A method and device for intelligently measuring flow velocity of river engineering model

Abstract

The invention discloses an intelligent flow velocity measurement method and device for a river engineering model, belonging to the technical field of river engineering model test measurement, and specifically comprises: obtaining a model flow field image sequence and performing preprocessing; automatically identifying and marking tracer particles by using an image recognition algorithm; calculating the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles in a continuous time period, in combination with an image acquisition frame rate and a model scale; and reconstructing a three-dimensional flow velocity field by using a radial basis function; the method realizes non-contact, high-precision measurement of the flow field of the river engineering model, avoids errors and interferences that may be caused by a traditional physical contact measurement method, and improves the automation degree and efficiency of the measurement; by using the displacement information of the tracer particles and the image acquisition frame rate, in combination with the river engineering model scale, the instantaneous velocity of the tracer particles can be accurately calculated, and the three-dimensional flow velocity field can be reconstructed by spatial interpolation and radial basis functions, thereby improving the accuracy and efficiency of the river engineering model test.

Description

Technical Field

[0001] The invention belongs to the technical field of river engineering model test measurement, in particular to a river engineering model flow velocity intelligent measurement method and device. Background Art

[0002] River engineering model test is an important means to study the movement law of river water flow, the evolution process of riverbed and the influence of hydraulic engineering buildings on water flow. In river engineering model test, flow velocity is a very critical parameter. Accurately measuring the flow velocity in river engineering model is of great significance for studying the hydrodynamic characteristics of rivers, the movement law of sediment and the influence of engineering buildings on water flow. Traditional river engineering model flow velocity measurement methods mainly use contact measuring instruments such as propeller flowmeters and ultrasonic flowmeters. These methods have the following disadvantages: manual operation is required, the measurement speed is slow, it is difficult to meet the needs of large-scale model tests, and the operator's proficiency and the choice of measurement location will affect the measurement results. At the same time, contact measuring instruments will destroy the model flow field and affect the accuracy of the measurement results.

[0003] For example, a Chinese patent with the authorization announcement number CN103644898B discloses a method for identifying and measuring the water boundary of a river engineering model test, including: a guide rail parallel to the two banks of the model to be tested, a measuring bridge installed on the guide rail across the left and right banks of the model to be tested, and the measuring bridge can move horizontally on the guide rail; a measuring vehicle is provided on the measuring bridge, and the measuring vehicle can move horizontally on the measuring bridge; a lifting drive device and a lifting frame are installed on the measuring vehicle, and the lifting frame can move vertically; a measuring rod and a pressure sensor are installed on the lifting frame; and a measuring electrode is installed on the measuring rod. The driving and positioning of the measuring bridge, the measuring vehicle, and the lifting frame, as well as the contact between the measuring rod and the water surface are all controlled by a computer. The technical scheme also discloses a method for measuring the water boundary using this water boundary identification and measurement system for river engineering model tests, and the data of each measuring point is obtained by using measuring electrodes and pressure sensors. The present invention has high accuracy and high automation, and provides necessary detection methods and detection data results for improving the automatic measurement system of the flow velocity distribution of the model test section and subsequent data processing.

[0004] The above existing technologies all have the following problems: they rely on the physical measuring device to directly contact the water surface to obtain data. Such contact may cause wear of the measuring device, thereby affecting the long-term stability and accuracy of the measurement; they are limited in adapting to river engineering models of different shapes or complex terrains; and they still rely on the movement and positioning of the physical device during the measurement process, and the degree of automation and intelligence is low. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes an intelligent measurement method and device for flow velocity of a river engineering model, which obtains a model flow field image sequence and performs preprocessing; automatically identifies and marks tracer particles using an image recognition algorithm; calculates the instantaneous velocity of the tracer particles based on the displacement information of the tracer particles in a continuous time period, combined with the image acquisition frame rate and the model scale; and reconstructs the three-dimensional flow velocity field using radial basis functions, thereby improving the accuracy and efficiency of river engineering model tests.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for intelligently measuring flow velocity of a river engineering model, comprising:

[0008] Step S1: obtaining a model flow field image sequence of a river engineering model test area and performing preprocessing;

[0009] Step S2: using an image recognition algorithm to automatically identify and mark tracer particles on the preprocessed model flow field image;

[0010] Step S3: Calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles in the continuous time period, combined with the image acquisition frame rate and the scale of the river engineering model;

