High-precision river channel surface flow velocity measuring device integrating LSPIV and optical flow method

Abstract

The invention provides a high-precision river surface flow velocity measuring device fusing LSPIV and an optical flow method, which comprises an image acquisition module, a preprocessing module, an LSPIV velocity measuring unit, an optical flow method velocity measuring unit and a fusion output module, and is characterized in that the image acquisition module is used for acquiring continuous video frames of a river surface and comprises a multispectral imaging unit and a polarized light imaging unit; the polarized light imaging unit generates 0-degree, 45-degree, 90-degree and 135-degree polarized images through a four-angle cross-polarization splitting prism. According to the invention, through a multispectral and polarization imaging technology, interference of water surface reflection, low illumination and the like is effectively inhibited, the contrast ratio of tracer particles and the background is improved, and the image acquisition quality under complex conditions is ensured; the large-scale cross-correlation stability of the LSPIV and the sub-pixel displacement precision of the optical flow method are fused, the weight is dynamically adjusted according to the tracer density, the flow velocity gradient and the environmental error, and the limitation of a single method in a sparse tracer or high flow velocity gradient scene is solved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of river surface flow velocity measurement, in particular to a high-precision river surface flow velocity measurement device fusing LSPIV and an optical flow method. BACKGROUND

[0002] The traditional river flow velocity measurement technology has the following limitations:

[0003] 1. LSPIV technology: dependent on tracer particle density, prone to matching errors in sparse distribution or high flow velocity gradient areas, and fixed window size is difficult to adapt to curved river morphology.

[0004] 2. Optical flow method: sensitive to image noise, prone to local mismatch under strong light reflection or turbulent conditions, and the resolution of the sparse optical flow field after interpolation is insufficient.

[0005] 3. Hardware limitations: fixed angle imaging is easily affected by terrain inclination or water surface fluctuation, resulting in image distortion; a single sensor is difficult to cover extreme scenes such as night and turbid water.

[0006] 4. Stiff fusion algorithm: traditional weight distribution relies on empirical values and cannot dynamically respond to environmental changes, leading to error accumulation.

[0007] Therefore, a high-precision river surface flow velocity measurement device fusing LSPIV and an optical flow method is proposed. SUMMARY

[0008] The application provides a high-precision river surface flow velocity measurement device fusing LSPIV and an optical flow method to solve the problems in the background art.

[0009] The specific technical solutions are as follows:

[0010] A high-precision river surface flow velocity measurement device fusing LSPIV and an optical flow method comprises:

[0011] An image acquisition module is used to acquire continuous video frames of the river surface, comprising a multi-spectral imaging unit and a polarized light imaging unit, the polarized light imaging unit generates 0°, 45°, 90° and 135° polarized images through a four-angle orthogonal polarization beam splitter to improve the contrast of tracer particles and the background;

[0012] A preprocessing module is used to perform gray scale conversion, image segmentation and tracer density statistics on the video frames, and extract the number of feature points in a unit grid through grid division, and select a velocity measurement mode according to the tracer density;

[0013] An LSPIV velocity measurement unit calculates the surface flow velocity based on the large-scale particle image cross-correlation algorithm, comprising an adaptive window adjustment module that dynamically adjusts the cross-correlation calculation window size according to the tracer distribution density.

[0014] The optical flow method velocity measurement unit extracts feature point optical flow vectors by using a frequency domain cross-correlation matching algorithm, and improves displacement estimation precision by combining multi-scale SUSAN feature point detection and sub-pixel positioning technology.

[0015] The fusion output module synthesizes the flow velocity results of the LSPIV and the optical flow method based on a weighted fusion algorithm, dynamically adjusts the weight coefficient through error analysis, and outputs the final flow velocity field.

[0016] As a preferred scheme of the present application, the preprocessing module comprises:

[0017] The spectral filtering unit selects the 360-520 nm or 760-1050 nm waveband imaging according to the ambient light intensity by using an electrically tunable filter LCTF, so as to enhance the image quality under low light or high reflection conditions.

[0018] The image distortion correction unit reconstructs the river channel three-dimensional terrain by using binocular stereo vision technology, and corrects the image deformation caused by the visual angle tilt by combining a virtual elevation iteration method.

[0019] As a preferred scheme of the present application, the LSPIV velocity measurement unit performs particle matching by using a robust least square difference MQD algorithm, and divides the triangular mesh to adapt to the curved river channel shape, so as to reduce invalid calculation nodes.

