A method, device, apparatus and device for visual measurement of a maglev wind tunnel test model

Abstract

This application discloses a method, device, apparatus, and apparatus for visually measuring a maglev wind tunnel test model, relating to the field of maglev wind tunnels. The method comprises: a measurement system acquiring images of the test model taken synchronously by two high-speed cameras using an improved genetic algorithm-optimized BP neural network calibration; extracting the pixel coordinates of several feature points from each test model image; determining several groups of corresponding feature points based on the relative position information of the feature points in the corresponding test model images; the corresponding relationships characterizing that feature points in the same group correspond to the same point on the test model; calculating the three-dimensional world coordinates of the test model based on the pixel coordinates of several groups of corresponding feature points using the principles of parallax and triangulation, and determining the velocity and attitude information of the test model based on the three-dimensional world coordinates. This application utilizes a measurement system constructed using high-speed cameras to achieve non-contact, high-precision measurement of a high-speed moving test model.

Description

Technical Field

[0001] The present invention relates to the field of maglev wind tunnels, and in particular to a method, device, equipment and apparatus for visually measuring a maglev wind tunnel test model. Background Art

[0002] The design concept of the maglev wind tunnel combines vacuum tube maglev train and dynamic model testing technologies. By creating a "body moving, wind-quiet" test state that resembles a real flight environment and motion characteristics, it meets the aerodynamic and interdisciplinary ground testing requirements for aerospace vehicles and high-speed trains over a wide range of Mach and Reynolds numbers, low noise, and low turbulence. Its key features are: first, "body moving, wind-quiet," with the moving object moving at supersonic speeds within the tube; second, non-contact, high-precision measurement requirements, requiring accurate position, velocity changes, and attitude information of the moving object.

[0003] Existing wind tunnel visual measurement technology is mostly used in ordinary wind tunnels, and the camera frame rates used are often below 1000 FPS (Frames Per Second). However, due to the high speed of the suspended rotors and the high measurement accuracy required in maglev wind tunnels, traditional wind tunnel visual measurement technology is not suitable for high-speed moving objects and cannot meet the high-precision measurement requirements. Secondly, when identifying features of high-speed moving objects, the addition of feature recognition objects may affect the aerodynamic performance of the high-speed moving object, and the impact on aerodynamic performance may be amplified at high speeds. Summary of the Invention

[0004] In view of this, the present invention aims to provide a method, device, apparatus, and device for visual measurement of a maglev wind tunnel test model. These methods, based on an improved genetic algorithm and optimized BP neural network, can calibrate a high-speed camera and utilize a high-speed camera-based measurement system to achieve non-contact, high-precision measurement of the high-speed moving test model. The specific implementation is as follows:

[0005] In a first aspect, the present application provides a visual measurement method for a maglev wind tunnel test model, which is applied to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously capture the same area in the maglev wind tunnel; wherein the method includes:

[0006] Acquire images of the test model taken synchronously by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on an improved genetic algorithm to optimize a BP neural network;

[0007] Feature points are extracted from each of the test model images to obtain pixel coordinates of a plurality of feature points, and a plurality of groups of feature points having corresponding relationships are determined based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicate that the feature points in the same group correspond to the same spatial point on the test model; the test model is a model mounted on a magnetically suspended mover that moves at high speed within the wind tunnel duct;

[0008] The three-dimensional world coordinates of the test model are calculated based on the pixel coordinates of the groups of corresponding feature points using the parallax principle and the triangulation principle, and the speed information and posture information of the test model are determined based on the three-dimensional world coordinates.

[0009] Optionally, before obtaining the calibrated test model images synchronously captured by the two high-speed cameras, the method further includes:

[0010] Aligning the coordinates of the infrared camera and the two high-speed cameras to establish a world coordinate system;

[0011] Accordingly, after obtaining the calibrated test model images synchronously photographed by the two high-speed cameras, the method further includes:

[0012] The test model images synchronously photographed by the two calibrated high-speed cameras are stored in the corresponding storage device.

[0013] Optionally, before obtaining the calibrated test model images synchronously captured by the two high-speed cameras, the method further includes:

[0014] Acquire a plurality of checkerboard images taken from different angles, and preprocess each of the checkerboard images to obtain a preprocessed image; the preprocessing includes grayscale transformation, Gaussian filtering, and image reduction processing based on a multi-grid algorithm;

[0015] Performing corner point detection on each of the preprocessed images, and matching the detected corner points with the same name based on grid motion statistics to obtain a plurality of corner points with the same name;

[0016] Converting the two-dimensional pixel coordinates of each corner point with the same name into corresponding three-dimensional world coordinates using the least squares method and based on a binocular vision model to obtain coordinate information corresponding to each corner point with the same name;

[0017] Constructing a corner point set based on coordinate information corresponding to each of the corner points with the same name, and dividing the corner point set into a training set and a test set;

[0018] Utilizing the training set and based on the improved genetic algorithm to perform optimization training on the BP neural network, so as to obtain a BP neural network optimized based on the improved genetic algorithm;

[0019] The BP neural network optimized based on the improved genetic algorithm is tested using the test set to obtain optimized world coordinates corresponding to the corner points with the same name in the test set, and the high-speed camera is calibrated based on the optimized world coordinates.