[0011] Step S4: performing spatial interpolation processing on the calculated tracer particle instantaneous velocity data, and using radial basis functions to reconstruct the three-dimensional velocity field in the river engineering model test area according to the known velocity data of the tracer particles;

[0012] The specific steps of step S3 include:

[0013] S3.1: Model flow field image sequences from consecutive time periods The position information of the tracer particles is extracted to obtain the coordinates of the tracer particles in each frame of the model flow field image, where: represents the first n Model flow field images, n Indicates the number of model flow field images in the model flow field image sequence;

[0014] S3.2: Obtain the frame rate of the model flow field image acquisition of the high-speed camera f and model scale S;

[0015] S3.3: Calculate the displacement of the tracer particles between adjacent frames based on the coordinate changes in each frame of the model flow field image. ;

[0016] S3.4: Use the optical flow method to calculate the displacement with sub-pixel accuracy to obtain the displacement information after sub-pixel accuracy calculation ;

[0017] S3.5: Use formula Convert the displacement information calculated with sub-pixel accuracy into actual displacement , and combined with the time interval , based on the formula Calculate the instantaneous velocity of the tracer particle ,in, and Respectively represent the tracer particles in each frame of the model flow field image x and y The velocity component in the direction;

[0018] S3.6: Use formula The error correction is performed on the calculated instantaneous speed, where: represents the final flow rate after error correction and optimization, represents the polynomial coefficients, m represents the polynomial order, j represents the polynomial order index, represents the salinity correction factor, represents the error in flow velocity measurement caused by salinity changes, represents the noise correction coefficient, Indicates the flow rate measurement error caused by noise interference;

[0019] S3.7: Based on the corrected instantaneous speed , output the instantaneous velocity of each tracer particle ,in, represents the direction of instantaneous velocity, and They represent the tracer particles in each frame of the model flow field image after error correction. x and y The component of velocity in the direction.

[0020] Specifically, the model flow field image sequence of the river engineering model test area is obtained by arranging a high-speed camera and an auxiliary light source in the river engineering model test area, and adopting a multi-angle synchronous shooting technology to obtain a multi-perspective image sequence of the model flow field to cover the complete flow field information of the test area.

[0021] Specifically, the automatic identification and marking of tracer particles in the preprocessed model flow field image using an image recognition algorithm includes:

[0022] S2.1: Obtain a preprocessed model flow field image sequence, and use a deep learning-based image recognition algorithm to detect tracer particles in the model flow field image;

[0023] S2.2: Based on the tracer particle identification results, initially mark the tracer particles and record the initial coordinates of each particle and feature information;

[0024] S2.3: Set the particles to move at a constant speed, combine the position information of the tracer particles in the continuous frame images, and use the adaptive threshold segmentation method to track the movement trajectory of each particle.

[0025] Specifically, the method of automatically identifying and marking the tracer particles in the preprocessed model flow field image using an image recognition algorithm also includes:

[0026] S2.4: According to the tracking results, dynamically update the position information of each particle and calculate its speed, while keeping the unique tag ID of the particle unchanged;

[0027] S2.5: Predict the particle motion trajectory based on the current position and velocity information of the particle;

[0028] S2.6: Correcting the particle motion trajectory according to the difference between the predicted result and the actual detection result;

[0029] S2.7: Match the corrected tracer particle motion trajectories, and integrate the motion trajectory information of all particles to generate a complete tracer particle motion data set; the complete tracer particle motion data set is a triplet data set including the tracer particle position, velocity and trajectory information.

[0030] Specifically, the specific steps of S2.3 include:

[0031] S2.31: Use the camera to obtain continuous frame images containing tracer particles at a preset frequency and method, and perform preprocessing to generate a model flow field image sequence of a continuous time period ,in, represents the first n Model flow field images, n Indicates the number of model flow field images in the model flow field image sequence;

[0032] S2.32: Segment the model flow field image using an adaptive threshold segmentation method based on the local mean to obtain a segmented binary image;

[0033] S2.33: Identify the position of the tracer particles in the segmented binary image;

[0034] S2.34: Combine the position information of the tracer particles in the continuous frame images to track the movement trajectory of each particle;

[0035] S2.35: Output the tracked particle motion trajectory.