[0020] As a preferred scheme of the present application, the optical flow method velocity measurement unit comprises:

[0021] The feature point screening module removes clustered feature points by using a sliding window, and retains uniformized feature points with local maximum R values.

[0022] The motion vector interpolation module interpolates the sparse optical flow field into a uniform meshed flow velocity field based on an inverse distance weighting method, so that the spatial resolution reaches the single-pixel level.

[0023] As a preferred scheme of the present application, the image acquisition module is mounted on a six-degree-of-freedom servo gimbal, the gimbal realizes vertical visual angle imaging through attitude feedback control, and is combined with a flight balloon suspended above the river channel, the takeoff height is adjustable in the range of 50-200 meters, and the wind resistance is greater than five levels.

[0024] As a preferred scheme of the present application, the error analysis of the fusion output module comprises:

[0025] The velocity interval judgment unit directly outputs the mean value if the LSPIV and the optical flow method result are located in the same flow velocity interval and the error is less than 5%.

[0026] The weighting correction unit dynamically allocates weights according to tracer density and flow velocity gradient if the error exceeds 5%, and the LSPIV result is preferentially used in the high-density area, and the optical flow result is preferentially used in the low-density area.

[0027] As a preferred scheme of the present application, the real-time monitoring feedback module is further included, the flow field, flow and water level change curve are displayed through the mobile workstation, an alarm signal is triggered based on the flow field abnormal value, and remote parameter adjustment is supported.

[0028] As a preferred scheme of the present application, the high-precision river surface flow velocity measuring device fusing LSPIV and optical flow method supports multi-sensor integrated calibration, and comprises the following:

[0029] The infrared thermal imaging unit is used for auxiliary imaging at night or in turbid water flow conditions.

[0030] The water surface elevation radar is fused with stereo vision data to calculate the cross-section flow.

[0031] As a preferred scheme of the present application, the optical flow velocity measuring unit further adopts a CLG (Combined Local-Global) optical flow algorithm, combines Horn-Schunck global smoothing constraint and local motion estimation, so as to improve the flow field continuity and noise resistance.

[0032] As a preferred scheme of the present application, the high-precision river surface flow velocity measuring device fusing LSPIV and optical flow method enhances tracer particle detection through a spectrum-polarization imaging technology in the form of an artificial compound eye, and specifically comprises the following:

[0033] The artificial dragonfly compound eye camera synchronously captures spectral intensity and polarization information through a four-channel polarization beam splitter prism.

[0034] The neural signal processing model simulates the transverse inhibition mechanism of the medulla layer of the compound eye, extracts the spatial position feature and apparent polarization feature of weak particles.

[0035] The present application has the following beneficial effects:

[0036] The application provides a high-precision river surface flow velocity measuring device fusing LSPIV and an optical flow method, which effectively suppresses interference such as water surface reflection and low light through multi-spectral and polarization imaging technology, improves the contrast of tracer particles and background, and ensures the image acquisition quality under complex conditions; fuses the large-scale cross-correlation stability of LSPIV and the sub-pixel displacement accuracy of the optical flow method, dynamically adjusts the weight according to the tracer density, flow velocity gradient and environmental error, and solves the limitations of a single method in sparse tracer or high flow velocity gradient scenes; combines binocular stereo vision and a virtual elevation iteration method to correct image distortion caused by viewing angle tilt or water surface fluctuation, and ensures the spatial consistency of three-dimensional terrain reconstruction and flow velocity field; hardware design (such as aerial balloon carrying and six-degree-of-freedom holder) supports vertical viewing angle imaging under different river width and terrain conditions, and the wind resistance and height adjustability adapt to stable operation under harsh environments; the flow velocity field, flow and abnormal alarm function are displayed in real time, combined with remote parameter adjustment, to improve the response efficiency and operation convenience of emergency hydrological monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 A composition block diagram of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method is provided for the embodiments of the application.

[0038] Figure 2 A connection relationship diagram between the modules of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method is provided for the embodiments of the application.

[0039] Figure 3 An error change diagram of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method under different tracer densities is provided for the embodiments of the application.

[0040] Figure 4 A multi-scene robustness expansion column chart of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method is provided for the embodiments of the application.

[0041] Figure 5 An error comparison column chart of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method under a strong reflection scene is provided for the embodiments of the application.

[0042] Figure 6 A spatial data precision improvement curve diagram of the high-precision river surface flow velocity measuring device fusing LSPIV and the optical flow method is provided for the embodiments of the application. DETAILED DESCRIPTION

[0043] The technical solutions of the application are further illustrated below by specific embodiments in combination with the drawings.

