evaluate the reliability of the confidence level claimed for the AEI (e.g., whether it truly achieves 95% coverage) under complex real-world conditions, the system performs Monte Carlo simulation (MCS). Monte Carlo simulation is a statistical method based on random sampling. The system creates a simplified probabilistic model for the currently measured hydrological environment (primarily based on the measured distribution characteristics of MEI, SNR, TUR, and SDEI). This model simulates: Signal generation: Simulates and generates an echo signal with statistical characteristics similar to the current measured data (such as average intensity, variance, and spectral shape), and superimposes random noise that meets the current SNR level.

[0143] Scattering effect: Based on the turbidity TUR and echo intensity characteristics MEI / SDEI, the contribution of different types of scatterers (such as particles of different sizes and velocities) to the Doppler spectrum broadening is simulated.

[0144] Flow velocity processing: Apply the same signal processing flow (e.g. FFT, beam decoupling) to the simulated signal as the actual device to estimate the simulated flow velocity value.

[0145] Error calculation: Since the "true" flow velocity is set in the simulation (as the ground truth), the error of the flow velocity estimation in each simulation can be calculated.

[0146] The system will perform thousands (e.g. 10000) of such Monte Carlo simulations. In each simulation, the same environmental parameters (SNR, MEI, TUR, etc.) are used but different random noise and scatterer distribution samples are used. For each depth cell, each flow velocity component, the system records the estimation error in each simulation. After the simulation is completed, the system analyzes these error samples: Calculate actual standard deviation: Calculate the standard deviation of the error sample set SD_simulated directly, which can be compared with the previous SD_actual (≈ CEC * SD_min) to verify the reasonableness of the correction coefficient.

[0147] Verify confidence interval: Check how many percentage of the simulation results, whose estimated flow velocity value falls within the corrected error interval AEI (i.e. measurement value ± HW_adjusted) calculated based on the parameters of this simulation. This proportion is the simulated actual confidence probability (e.g. simulated confidence probability SCP).

[0148] The system compares the simulated confidence probability SCP with the target confidence probability (usually 95%). If SCP is very close to 95% (e.g. between 94%-96%), it is considered that the correction model and error interval are reliable. If SCP is significantly lower than 95%, it means that AEI is too optimistic, and the system may need to automatically adjust CEC or adjust the model; if SCP is higher than 95%, it means that AEI may be too conservative, but it is usually acceptable. This process verifies the effectiveness of the confidence probability, ensuring that the final confidence label has a solid statistical basis. The verification result can also be used to feedback to optimize the correction model MCECM.

[0149] According to the pre-set confidence level, label the confidence of the flow velocity value of each depth cell, and output the final flow velocity profile.

[0150] The half-width of the adjusted error interval (AEI) HW_adjusted, which is verified by Monte Carlo simulation (or directly used after confirming the reliability of the model), is a quantitative expression of the measurement accuracy of each depth unit and each flow velocity component (U, V, W). In order to intuitively show the quality and reliability of the measurement results to the user, the system needs to evaluate and classify these accuracies according to the preset standards. The user or system administrator will usually set one or more confidence levels (Confidence Level, CL) in advance. These levels are divided based on the relative size (usually the relative error percentage RE% = (HW_adjusted / |V|) * 100% obtained by dividing the half-width of the error interval HW_adjusted by the absolute value of the flow velocity measurement value V) or the absolute size (HW_adjusted itself) of the flow velocity measurement error. Common confidence levels may have three or four levels, for example: High Confidence (HC): relative error RE% < 5% and HW_adjusted < 0.05 m / s (for example, suitable for precise scientific research or engineering monitoring); Medium Confidence (MC): 5% ≤ RE% < 10% or 0.05 m / s ≤ HW_adjusted < 0.10 m / s (for example, suitable for general hydrological investigation); Low Confidence (LC): 10% ≤ RE% < 20% or 0.10 m / s ≤ HW_adjusted < 0.20 m / s (for example, the result is only for reference and needs to be used with caution); Unreliable (UR): RE% ≥ 20% or HW_adjusted ≥ 0.20 m / s or SNR is lower than a certain absolute threshold (such as SNR < 3 dB) (for example, the data quality is poor and is recommended to be discarded).