[0036] Specifically, the specific steps of S3.4 include:

[0037] S3.41: Acquisition and the displacement of the tracer particles between adjacent frames ;

[0038] S3.42: Use Calculate the optical flow vector of each pixel in the model flow field image sequence ,in, , and Respectively represent the model flow field images in x , y , t The gradient in direction, t Indicates the current moment, u , v Represents the optical flow vector x , y Quantity, and represents the weight coefficient of the spatial smoothness constraint, , , and Respectively u , v exist x , y The second derivative in the direction, and They represent the noise weight coefficients, and Respectively u , v Directional noise;

[0039] S3.43: From the calculated optical flow vector, Extract the displacement information at the integer pixel level, where represents the rounding function, and Represents the optical flow vector after rounding;

[0040] S3.44: interpolate pixels around integer pixel displacements to estimate sub-pixel intensity values;

[0041] S3.45: Determine the sub-pixel displacement by a correlation search method based on the pixel intensity value obtained by interpolation, and output the obtained sub-pixel displacement information.

[0042] A river engineering model flow velocity intelligent measurement device, comprising: a power supply and control system, an auxiliary light source, a display, an image acquisition module, a tracer particle identification and marking module, a flow velocity calculation module, a flow velocity field reconstruction module, and a visual analysis module;

[0043] The power supply and control system is used to provide a stable power supply for the entire measuring device and to monitor and control the working status of each module in real time;

[0044] Auxiliary light source, used to provide uniform lighting conditions;

[0045] A display for displaying dynamic graphics of a three-dimensional flow velocity field;

[0046] An image acquisition module is used to capture a model flow field image sequence of a river engineering model test area through a high-speed camera;

[0047] The tracer particle identification and marking module is used to automatically detect, track and mark the tracer particles on the pre-processed image using an image recognition algorithm to obtain the position and motion trajectory information of the particles;

[0048] The velocity calculation module is used to calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles, the image acquisition frame rate and the model scale, and to improve the accuracy of the displacement calculation by using the optical flow method;

[0049] The velocity field reconstruction module is used to perform spatial interpolation processing on the calculated velocity data and reconstruct the three-dimensional velocity field in the river engineering model test area using radial basis functions;

[0050] The visualization analysis module is used to display the reconstructed three-dimensional velocity field in a graphical form and provide data analysis functions.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention proposes an intelligent flow velocity measurement device for river engineering models, which integrates multiple functional modules such as image acquisition, processing, calculation, storage and display, realizes the intelligent and automatic flow velocity measurement of river engineering models, is simple and convenient to operate, reduces manual intervention, and reduces measurement errors.

[0053] 2. The present invention proposes an intelligent method for measuring the flow velocity of a river engineering model. By acquiring and processing an image sequence of a river engineering model test area, and automatically identifying and marking tracer particles, the method realizes non-contact, high-precision measurement of the flow field of the river engineering model, which not only avoids the errors and interferences that may be caused by traditional physical contact measurement methods, but also improves the degree of automation and efficiency of the measurement. By utilizing the displacement information and image acquisition frame rate of the tracer particles and combining the scale of the river engineering model, the instantaneous velocity of the tracer particles can be accurately calculated, and the three-dimensional flow velocity field can be reconstructed through spatial interpolation and radial basis functions, providing rich flow velocity data. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the method of the present invention;

[0055] Figure 2 A flow chart of automatic identification and labeling of tracer particles of the present invention;

[0056] Figure 3 This is a module architecture diagram of the river model flow velocity intelligent measurement device of the present invention. DETAILED DESCRIPTION Example 1

[0057] See also Figure 1 and Figure 2 The present invention provides an embodiment: a method for intelligently measuring flow velocity of a river engineering model, comprising the following steps:

[0058] Step S1: obtaining a model flow field image sequence of a river engineering model test area and performing preprocessing;

[0059] The method of obtaining the model flow field image sequence of the river engineering model test area is to arrange a high-speed camera and an auxiliary light source in the river engineering model test area, and adopt a multi-angle synchronous shooting technology to obtain a multi-view image sequence of the model flow field to cover the complete flow field information of the test area.

[0060] Step S2: using an image recognition algorithm to automatically identify and mark tracer particles on the preprocessed model flow field image;

[0061] Step S3: Calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles in the continuous time period, combined with the image acquisition frame rate and the scale of the river engineering model;

[0062] Step S4: Reconstruct the three-dimensional velocity field in the river engineering model test area according to the velocity data of known tracer particles using radial basis functions.