[0044] Among them, the drawings are only used for illustrative description, and the representation is only a schematic diagram, not a physical diagram, and cannot be understood as a limitation on the patent; in order to better illustrate the embodiments of the present application, some components in the drawings will be omitted, enlarged or reduced, and do not represent the size of the actual product; for those skilled in the art, it is understandable that some well-known structures in the drawings and their descriptions may be omitted.

[0045] The same or similar reference numerals in the drawings of the embodiments of the present application correspond to the same or similar components; in the description of the present application, it should be understood that if the terms "upper", "lower", "left", "right", "inner", "outer" and the like indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the drawings, only for the convenience of describing the present application and simplifying the description, and not indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore the positional relationship described in the drawings is only used for illustrative description, and cannot be understood as a limitation on the patent, for those skilled in the art, the specific meaning of the above terms can be understood according to the specific situation.

[0046] In the description of the present application, unless otherwise explicitly specified and limited, if the term "connection" and the like indicating the connection relationship between components appears, the term should be broadly understood, for example, it can be fixedly connected, or it can be detachably connected, or it can be integrated; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two components or the interaction relationship between two components. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific situation.

[0047] The high-precision river surface flow velocity measuring device provided by the embodiment fuses LSPIV and optical flow method, as shown in Figures 1-6 , wherein Figure 1 is a composition block diagram of the device, showing the main composition modules; Figure 2 is a connection relationship diagram between the modules of the device, showing the connection relationship between the modules; Figure 3 is an error change diagram of the device under different tracer densities, showing the error change of different algorithms (LSPIV, optical flow method, fusion) under different tracer densities, it can be seen that the fusion algorithm shows lower average error in different density areas, which reflects the advantage of dynamic algorithm cooperative optimization; Figure 4 is a multi-scene robustness expansion column chart of the device, showing the wind resistance of the device under different lift-off heights (50-200 meters), all greater than five levels, which reflects the characteristics of multi-scene robustness expansion; Figure 5For the error contrast column chart of the device in the strong reflection light scene, it can be clearly seen from the column chart that the error after the technical improvement is significantly reduced from 12.7% to 5.3% in the strong reflection light scene, which reflects the enhancement of the adaptability in complex environment; Figure 6 For the spatial data precision improvement curve chart of the device, the improvement of spatial consistency of the simulation data is displayed, that is, the change of the spatial consistency index before and after correction. Through the simulation data, the curve chart shows the change of the spatial consistency index (assuming the correlation coefficient) before and after correction. After correction, the spatial consistency index is significantly improved, which reflects the enhancement of the spatial data precision. The high-precision river surface flow velocity measurement device fusing LSPIV and optical flow method includes an image acquisition module, a preprocessing module, an LSPIV velocity measurement unit, an optical flow method velocity measurement unit and a fusion output module, wherein:

[0048] The image acquisition module is used to acquire continuous video frames of the river surface, including a multi-spectral imaging unit and a polarized light imaging unit. The polarized light imaging unit generates 0°, 45°, 90° and 135° polarized images through a four-angle orthogonal polarization beam splitter to improve the contrast between the tracer particles and the background;

[0049] The preprocessing module is used for gray scale conversion, image segmentation and tracer density statistics of the video frames, and the number of feature points in a unit grid is extracted through grid division. The velocity measurement mode is selected according to the tracer density;

[0050] The LSPIV velocity measurement unit calculates the surface flow velocity based on the large-scale particle image mutual correlation algorithm, including an adaptive window adjustment module, which dynamically adjusts the mutual correlation calculation window size according to the tracer distribution density;

[0051] The optical flow method velocity measurement unit extracts feature point optical flow vectors using a frequency domain mutual correlation matching algorithm, and combines multi-scale SUSAN feature point detection and sub-pixel positioning technology to improve displacement estimation accuracy;

[0052] The fusion output module integrates the flow velocity results of LSPIV and optical flow method based on a weighted fusion algorithm, dynamically adjusts the weight coefficient through error analysis, and outputs the final flow velocity field.

[0053] The high-precision river surface flow velocity measurement device fusing LSPIV and optical flow method using the above technical scheme enhances the image acquisition capability in complex environment through multi-spectral and polarized imaging technology, combines the large-scale mutual correlation advantage of LSPIV and the sub-pixel displacement estimation characteristics of optical flow method, and realizes high-precision flow velocity measurement. The dynamic fusion algorithm solves the limitations of single method in the scene where the tracer distribution is uneven or the flow velocity gradient is large, and can improve the adaptability and measurement reliability of the whole system.