[0151] The specific threshold division can be pre-set through the system configuration interface according to the accuracy requirements of the application scenario.

[0152] The system evaluates and labels each flow velocity component (U, V, W) of each depth unit in the absolute flow velocity distribution independently according to the division standard of the pre-set confidence level (CL). The specific process is: Read the current measurement value V (representing U, V or W) of the component; Read the half-width of the adjusted error interval HW_adjusted corresponding to the component; Calculate the relative error RE% = (HW_adjusted / |V|) * 100% (if V is 0 or close to 0, use the absolute error HW_adjusted directly); Compare the calculated RE% and HW_adjusted values with the preset confidence level thresholds.

[0153] According to the comparison results, classify the measurement results of the flow velocity components into corresponding confidence levels (such as HC, MC, LC, UR); Assign a corresponding confidence indicator (CI) to the flow velocity component. This indicator can be a numerical code (such as 3=HC, 2=MC, 1=LC, 0=UR), a letter code (such as 'H', 'M', 'L', 'U') or a more intuitive color coding (such as green=HC, yellow=MC, orange=LC, red=UR). At the same time, the corrected error interval value (such as ± value) of the component is also retained.

[0154] After completing the confidence evaluation and identification labeling of all flow velocity components of all depth units, the system integrates all information and outputs the final flow velocity profile (FVP). This final profile data includes: Depth coordinates: the center depth or range of each depth unit; flow velocity data: the measured values of eastward flow velocity U, northward flow velocity V and vertical flow velocity W of each depth unit; error estimation: the corresponding corrected error interval half-width value (such as U_err, V_err, W_err) of each flow velocity component; confidence indicator: the corresponding confidence indicator (CI_U, CI_V, CI_W) of each flow velocity component; auxiliary data (optional): such as the signal-to-noise ratio SNR, echo intensity MEI and turbidity TUR of the unit.

[0155] The final flow velocity profile FVP can be output in standard data formats (such as NetCDF, ASCII table) or displayed through the device's graphical user interface (GUI) or data visualization software. In graphical display, the flow velocity curve is usually drawn in different colors according to the confidence indicator, and the detailed error interval value and confidence level are displayed when the mouse hovers over, or a color bar of the confidence indicator is drawn beside the profile. The final flow velocity profile with confidence indicator provides users with clear, intuitive and reliable information on the vertical structure of water flow velocity, and clearly indicates the advantages and disadvantages of data quality at different positions, greatly improving the interpretability and practical value of measurement results, and supporting users to make more informed decisions (such as which data can be directly used for modeling, which data needs to be treated with caution or discarded).

[0156] Based on the acoustic scattering theory and signal processing principle, the signal-to-noise ratio, turbidity and other parameters of each depth unit are comprehensively analyzed, the limit is measured through the Cramer-Rao lower bound calculation theory, the data fluctuation under different environmental interference is simulated by using the Monte Carlo method, the error interval with statistical significance is generated, and the confidence level is marked for each data point, and the quantitative reliability evaluation of the flow rate measurement result is realized for the first time, and the accuracy basis is provided for the subsequent data application. The confidence level identification can assist the user to identify the reliable data interval, and avoid the decision-making error caused by the low-quality data.

[0157] It can be seen that according to the depth characteristics and accuracy requirements of the target water area, the transmission parameter combination suitable for the current hydrological environment is generated; the original echo signal is collected based on the transmission parameter combination, and the temperature-compensated three-dimensional flow rate vector is output; the three-dimensional flow rate vector is input into the dynamic segmentation network to generate a layered flow rate distribution model; the layered flow rate distribution model is called, the flow direction compensation is performed by fusing the magnetic flux gate compass azimuth data, the correction algorithm is triggered when the device inclination exceeds the limit, and the absolute flow rate distribution after pose calibration is output; according to the absolute flow rate distribution and the echo intensity dynamic range constraint, the error interval of each depth unit is calculated, and the final flow rate profile with confidence level identification is output, so that the environmental adaptability and data reliability of the flow rate measurement can be improved.