[0063] Furthermore, the specific steps of step S4 include:

[0064] (1) Collect the velocity data of known tracer particles in the river engineering model test area, including the position coordinates of the particles and the corresponding velocity values;

[0065] (2) Using the known position and velocity data of tracer particles, a radial basis function interpolation model is constructed. ,in, represents the function to be interpolated, z represents the position to be predicted, represents the coefficients of the radial basis function, represents the radial basis function, express z and The Euclidean distance between represents the position of the known tracer particle, represents a low-order polynomial, M represents the number of known tracer particles;

[0066] (3) By Solve the linear equations and determine the coefficients of the radial basis functions, where A represents the coefficient matrix consisting of the radial basis functions and the polynomials. w represents the coefficient vector to be solved;

[0067] (4) Use the constructed radial basis function interpolation model to predict the flow velocity at any location within the river engineering model test area;

[0068] (5) Based on the prediction results, the three-dimensional velocity field is reconstructed.

[0069] Exemplarily, the high-resolution camera and auxiliary light source in the image acquisition module are installed at the corresponding position of the river engineering model test area according to the design requirements, and are debugged to ensure that the camera can normally capture clear images and the auxiliary light source can provide sufficient lighting; before the river engineering model test begins, tracer particles are evenly released into the water flow; the measuring device is started, and the image acquisition module is remotely operated through the power supply and control system to collect images of the river engineering model test area according to the predetermined time interval and shooting angle; the image acquisition module transmits the collected images to the tracer particle identification and marking module in real time, and the module uses the built-in image processing software and flow velocity calculation program to process and analyze the images, and calculate the flow velocity at the location of each tracer particle; after calculating the flow velocity at the location of each tracer particle, the flow velocity data is transmitted to the data storage and display module for preliminary storage, and the preprocessing operation mentioned in step S1 is performed; the image recognition algorithm based on deep learning is used to preprocess the preprocessed The tracer particles are automatically identified and marked based on the model flow field image after the model is retrieved, and the model is trained with the labeled tracer particle images so that it can accurately distinguish the tracer particles from the background; the displacement information of the tracer particles in a continuous time period is combined, the image acquisition frame rate and the model scale are considered, and the instantaneous velocity of the tracer particles is calculated using the velocity calculation formula, and then the velocity at the location of each tracer particle is determined; these velocity data are then reintegrated into the data storage and display module for more comprehensive data storage; finally, the stored velocity data are spatially interpolated, and the data blanks are filled with a suitable interpolation algorithm to make the data distribution more uniform, and then the radial basis function is used to reconstruct the three-dimensional velocity field in the river engineering model test area, presenting the velocity conditions of the entire test area in an intuitive and three-dimensional manner, which is convenient for researchers to deeply analyze the water flow characteristics. At the same time, the display terminal displays the reconstructed three-dimensional velocity field and key velocity parameters in real time, and researchers can further carry out river engineering model test research based on these data.

[0070] The method of automatically identifying and marking the tracer particles in the preprocessed model flow field image using an image recognition algorithm includes:

[0071] S2.1: Obtain a preprocessed model flow field image sequence, and use a deep learning-based image recognition algorithm to detect tracer particles in the model flow field image;

[0072] Furthermore, the specific steps of S2.1 include:

[0073] (1) Obtain the preprocessed model flow field image sequence, standardize the model flow field image sequence, and adjust the image size and pixel value range;

[0074] (2) Loading a pre-built target detection model and preparing a labeled training data set, including the position and category information of the tracer particles. The target detection model uses the YOLO model. The YOLO model is a prior art in the art and is not an inventive solution of the present application, and is not described in detail herein.

[0075] (3) Use the labeled training data set to train the target detection model, optimize the target detection model parameters, and generate a trained target detection model;

[0076] (4) Input the preprocessed model flow field image sequence into the trained target detection model to detect the tracer particles and obtain the output results of the target detection model, including the position coordinates and confidence scores of the tracer particles.

[0077] S2.2: Based on the tracer particle identification results, initially mark the tracer particles and record the initial coordinates of each particle and characteristic information. In the present invention, the initial marking of the tracer particles is achieved by assigning a unique identifier to each tracer particle on the continuous frame images, so as to accurately track and identify these particles in the subsequent analysis and processing;

[0078] S2.3: Set the particles to move at a uniform speed, combine the position information of the tracer particles in the continuous frame images, and use the adaptive threshold segmentation method to track the movement trajectory of each particle;

[0079] S2.4: According to the tracking results, dynamically update the position information of each particle and calculate its speed, while keeping the unique tag ID of the particle unchanged;

[0080] Furthermore, the specific steps of S2.4 include:

[0081] (1) Obtain the current frame position information of each particle;

[0082] (2) Matching the position information of the current frame with the position information of the previous frame based on the particle’s unique tag ID ensures that the identity of the particle can be continuously tracked even if the particle moves in the image;

[0083] (3) For each matched particle, update its position record with the position information of the current frame, usually storing or updating a data structure such as a list, array, or database, where each entry contains the particle ID, position coordinates, and speed;

[0084] (4) Calculate the velocity of the particle based on the change in its position between two consecutive frames. For example, if the tracer particle x and y The positions in the directions are and , and the inter-frame time is , then the speed It is expressed as: ,in, and Respectively represent the tracer particles in x , y The component of velocity in the direction.

[0085] S2.5: Based on the current position and velocity information of the particle, the particle tracking algorithm is used to predict the particle motion trajectory;

[0086] Furthermore, the specific steps of S2.5 include:

[0087] (1) Collect the current position and velocity information of the tracer particles;

[0088] (2) Initialize the parameters required by the particle tracking algorithm, such as time step and number of prediction steps;

[0089] (3) Using the current velocity of the particle and the predicted time step, the predicted position of the particle at the next time step is calculated according to the kinematic formula, wherein the kinematic formula is used to predict the position of the particle at the next time step. The specific calculation formula of the kinematic formula is the prior art content in this field, which is not the inventive solution of the present application and is not elaborated here.

[0090] S2.6: According to the difference between the predicted result and the actual detection result, the particle motion trajectory is corrected using the least square method, wherein the least square method is a prior art content in this field, is not an inventive solution of the present application, and is not described in detail here;

[0091] S2.7: Match the corrected tracer particle motion trajectories, and integrate the motion trajectory information of all particles to generate a complete tracer particle motion data set; the complete tracer particle motion data set is a triplet data set including the tracer particle position, velocity and trajectory information.

[0092] The specific steps of S2.3 include:

[0093] S2.31: Use the camera to obtain continuous frame images containing tracer particles at a preset frequency and method, and perform preprocessing to generate a model flow field image sequence of a continuous time period ,in, represents the first n Model flow field images, n Indicates the number of model flow field images in the model flow field image sequence;

[0094] S2.32: Segment the model flow field image using an adaptive threshold segmentation method based on a local mean to obtain a segmented binary image, wherein the adaptive threshold segmentation method based on a local mean is a prior art in the art and is not an inventive solution of the present application, and is not described in detail herein;

[0095] In the present invention, the pixel values ​​of the segmented binary image have only two possible states, 0 and 1, which correspond to pure black and pure white.

[0096] S2.33: identifying the position of the tracer particles in the segmented binary image by a contour recognition method, wherein the contour recognition method is a prior art in the art and is not an inventive solution of the present application, and is not described in detail herein;

[0097] S2.34: combining the position information of the tracer particles in the continuous frame images, using nearest neighbor matching to track the motion trajectory of each particle, wherein nearest neighbor matching is a prior art content in this field, is not an inventive solution of the present application, and is not described in detail here;

[0098] S2.35: Output the tracked particle motion trajectory.

[0099] The specific steps of step S3 include:

[0100] S3.1: Model flow field image sequences from consecutive time periods Extract the position information of the tracer particles, and obtain the coordinates of the tracer particles in each frame of the model flow field image;

[0101] Furthermore, the specific process of extracting the position information of the tracer particles from the model flow field image sequence of continuous time periods includes:

[0102] (1) Preprocessing the model flow field image sequence of continuous time periods, including denoising and contrast enhancement, to improve the recognition accuracy of tracer particles;

[0103] (2) Using deep learning-based target detection algorithms to identify tracer particles in the preprocessed model flow field images, including:

[0104] The image is divided into small blocks, and each small block is classified to determine whether it contains tracer particles, wherein the target detection algorithm based on deep learning is the prior art content in this field, and is not the inventive solution of the present application, and will not be described in detail here;

[0105] (3) For each identified tracer particle, mark the position of the tracer particle;

[0106] (4) Record the pixel coordinates of each identified tracer particle and associate it with the corresponding timestamp;

[0107] (5) Store the extracted location information in a database or file.