[0054] Specifically, in the embodiment, the preprocessing module includes a spectral filter unit and an image distortion correction unit, wherein:

[0055] The spectral filtering unit selects 360-520nm or 760-1050nm waveband imaging according to the intensity of ambient light by an electrically tunable filter LCTF, so as to enhance the image quality under low light or high reflection conditions.

[0056] The image distortion correction unit reconstructs the three-dimensional terrain of the river channel by using binocular stereo vision technology, and corrects the image deformation caused by the tilt of the viewing angle by combining a virtual elevation iteration method.

[0057] The preprocessing module adopting the technical scheme can optimize the imaging quality under different light conditions through the spectral filtering unit, so as to reduce the interference of ambient light; the image distortion correction unit is set to eliminate the image deformation caused by the tilt of the viewing angle or the fluctuation of the water surface by three-dimensional terrain reconstruction and iterative calibration, so as to improve the spatial data accuracy.

[0058] Specifically, in the embodiment, the LSPIV velocity measurement unit adopts a robust least square difference MQD algorithm for particle matching, and triangular mesh division is adopted to adapt to the curved river channel shape, so as to reduce invalid calculation nodes.

[0059] The robust least square difference algorithm and triangular mesh division are adopted to adapt to the complex shape of the curved river channel, reduce the proportion of invalid calculation regions, and improve the calculation efficiency and result consistency of the LSPIV algorithm in irregular river channels.

[0060] Specifically, in the embodiment, the optical flow method velocity measurement unit includes a feature point screening module and a motion vector interpolation module, wherein:

[0061] The feature point screening module removes clustered feature points by a sliding window and retains uniformized feature points with local maximum R values;

[0062] The motion vector interpolation module interpolates the sparse optical flow field into a uniform meshed flow velocity field based on an inverse distance weighting method, so that the spatial resolution reaches a single-pixel level.

[0063] The feature point screening module avoids the interference of local clustered feature points on the optical flow vector, which can improve the uniformity of displacement estimation; the motion vector interpolation module is set to convert the sparse optical flow field into a high-resolution flow velocity field by spatial interpolation technology, which can enhance the ability to capture detailed flow velocity changes.

[0064] Specifically, in the embodiment, the image acquisition module is mounted on a six-degree-of-freedom servo gimbal, the gimbal realizes vertical viewing angle imaging through attitude feedback control, and is combined with a aerial balloon suspended above the river channel, the takeoff height is adjustable in the range of 50-200 meters, and the wind resistance is greater than five levels.

[0065] The above technical scheme adopts the six-degree-of-freedom servo holder in combination with the adjustable suspension design of the aerial balloon to ensure the stability of vertical visual angle imaging, and can adapt to different river width and terrain conditions; and the anti-wind design can ensure reliable operation of the device in bad weather.

[0066] Specifically, in the embodiment, the error analysis of the fusion output module includes a speed interval judgment unit and a weighted correction unit, wherein:

[0067] The speed interval judgment unit directly outputs the mean value if the LSPIV and the optical flow method result are located in the same flow speed interval and the error is less than 5%.

[0068] The weighted correction unit dynamically allocates weights according to the tracer density and the flow speed gradient if the error exceeds 5%, and the LSPIV result is preferentially used in high-density areas, and the optical flow method result is preferentially used in low-density areas.

[0069] The above technical scheme adopts the dynamic weight allocation mechanism based on the flow speed interval and the error threshold to automatically optimize the priority of the algorithm in areas with large differences in tracer density, reduce the influence of single method error on the final result, and thus improve the robustness of the fusion result.

[0070] Specifically, in the embodiment, a real-time monitoring feedback module is further included, which displays the flow speed field, flow and water level change curve through a mobile workstation, triggers an alarm signal based on flow field abnormal value, and supports remote parameter adjustment.

[0071] The above technical scheme sets the real-time monitoring module to provide visual flow field data and abnormal alarm function, support remote parameter adjustment, enhance the real-time response capability and operation convenience of the system, and be applicable to emergency hydrological monitoring scenes.

[0072] Specifically, in the embodiment, the high-precision river surface flow speed measuring device fusing LSPIV and optical flow method supports multi-sensor integrated calibration, including an infrared thermal imaging unit and a water surface elevation radar, wherein:

[0073] The infrared thermal imaging unit is used for auxiliary imaging at night or in turbid water flow conditions.

[0074] The water surface elevation radar and the stereo vision data are fused to calculate the cross-section flow.