[0158] Another embodiment of the application provides an acoustic Doppler flow rate measurement system, referring to Figure 3 , the system can include: The activation module 301 is used for automatically activating the preset high-resolution broadband mode or large-range narrowband mode according to the depth characteristics and accuracy requirements of the target water area, generating a transmission parameter combination suitable for the current hydrological environment; The correction module 302 is used for collecting the original echo signal based on the transmission parameter combination, combining the real-time temperature data of the transducer, correcting the Doppler frequency shift through the pre-trained compensation model, and outputting the temperature-compensated three-dimensional flow rate vector; The processing module 303 is used for inputting the three-dimensional flow rate vector into the dynamic segmentation network, segmenting the profile layer according to the preset depth unit size, completing the echo intensity normalization processing, and generating a layered flow rate distribution model; The compensation module 304 is used for calling the layered flow rate distribution model, fusing the magnetic flux gate compass azimuth data to perform flow direction compensation, triggering the correction algorithm when the device inclination exceeds the limit, and outputting the absolute flow rate distribution after pose calibration; The output module 305 is used for calculating the error interval of each depth unit according to the absolute flow rate distribution and the echo intensity dynamic range constraint, and outputting the final flow rate profile with confidence level identification.

[0159] The embodiment of the present application also provides a storage medium, and the storage medium stores a computer program, wherein the computer program is configured to execute the steps in any of the method embodiments.

[0160] Specifically, in the embodiment, the storage medium can be configured to store a computer program for executing the following steps: S201, automatically activating a preset high-resolution wideband mode or a large-range narrowband mode according to depth characteristics of a target water area and precision requirements, and generating a transmission parameter combination adaptive to a current hydrological environment; S202, collecting original echo signals based on the transmission parameter combination, combining real-time temperature data of a transducer, correcting Doppler frequency shift through a pre-trained compensation model, and outputting a temperature-compensated three-dimensional flow velocity vector; S203, inputting the three-dimensional flow velocity vector into a dynamic segmentation network, segmenting a profile layer according to a preset depth unit size, completing echo intensity normalization processing, and generating a layered flow velocity distribution model; S204, calling the layered flow velocity distribution model, fusing magnetic fluxgate compass azimuth data for flow direction compensation, triggering a deviation correction algorithm when device inclination exceeds a limit, and outputting an absolute flow velocity distribution after pose calibration; S205, calculating error intervals of each depth unit according to the absolute flow velocity distribution and echo intensity dynamic range constraints, and outputting a final flow velocity profile with a confidence identifier.

[0161] The embodiment of the present application also provides an electronic device, including a memory and a processor, the memory stores a computer program, and the processor is configured to execute the computer program to execute the steps in any of the method embodiments.

[0162] Specifically, the electronic device can further include a transmission device and an input-output device, wherein the transmission device is connected with the processor, and the input-output device is connected with the processor.

[0163] Specifically, in the embodiment, the processor can be configured to execute the following steps through the computer program: S201, automatically activating a preset high-resolution wideband mode or a large-range narrowband mode according to depth characteristics of a target water area and precision requirements, and generating a transmission parameter combination adaptive to a current hydrological environment; S202, collecting original echo signals based on the transmission parameter combination, combining real-time temperature data of a transducer, correcting Doppler frequency shift through a pre-trained compensation model, and outputting a temperature-compensated three-dimensional flow velocity vector; S203, inputting the three-dimensional flow velocity vector into a dynamic segmentation network, segmenting a profile layer according to a preset depth unit size, completing echo intensity normalization processing, and generating a layered flow velocity distribution model; S204, calling the layered flow rate distribution model, fusing the fluxgate compass azimuth data for flow direction compensation, triggering the deviation correction algorithm when the device inclination exceeds the limit, and outputting the absolute flow rate distribution after pose calibration; S205, calculating the error interval of each depth unit according to the absolute flow rate distribution and the echo intensity dynamic range constraint, and outputting the final flow rate profile with confidence identification.