[0108] S3.2: Obtain the frame rate of the model flow field image acquisition of the high-speed camera f and model scale S;

[0109] S3.3: Calculate the displacement of the tracer particles between adjacent frames according to the coordinates in each frame of the model flow field image. ,in, , and They respectively represent the position coordinates of the tracer particles in the next frame of image;

[0110] S3.4: Use the optical flow method to calculate the displacement with sub-pixel accuracy to obtain the displacement information after sub-pixel accuracy calculation ;

[0111] S3.5: Use formula Convert the displacement information calculated with sub-pixel accuracy into actual displacement , and combined with the time interval , based on the formula Calculate the instantaneous velocity of the tracer particle ,in, and Respectively represent the tracer particles in each frame of the model flow field image x and y The velocity component in the direction;

[0112] S3.6: Use formula The error correction is performed on the calculated instantaneous speed, where: represents the final flow rate after error correction and optimization, represents the polynomial coefficients, m represents the polynomial order, j represents the polynomial order index, represents the salinity correction factor, represents the error in flow velocity measurement caused by salinity changes, represents the noise correction coefficient, Indicates the flow rate measurement error caused by noise interference;

[0113] S3.7: Based on the corrected instantaneous speed , output the instantaneous velocity of each tracer particle ,and ,in, represents the direction of instantaneous velocity, represents the inverse tangent function, and They represent the tracer particles in each frame of the model flow field image after error correction. x and y The component of velocity in the direction.

[0114] In the present invention, because the flow velocity is a vector with two attributes, magnitude and direction, a single value can only represent the magnitude of the velocity but cannot describe the direction information. The form of is because when performing flow field analysis, it is necessary to use a rectangular coordinate system, the form is , which is convenient for calculating physical quantities such as the gradient, divergence and curl of the flow field. When visualizing it, polar coordinates are usually used, in the form of , which is convenient for drawing, so it is stored as form.

[0115] The specific steps of S3.4 include:

[0116] S3.41: Acquisition and the displacement of the tracer particles between adjacent frames ;

[0117] S3.42: Use Calculate the optical flow vector of each pixel in the model flow field image sequence ,in, , and Respectively represent the model flow field images in x , y , t The gradient in direction, t Indicates the current moment, u , v Represents the optical flow vector x , y Quantity, and represents the weight coefficient of the spatial smoothness constraint, , , and Respectively u , v exist x , y The second derivative in the direction, and They represent the noise weight coefficients, and Respectively u , v Directional noise;

[0118] It should be noted that the formula in the present invention comprehensively considers the spatial and temporal smoothness of the optical flow, as well as the influence of noise and uncertainty on the optical flow calculation. The spatial smoothness of the optical flow field is enhanced by introducing the spatial second-order derivative term, so that the optical flow vectors of adjacent pixels are more consistent, reducing the mutation in the optical flow field. The addition of the noise term enables the formula to better handle the noise in the image data, thereby improving the robustness of the optical flow calculation.

[0119] S3.43: From the calculated optical flow vector, Extract the displacement information at the integer pixel level, where represents the rounding function, and Represents the optical flow vector after rounding;

[0120] S3.44: interpolating pixels around the integer pixel displacement to estimate the intensity value at the sub-pixel level. In the present invention, a bilinear interpolation method is used for interpolation. The bilinear interpolation method is a prior art in the art and is not an inventive solution of the present application, and is not described in detail here.

[0121] S3.45: According to the pixel intensity value obtained by interpolation, the displacement at the sub-pixel level is determined by a correlation search method, and the obtained sub-pixel displacement information is output, wherein the correlation search method is the prior art content in this field, is not the creative solution of the present application, and is not elaborated here. Example 2

[0122] See also Figure 3 , another embodiment provided by the present invention: a river engineering model flow velocity intelligent measurement device, comprising: a power supply and control system, an auxiliary light source, a display, an image acquisition module, a tracer particle identification and marking module, a flow velocity calculation module, a flow velocity field reconstruction module, and a visualization analysis module;

[0123] The power supply and control system is used to provide a stable power supply for the entire measuring device and to monitor and control the working status of each module in real time;

[0124] Auxiliary light source, used to provide uniform lighting conditions;

[0125] A display for displaying dynamic graphics of a three-dimensional flow velocity field;

[0126] An image acquisition module is used to capture a model flow field image sequence of a river engineering model test area through a high-speed camera;

[0127] The tracer particle identification and marking module is used to automatically detect, track and mark the tracer particles in the pre-processed image using an image recognition algorithm to obtain the position and motion trajectory information of the particles;

[0128] The velocity calculation module is used to calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles, the image acquisition frame rate and the model scale, and to improve the accuracy of the displacement calculation by using the optical flow method;

[0129] The velocity field reconstruction module is used to perform spatial interpolation processing on the calculated velocity data and reconstruct the three-dimensional velocity field in the river engineering model test area using radial basis functions;

[0130] The visualization analysis module is used to display the reconstructed three-dimensional velocity field in graphical form and provide data analysis functions.