[0075] The above technical scheme adopts multi-sensor integration to expand the applicability of the device in low light, turbid water and other extreme environments; and the fusion of the water surface elevation radar and the stereo vision data can improve the comprehensive precision of cross-section flow calculation.

[0076] Specifically, in the embodiment, the optical flow method velocity measurement unit further adopts a CLG (Combined Local-Global) optical flow algorithm, which combines the Horn-Schunck global smoothing constraint and local motion estimation, to improve the continuity of the flow field and the noise resistance.

[0077] By adopting the above technical solution, the CLG optical flow algorithm combines the global smoothing constraint and the local motion estimation, which can reduce the noise interference and the motion discontinuity problem, and enhance the continuity and the calculation stability of the optical flow field under complex water surface texture.

[0078] Specifically, in the embodiment, the high-precision river surface flow velocity measurement device fusing LSPIV and optical flow method enhances the detection of tracer particles through the spectrum-polarization imaging technology of the mimic compound eye, specifically including: a mimic dragonfly compound eye camera and a neural signal processing model, wherein:

[0079] The mimic dragonfly compound eye camera synchronously captures the spectral intensity and polarization information through a four-channel polarization beam splitter prism;

[0080] The neural signal processing model simulates the horizontal inhibition mechanism of the medulla layer of the compound eye, extracts the spatial position features and apparent polarization features of the weak tracer particles.

[0081] By adopting the above technical solution, the mimic compound eye imaging technology simulates the biological visual mechanism, enhances the detection ability of the weak tracer particles through the polarization beam splitting and horizontal inhibition model, and can improve the image resolution in the low-contrast or high-reflectivity area.

[0082] Specifically, in the embodiment, the fusion output module adopts a dynamic confidence weight equation for flow velocity field fusion calculation, and the core equation is:

[0083]

[0084] In the formula:

[0085] and R and R are dynamic weight coefficients of LSPIV and optical flow method respectively;

[0086] R is the peak signal-to-noise ratio of the LSPIV cross-correlation calculation (range 0~1);

[0087] ΔE is the normalized standard deviation of the gradient residual of the adjacent frames of the optical flow method (range 0~1);

[0088] R0 is the cross-correlation reliability threshold (default 0.6);

[0089] k1 is the S-type function gain coefficient (default 8);

[0090] k2 is the error suppression coefficient (default 3).

[0091] Examples are as follows:

[0092] Data input:

[0093] The LSPIV unit outputs the correlation peak signal-to-noise ratio RR (obtained by calculating the ratio of the maximum peak to the secondary peak of the correlation matrix) of the current window; the optical flow unit outputs the gradient residual standard deviation ΔE (obtained by residual statistics of the optical flow equation ) of adjacent frames;

[0094] 2. Weight calculation:

[0095] When R > R0, tends to 0, and wLSPIV approaches 1, the LSPIV result is preferentially adopted;

[0096] When ΔE increases (the error of the optical flow method is significant), the denominator term 1+k2·ΔE suppresses wLSPIV, and the weight of LSPIV is reduced;

[0097] 3. Fusion execution:

[0098] Final flow rate .

[0099] The parameter design principle is as follows:

[0100] 1. S-shaped function controls the correlation reliability:

[0101] When the correlation peak signal-to-noise ratio R of LSPIV is higher than the threshold R0, the weight exponentially increases, and the dominance of the high-reliability LSPIV result is strengthened;

[0102] The gain coefficient k1=8 ensures that the weight exceeds 0.88 when R reaches 0.75 (calibrated through experiments);

[0103] 2. Error suppression term optimizes robustness:

[0104] The gradient residual standard deviation ΔE of the optical flow method reflects the consistency of motion estimation, and k2=3 ensures that the weight decreases by more than 50% when ΔE>0.3.

[0105] Technical effects

[0106] 1. Dynamic adaptability: in the dense region of tracer particles (R is high), the weight of LSPIV is dominant, and the large-scale correlation stability is utilized; in the low-contrast region (ΔE is low), the weight of the optical flow method is improved, and the advantage of sub-pixel displacement estimation is exerted;

[0107] 2. Error suppression: when local mismatching (ΔE suddenly increases) occurs in the optical flow method, the contribution weight is automatically reduced, and error propagation is avoided; compared with the traditional fixed-weight fusion (such as 0.5:0.5), the average error is reduced by 32% (verified by experimental data).

[0108] The working principle of the equation is as follows:

[0109] 1. Data synchronization: LSPIV and optical flow method unit are calculated in parallel, and V LSPIV , R, V OF , ΔE are output.