[0164] The above describes the structure, features and effects of the present application in detail according to the embodiments shown in the drawings. The above description is only the preferred embodiments of the present application, but the present application is not limited to the embodiments shown in the drawings. Any changes or modifications made in accordance with the concept of the present application, or equivalent embodiments with equivalent changes, are still within the scope of protection of the present application as long as they do not exceed the spirit of the specification and drawings.

Claims

1. An acoustic Doppler flow velocity measurement method, characterized in that: The method comprises: Automatically activates the preset high-resolution broadband mode or wide-range narrowband mode based on the depth characteristics and accuracy requirements of the target waters, generating a transmission parameter combination that is suitable for the current hydrological environment; The original echo signal is collected based on the transmission parameter combination, combined with the real-time temperature data of the transducer, the Doppler frequency shift is corrected through a pre-trained compensation model, and a temperature-compensated three-dimensional flow velocity vector is output; Inputting the three-dimensional velocity vector into a dynamic segmentation network, segmenting the profile layer according to a preset depth unit size, completing echo intensity normalization processing, and generating a layered velocity distribution model; The layered flow velocity distribution model is called, and the fluxgate compass azimuth data is integrated to perform flow direction compensation. When the device tilt exceeds the limit, the correction algorithm is triggered, and the absolute flow velocity distribution of the posture calibration is output; According to the absolute velocity distribution and the echo intensity dynamic range constraint, the error interval of each depth unit is calculated, and the final velocity profile with a confidence mark is output.

2. The method according to claim 1, characterized in that According to the depth characteristics and accuracy requirements of the target waters, the preset high-resolution broadband mode or large-range narrowband mode is automatically activated to generate a combination of transmission parameters adapted to the current hydrological environment, including: Based on the sonar depth detection data of the target waters and the preset accuracy threshold, the depth-accuracy feature matrix is ​​calculated and the mode selection decision factor is output; Based on the decision factor, the frequency configuration library of high-resolution broadband mode or large-range narrowband mode is activated, and the frequency parameter set including center frequency, bandwidth and pulse width is extracted; Combined with the current water salinity sensor data, the pulse repetition period parameters are dynamically adjusted through sound velocity compensation to generate a basic transmission parameter set; According to the basic transmission parameter set and frequency parameter set, the combination of sound beam opening angle and transmission power is optimized, and the final transmission parameter combination is output.

3. The method according to claim 2, characterized in that The method comprises: collecting the original echo signal based on the transmission parameter combination, combining the real-time temperature data of the transducer, correcting the Doppler frequency shift through a pre-trained compensation model, and outputting a temperature-compensated three-dimensional flow velocity vector, including: According to the combination of transmission parameters, the transducer array is driven to transmit the frequency modulated pulse signal and the IQ quadrature components of the original echo signal are collected; Synchronously acquire real-time data from the transducer temperature sensor, input it into a pre-trained frequency shift-temperature mapping model, and output the temperature drift compensation coefficient; The IQ components are demodulated into complex signals, and the Doppler frequency shift spectrum is calculated by window function weighted FFT to generate the initial frequency shift matrix; Apply the temperature drift compensation coefficient to the initial frequency shift matrix and use the three-beam spatial decoupling algorithm to decompose the XYZ axial components; The decomposed axial component is subjected to motion interference filtering, and a temperature-compensated three-dimensional velocity vector containing a directional component is output.

4. The method according to claim 3, characterized in that The three-dimensional velocity vector is input into a dynamic segmentation network, the cross-section layer is segmented according to a preset depth unit size, and the echo intensity normalization processing is completed to generate a layered velocity distribution model, including: The three-dimensional velocity vector is input into the dynamic segmentation network, and the profile layer is divided according to the preset depth unit size to generate the depth index vector; Based on the depth index vector, the signal-to-noise ratio characteristics of the echo signals of each layer are extracted, and outliers are eliminated through adaptive thresholding; For the removed stratified signals, the energy-weighted least squares method is used to fit the velocity distribution curve and output the unnormalized stratified velocity. Combined with the transducer emission energy attenuation model, the depth compensation coefficient is calculated and the velocity amplitude is normalized to generate a stratified velocity distribution model with directional components.