[0131] Further, the tracer particle identification and labeling module includes: a tracer particle detection unit, a particle tracking unit, and a dynamic labeling unit;

[0132] A tracer particle detection unit, used to detect tracer particles in an image using a deep learning algorithm;

[0133] The particle tracking unit tracks the tracer particles in real time by combining an adaptive threshold segmentation algorithm;

[0134] The dynamic marking unit is used to dynamically mark the tracked particles and record the position and motion trajectory information of the particles.

[0135] The flow rate calculation module includes: a displacement calculation unit, an optical flow calculation unit, a speed calculation unit, and an error correction unit;

[0136] A displacement calculation unit, used to calculate the pixel displacement of the particle according to the displacement information of the tracer particle in a continuous time period;

[0137] An optical flow calculation unit, used to calculate displacement with sub-pixel accuracy using the optical flow method, thereby improving the resolution of displacement measurement;

[0138] The velocity calculation unit calculates the instantaneous velocity of the tracer particles by combining the image acquisition frame rate and the model scale;

[0139] The error correction unit is used to perform error correction on the speed data to eliminate errors caused by factors such as image distortion and noise.

[0140] The velocity field reconstruction module includes: spatial interpolation unit, radial basis function calculation unit, and data fusion unit;

[0141] A spatial interpolation unit is used to perform spatial interpolation on discrete tracer particle velocity data to generate continuous velocity field data;

[0142] A radial basis function calculation unit is used to optimize the interpolation result using the radial basis function and reconstruct a high-precision three-dimensional velocity field;

[0143] The data fusion unit is used to combine the velocity field data with the model terrain data to generate a velocity field model with spatial distribution characteristics.

[0144] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. A method for intelligently measuring flow velocity of a river model, characterized in that: include: Step S1: obtaining a model flow field image sequence of a river engineering model test area and performing preprocessing; Step S2: using an image recognition algorithm to automatically identify and mark tracer particles on the preprocessed model flow field image; Step S3: Calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles in the continuous time period, combined with the image acquisition frame rate and the scale of the river engineering model; Step S4: performing spatial interpolation processing on the calculated tracer particle instantaneous velocity data, and using radial basis functions to reconstruct the three-dimensional velocity field in the river engineering model test area according to the known velocity data of the tracer particles; The specific steps of step S3 include: S3.1: Model flow field image sequences from consecutive time periods The position information of the tracer particles is extracted to obtain the coordinates of the tracer particles in each frame of the model flow field image, where: represents the first n Model flow field images, n Indicates the number of model flow field images in the model flow field image sequence; S3.2: Obtain the frame rate of the model flow field image acquisition of the high-speed camera f and model scale S; S3.3: Calculate the displacement of the tracer particles between adjacent frames based on the coordinate changes in each frame of the model flow field image. ; S3.4: Use the optical flow method to calculate the displacement with sub-pixel accuracy to obtain the displacement information after sub-pixel accuracy calculation ; S3.5: Use formula Convert the displacement information calculated with sub-pixel accuracy into actual displacement , and combined with the time interval , based on the formula Calculate the instantaneous velocity of the tracer particle ,in, and Respectively represent the tracer particles in each frame of the model flow field image x and y The velocity component in the direction; S3.6: Use formula The error correction is performed on the calculated instantaneous speed, where: represents the final flow rate after error correction and optimization, represents the polynomial coefficients, m represents the polynomial order, j represents the polynomial order index, represents the salinity correction factor, represents the error in flow velocity measurement caused by salinity changes, represents the noise correction coefficient, Indicates the flow rate measurement error caused by noise interference; S3.7: Based on the corrected instantaneous speed , output the instantaneous velocity of each tracer particle ,in, represents the direction of instantaneous velocity, and They represent the tracer particles in each frame of the model flow field image after error correction. x and y The component of velocity in the direction.

2. A method for intelligently measuring flow velocity of a river model as claimed in claim 1, characterized in that: The model flow field image sequence of the river engineering model test area is obtained by: High-speed cameras and auxiliary light sources are arranged in the river engineering model test area, and multi-angle synchronous shooting technology is used to obtain multi-perspective image sequences of the model flow field to cover the complete flow field information of the test area.