[0110] 2. Confidence evaluation: real-time calculation of dynamic weight wLSPIV and w OF

[0111] 3. Field fusion: for each calculation grid, wLSPIV and w are calculated.

[0112] 4. Abnormal processing: when wLSPIV and w OF are lower than 0.2 at the same time, the real-time monitoring module is triggered to reacquire data.

[0113] The equation relates the cross-correlation signal-to-noise ratio (a parameter unique to LSPIV) and the gradient residual (a parameter unique to the optical flow method) through a fractional-point array combination, breaking through the limitations of simple linear superposition in the prior art (such as patent CN105486328A); through the physical meaning definition of R0, k1 and k2, the equation has the ability of parameter self-adaptive adjustment, which is superior to the traditional fixed coefficient model (such as patent US20170261324A1); the suppression term is designed for the common problems of uneven distribution of tracers (leading to RR fluctuation) and water surface reflection interference (leading to ΔE mutation) in river measurement, which can reduce the fusion error from 12.7% to 5.3% in strong reflection scene.

[0114] By adopting the above technical scheme, the dynamic confidence weight equation realizes self-adaptive adjustment of the weight by nonlinearly coupling the cross-correlation signal-to-noise ratio of LSPIV and the gradient residual of the optical flow method, solves the rigidity problem of the traditional fixed weight fusion, and reduces the influence of local error matching on the global result, thereby significantly improving the fault tolerance and precision of the fusion algorithm in complex environment.

[0115] Experimental verification method

[0116] 1. Comparative experiment design

[0117] Control group: traditional fixed weight fusion method (LSPIV and optical flow method weight each 0.5), independent LSPIV algorithm, independent optical flow algorithm;

[0118] Experimental group: dynamic weight fusion method of the application;

[0119] Benchmark true value: acoustic Doppler current profiler (ADCP) or particle tracking velocimetry (PTV) is used as the true value of flow velocity.

[0120] Test scene:

[0121] Scenario 1: Dense tracer particles (R>0.8), low flow velocity gradient (AV<0.2m / s)

[0122] Scenario 2: Sparse tracer particles (R<0.4), high flow velocity gradient (AV>1.5m / s)

[0123] Scenario 3: Strong water surface reflection (ΔE>0.5), obvious turbulent flow.

[0124] 2. Data acquisition and processing

[0125] Data volume: 100 sets of effective flow field data are collected for each scenario (each scene contains 1000-5000 measurement points);

[0126] Error calculation:

[0127] Absolute error:

[0128] Relative error:

[0129] Statistical method: take the mean and standard deviation of the relative error of all measurement points under each scenario.

[0130] Experimental results table

[0131]

[0132] Error reduction amplitude calculation:

[0133] Dynamic weight vs. fixed weight:

[0134] Dynamic weight vs. best independent algorithm:

[0135] In summary, the high-precision river surface flow measurement device fusing LSPIV and optical flow method provided by the embodiment has the following advantages:

[0136] 1. Enhanced adaptability to complex environments: through multispectral and polarization imaging technology, effectively suppress water surface reflection, low light interference, etc., improve the contrast between tracer particles and background, and ensure the image acquisition quality under complex conditions.

[0137] 2. Dynamic algorithm collaborative optimization: fuse the large-scale cross-correlation stability of LSPIV and the sub-pixel displacement precision of optical flow method, dynamically adjust the weight according to the tracer density, flow velocity gradient and environmental error, and solve the limitations of single method in sparse tracer or high flow velocity gradient scene.

[0138] ​​​​3. Spatial data precision improvement: Combining binocular stereo vision and virtual elevation iteration method, corrects image distortion caused by viewing angle tilt or water surface fluctuation, ensuring the spatial consistency of three-dimensional terrain reconstruction and flow field.

[0139] 4. Multi-scene robustness expansion: Hardware design (such as aerial balloon carrying, six-degree-of-freedom gimbal) supports vertical angle imaging of different river widths and terrain conditions, wind resistance and height adjustability adapt to stable operation in harsh environments.

[0140] 5. Intelligent monitoring and feedback: Real-time display of flow field, flow rate and abnormal alarm function, combined with remote parameter adjustment, improves the response efficiency and operational convenience of emergency hydrological monitoring.

[0141] Among them, the connection relationship of each module is as follows:

[0142] 1. Connection of image acquisition module and other modules

[0143] Physical connection:

[0144] Multispectral imaging unit and polarized light imaging unit are integrated in the same hardware platform through optical interface, and polarized light splitting prism and imaging sensor are directly coupled.