5. The method according to claim 4, characterized in that The layered velocity distribution model is called, and the fluxgate compass azimuth data is integrated to perform flow direction compensation. When the device tilt exceeds the limit, the correction algorithm is triggered, and the absolute velocity distribution of the posture calibration is output, including: Call the depth unit data in the layered velocity distribution model, synchronously read the three-axis azimuth of the fluxgate compass, and generate the flow direction reference vector; Combining the flow direction reference vector with the stratified flow velocity direction component, the flow direction compensation is completed through spherical coordinate transformation, and the relative flow velocity distribution is output; Real-time monitoring of tilt sensor data. When the device's tilt angle exceeds a preset threshold, quaternion rotation correction is triggered to generate a rotation correction matrix. Apply a rotation correction matrix to the relative velocity distribution to transform the device coordinate system to the earth coordinate system; The Coriolis force deviation caused by the earth's rotation is eliminated through coordinate projection, and the absolute flow velocity distribution with posture calibration is output.

6. The method according to claim 5, characterized in that The calculation of the error interval of each depth unit according to the absolute velocity distribution and the echo intensity dynamic range constraint, and outputting the final velocity profile with a confidence mark, includes: According to the data of each depth unit in the absolute velocity distribution, the dynamic range characteristics of the echo intensity are extracted to generate the signal-to-noise ratio attenuation curve; Based on the signal-to-noise ratio attenuation curve, the velocity measurement variance is calculated using the Cramer-Rao lower bound theory, and the initial error range is output; The acoustic scattering intensity and water turbidity data are integrated to construct a multi-physics field coupled error correction model and generate correction error coefficients. Apply the corrected error coefficient to the initial error interval and use Monte Carlo simulation to verify the confidence probability; According to the preset confidence level, the confidence level of the velocity value of each depth unit is marked, and the final velocity profile is output.

7. An acoustic Doppler flow velocity measurement system, characterized in that: The system comprises: The activation module is used to automatically activate the preset high-resolution broadband mode or large-range narrowband mode according to the depth characteristics and accuracy requirements of the target water area, and generate a transmission parameter combination adapted to the current hydrological environment; a correction module for collecting original echo signals based on the transmission parameter combination, combining the real-time temperature data of the transducer, correcting the Doppler shift using a pre-trained compensation model, and outputting a temperature-compensated three-dimensional flow velocity vector; a processing module, configured to input the three-dimensional velocity vector into a dynamic segmentation network, segment the profile layer according to a preset depth unit size, perform echo intensity normalization processing, and generate a layered velocity distribution model; A compensation module is used to call the layered flow velocity distribution model, integrate the fluxgate compass azimuth data to perform flow direction compensation, trigger the correction algorithm when the device tilt exceeds the limit, and output the absolute flow velocity distribution of the posture calibration; The output module is used to calculate the error interval of each depth unit according to the absolute velocity distribution and the echo intensity dynamic range constraint, and output the final velocity profile with a confidence mark.

8. The system according to claim 7, characterized in that The activation module is specifically used to: Based on the sonar depth detection data of the target waters and the preset accuracy threshold, the depth-accuracy feature matrix is ​​calculated and the mode selection decision factor is output; Based on the decision factor, the frequency configuration library of high-resolution broadband mode or large-range narrowband mode is activated, and the frequency parameter set including center frequency, bandwidth and pulse width is extracted; Combined with the current water salinity sensor data, the pulse repetition period parameters are dynamically adjusted through sound velocity compensation to generate a basic transmission parameter set; According to the basic transmission parameter set and frequency parameter set, the combination of sound beam opening angle and transmission power is optimized, and the final transmission parameter combination is output.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 6 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 6.

Images