3. A method for intelligently measuring flow velocity of a river model as claimed in claim 2, characterized in that: The method of automatically identifying and marking the tracer particles in the preprocessed model flow field image using an image recognition algorithm includes: S2.1: Obtain a preprocessed model flow field image sequence, and use a deep learning-based image recognition algorithm to detect tracer particles in the model flow field image; S2.2: Based on the tracer particle identification results, initially mark the tracer particles and record the initial coordinates of each particle and feature information; S2.3: Set the particles to move at a constant speed, combine the position information of the tracer particles in the continuous frame images, and use the adaptive threshold segmentation method to track the movement trajectory of each particle.

4. A method for intelligently measuring flow velocity of a river model as claimed in claim 3, characterized in that: The method of automatically identifying and marking the tracer particles in the preprocessed model flow field image by using an image recognition algorithm also includes: S2.4: According to the tracking results, dynamically update the position information of each particle and calculate its speed, while keeping the unique tag ID of the particle unchanged; S2.5: Predict the particle motion trajectory based on the current position and velocity information of the particle; S2.6: Correcting the particle motion trajectory according to the difference between the predicted result and the actual detection result; S2.7: Match the corrected tracer particle motion trajectories, and integrate the motion trajectory information of all particles to generate a complete tracer particle motion data set; the complete tracer particle motion data set is a triplet data set including the tracer particle position, velocity and trajectory information.

5. The intelligent method for measuring flow velocity of a river model according to claim 4, characterized in that: The specific steps of S2.3 include: S2.31: Use the camera to obtain continuous frame images containing tracer particles at a preset frequency and method, and perform preprocessing to generate a model flow field image sequence of a continuous time period ; S2.32: Segment the model flow field image using an adaptive threshold segmentation method based on the local mean to obtain a segmented binary image; S2.33: Identify the position of the tracer particles in the segmented binary image; S2.34: Combine the position information of the tracer particles in the continuous frame images to track the movement trajectory of each particle; S2.35: Output the tracked particle motion trajectory.

6. A method for intelligently measuring flow velocity of a river model as claimed in claim 5, characterized in that: The specific steps of S3.4 include: S3.41: Acquisition and the displacement of the tracer particles between adjacent frames ; S3.42: Use Calculate the optical flow vector of each pixel in the model flow field image sequence ,in, , and Respectively represent the model flow field images in x , y , t The gradient in direction, t Indicates the current moment, u , v Represents the optical flow vector x , y Quantity, and represents the weight coefficient of the spatial smoothness constraint, , , and Respectively u , v exist x , y The second derivative in the direction, and They represent the noise weight coefficients, and Respectively u , v Directional noise; S3.43: From the calculated optical flow vector, Extract the displacement information at the integer pixel level, where represents the rounding function, and Represents the optical flow vector after rounding; S3.44: interpolate pixels around integer pixel displacements to estimate sub-pixel intensity values; S3.45: Determine the sub-pixel displacement by a correlation search method based on the pixel intensity value obtained by interpolation, and output the obtained sub-pixel displacement information.

7. A river model flow velocity intelligent measurement device, which is used to implement a river model flow velocity intelligent measurement method according to any one of claims 1 to 6, characterized in that: include: Power supply and control system, auxiliary light source, display, image acquisition module, tracer particle identification and marking module, flow velocity calculation module, flow velocity field reconstruction module, and visualization analysis module; The power supply and control system is used to provide a stable power supply for the entire measuring device and to monitor and control the working status of each module in real time; Auxiliary light source, used to provide uniform lighting conditions; A display for displaying dynamic graphics of a three-dimensional flow velocity field; An image acquisition module is used to capture a model flow field image sequence of a river engineering model test area through a high-speed camera; The tracer particle identification and marking module is used to automatically detect, track and mark the tracer particles on the pre-processed image using an image recognition algorithm to obtain the position and motion trajectory information of the particles; The velocity calculation module is used to calculate the instantaneous velocity of the tracer particles according to the displacement information of the tracer particles, the image acquisition frame rate and the model scale, and to improve the accuracy of the displacement calculation by using the optical flow method; The velocity field reconstruction module is used to perform spatial interpolation processing on the calculated velocity data and reconstruct the three-dimensional velocity field in the river engineering model test area using radial basis functions; The visualization analysis module is used to display the reconstructed three-dimensional velocity field in a graphical form and provide data analysis functions.

Images