[0145] Six-degree-of-freedom servo gimbal is connected with the main control unit through cable or wireless communication protocol (such as RS485, CAN bus), and receives attitude adjustment instructions.

[0146] Data flow:

[0147] The collected raw video frames are transmitted in real time to the preprocessing module through high-speed data transmission interface (such as HDMI or gigabit Ethernet).

[0148] 2. Input and output of preprocessing module

[0149] Input: Receive raw multispectral and polarized image data from the image acquisition module.

[0150] Output:

[0151] Processed gray-scale images, segmented images, and tracer density statistics are sent to the LSPIV speed measurement unit and the optical flow method speed measurement unit through parallel data channels, respectively.

[0152] Distortion-corrected three-dimensional terrain data are stored in shared memory for fusion output module calling.

[0153] 3. Data interaction of LSPIV speed measurement unit and optical flow method speed measurement unit

[0154] Input:

[0155] LSPIV unit receives the grayscale image and tracer density data from the preprocessing module, and the adaptive window adjustment module dynamically adjusts the calculation parameters according to the density.

[0156] Optical flow unit receives the segmented image and feature point data from the preprocessing module, and performs optical flow calculation combined with multi-scale SUSAN detection results.

[0157] Output:

[0158] LSPIV unit outputs the gridded flow field and cross-correlation signal-to-noise ratio (R value).

[0159] Optical flow unit outputs the interpolated high-resolution flow field and gradient residual standard deviation (ΔE).

[0160] 4. Core connection of fusion output module

[0161] Input:

[0162] Receive the flow field and R value from the LSPIV unit, and the flow field and ΔE from the optical flow unit.

[0163] Data processing:

[0164] Dynamic weight equation calculates the fusion weight based on R and ΔE, and performs flow field weighted fusion.

[0165] Call the three-dimensional terrain data of the preprocessing module to perform spatial coordinate mapping and error correction.

[0166] Output:

[0167] The final flow field is transmitted to the real-time monitoring feedback module through a high-speed data interface.

[0168] 5. Closed-loop control of real-time monitoring feedback module

[0169] Input: Receive the final flow field and flow calculation results of the fusion output module.

[0170] Output:

[0171] Mobile workstations display flow field, water level curve and alarm signals (through GUI interface).

[0172] After abnormality detection is triggered, parameter adjustment instructions (such as adjusting the pan-tilt angle and switching the imaging waveband) are sent to the image acquisition module through wireless network (such as 4G / 5G).

[0173] 6. Connection of multi-sensor integrated calibration

[0174] Infrared thermal imaging unit and water surface elevation radar access the preprocessing module through independent data channels, and after data and stereo vision information fusion, they are used for auxiliary imaging and cross-section flow calculation under night or turbid water flow conditions.

[0175] The working principle of the high-precision river surface flow velocity measurement device that fuses LSPIV and optical flow method is as follows:

[0176] 1. Image acquisition and enhancement:

[0177] The multispectral imaging unit switches the waveband according to the ambient light (such as near-infrared for low light), and the polarized light imaging generates multi-directional polarization images through a four-angle light splitting prism, suppresses water surface reflection, and enhances the detection of tracer particles.

[0178] The aerial balloon is equipped with a six-degree-of-freedom gimbal to adjust the height and viewing angle, ensuring the stability of vertical imaging.

[0179] 2. Preprocessing and correction:

[0180] After optimizing the image quality through spectral filtering, perform gray scale conversion, image segmentation, and tracer density statistics.

[0181] Reconstruct the river three-dimensional terrain through binocular stereo vision, correct image distortion through virtual elevation iteration method, and eliminate viewing angle tilt error.

[0182] 3. Parallel velocity calculation:

[0183] LSPIV unit: adaptive window adjustment and triangular mesh division are used to adapt to the curved shape of the river, and the MQD algorithm is used to improve the particle matching accuracy.

[0184] Optical flow method unit: combined with multi-scale feature point detection and CLG optical flow algorithm, sub-pixel displacement is extracted, and high-resolution flow velocity field is generated through inverse distance weighted interpolation.

[0185] 4. Dynamic fusion and output:

[0186] According to the cross-correlation signal-to-noise ratio (R) of LSPIV and the gradient residual (ΔE) of optical flow method, the fusion weight is calculated using the dynamic confidence weight equation.

[0187] High-density areas prefer LSPIV results, low-density or high-error areas focus on optical flow method, and output a uniform high-precision flow velocity field.

[0188] 5. Real-time monitoring and feedback:

[0189] The mobile workstation visualizes the flow velocity field and flow curve, triggers an alarm for abnormal values, and supports remote parameter adjustment to achieve closed-loop control.

[0190] The above merely preferred embodiments of the present application and are not intended to limit the embodiments and protection scope of the present application. Those skilled in the art should be able to understand that any equivalent substitutions and obvious changes made according to the present application description and drawings should be included in the protection scope of the present application.

Claims

1. A high-precision river surface velocity measurement device integrating LSPIV and optical flow method, characterized in that, include: The image acquisition module is used to acquire continuous video frames of the river surface, including a multispectral imaging unit and a polarization imaging unit. The polarization imaging unit generates 0°, 45°, 90° and 135° polarization images through a four-angle orthogonal polarization beam splitter to improve the contrast between the tracer particles and the background. The preprocessing module is used to perform grayscale conversion, image segmentation, and tracer density statistics on video frames, and extract the number of feature points within a unit grid by grid division, and select the speed measurement mode according to the tracer density. The LSPIV velocimetry unit calculates surface flow velocity based on a large-scale particle image cross-correlation algorithm, including an adaptive window adjustment module that dynamically adjusts the cross-correlation calculation window size according to the tracer distribution density. The optical flow velocimetry unit uses a frequency domain cross-correlation matching algorithm to extract the optical flow vector of feature points, and combines multi-scale SUSAN feature point detection and sub-pixel positioning technology to improve the displacement estimation accuracy. The fusion output module integrates the velocity results of LSPIV and optical flow based on a weighted fusion algorithm, and dynamically adjusts the weighting coefficients through error analysis to output the final velocity field.

2. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The preprocessing module includes: The spectral filtering unit, through an electrically adjustable filter LCTF, selects the 360-520nm or 760-1050nm band for imaging based on the ambient light intensity, in order to enhance image quality under low light or high reflectivity conditions. The image distortion correction unit uses binocular stereo vision technology to reconstruct the three-dimensional topography of the river channel and combines virtual elevation iteration method to correct image distortion caused by viewing angle tilt.

3. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The LSPIV velocity measurement unit uses the robust least squares difference (MQD) algorithm for particle matching and adapts to the curved river morphology through triangular mesh partitioning to reduce invalid computation nodes.

4. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The optical flow velocimetry unit includes: The feature point filtering module uses a sliding window to remove clustered feature points and retains homogeneous feature points with local maximum R values. The motion vector interpolation module interpolates the sparse optical flow field into a uniform gridded velocity field based on the inverse distance weighting method, so that the spatial resolution reaches the single-pixel level.

5. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The image acquisition module is mounted on a six-degree-of-freedom servo gimbal. The gimbal achieves vertical perspective imaging through attitude feedback control and, combined with a drone balloon, floats above the river. The altitude is adjustable from 50 to 200 meters and has a wind resistance greater than level 5.

6. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The error analysis of the fusion output module includes: The velocity range judgment unit directly outputs the average value if the LSPIV and optical flow results are in the same velocity range and the error is less than 5%. The weighted correction unit dynamically allocates weights based on tracer density and flow velocity gradient if the error exceeds 5%. In high-density regions, LSPIV results are preferred, while in low-density regions, optical flow results are preferred.

7. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, It also includes a real-time monitoring and feedback module, which displays the velocity field, flow rate and water level change curves through a mobile workstation, and triggers alarm signals based on abnormal flow field values, and supports remote parameter adjustment.

8. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The high-precision river surface velocity measurement device integrating LSPIV and optical flow methods supports multi-sensor integrated calibration, including: Infrared thermal imaging unit for auxiliary imaging at night or in turbid water conditions; Water surface elevation radar is fused with stereo vision data to calculate cross-sectional flow.

9. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 4, characterized in that, The optical flow velocimetry unit further employs the CLG optical flow algorithm, combined with Horn-Schunck global smoothing constraints and local motion estimation, to improve flow field continuity and noise resistance.

10. The high-precision river surface velocity measurement device integrating LSPIV and optical flow method according to claim 1, characterized in that, The high-precision river surface velocity measurement device integrating LSPIV and optical flow methods enhances tracer particle detection through spectral-polarization imaging technology mimicking the shape of a compound eye, specifically including: A dragonfly compound eye-like camera simultaneously captures spectral intensity and polarization information using a four-channel polarization beam splitter. A neural signal processing model was developed to simulate the transverse inhibition mechanism of the medullary layer of the compound eye, and to extract the spatial location and apparent polarization features of occult particles.

Images