[0020] Optionally, extracting feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points includes:

[0021] Performing Gaussian filtering on each of the test model images to obtain each processed model image, and identifying a plurality of feature markers from each of the processed model images;

[0022] Feature points are extracted from each of the feature marks to obtain pixel coordinates of several feature points.

[0023] Optionally, extracting feature points from each of the feature markers to obtain pixel coordinates of several feature points includes:

[0024] respectively determining all pixel coordinates within each of the feature marks and the grayscale values ​​corresponding to all the pixel coordinates;

[0025] Based on all pixel coordinates in each of the feature marks and the grayscale values ​​respectively corresponding to all of the pixel coordinates, the pixel coordinates of the center point in each of the feature marks are calculated to obtain the pixel coordinates of several feature points.

[0026] Optionally, the calculating of the corresponding three-dimensional world coordinates of the test model based on the pixel coordinates of the plurality of groups of feature points having corresponding relationships by using the parallax principle and the triangulation principle includes:

[0027] Utilizing a pre-established world coordinate calculation formula based on the parallax principle and the triangulation principle, the pixel coordinates of any set of corresponding feature points are calculated to obtain the three-dimensional world coordinates of the test model;

[0028] Before calculating the pixel coordinates of any set of corresponding feature points using the pre-constructed world coordinate calculation formula based on the parallax principle and the triangulation principle, the method further includes:

[0029] Obtaining the focal length of the high-speed camera and the distance between the optical centers of the two high-speed cameras; the focal length and field of view angle of the two high-speed cameras are the same;

[0030] The world coordinate calculation formula is constructed based on the focal length and the distance by utilizing the parallax principle and the triangulation principle.

[0031] Optionally, the method further includes:

[0032] Obtaining the original temperature of the test model measured by the infrared camera;

[0033] An angle between the infrared camera and the test model is determined, and the original temperature is adjusted based on the angle and a coefficient related to the wavelength of electromagnetic radiation to obtain temperature data of the test model.

[0034] In a second aspect, the present invention provides a visual measurement device for a maglev wind tunnel test model, which is applied to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously capture the same area in the maglev wind tunnel; wherein the device includes:

[0035] An image acquisition module is used to acquire images of the test model taken synchronously by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on a BP neural network optimized by an improved genetic algorithm;

[0036] a correspondence determination module for extracting feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points, and determining a plurality of groups of feature points having corresponding relationships based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicating that the feature points in the same group correspond to the same spatial point on the test model; the test model being a model mounted on a magnetically suspended mover that moves at high speed within the wind tunnel duct;

[0037] The information determination module is used to calculate the three-dimensional world coordinates of the corresponding test model based on the pixel coordinates of the plurality of groups of corresponding feature points by utilizing the parallax principle and the triangulation principle, and to determine the speed information and posture information of the test model based on the three-dimensional world coordinates.

[0038] In a third aspect, an electronic device is provided, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to execute the computer program to implement the aforementioned visual measurement method for the magnetic levitation wind tunnel test model.

[0041] In a fourth aspect, a computer-readable storage medium is provided for storing a computer program, which, when executed by a processor, implements the aforementioned visual measurement method for a magnetic levitation wind tunnel test model.

[0042] The present invention provides a method for visually measuring a maglev wind tunnel test model applied to a measurement system; the measurement system comprises an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each of the high-speed cameras; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel duct and synchronously photograph the same area in the maglev wind tunnel; wherein the method comprises: obtaining images of the test model synchronously photographed by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on an improved genetic algorithm to optimize the BP neural network for the high-speed cameras; and performing feature point tracing on each of the test model images. Extracting to obtain pixel coordinates of several feature points, determining several groups of feature points with corresponding relationships based on the relative position information of the feature points in the corresponding test model image; the corresponding relationship indicates that the feature points in the same group correspond to the same spatial point on the test model; the test model is a model mounted on a magnetic levitation mover that moves at high speed in the wind tunnel duct; using the parallax principle and the triangulation principle and based on the pixel coordinates of the several groups of feature points with corresponding relationships, calculating the three-dimensional world coordinates of the corresponding test model, and determining the velocity information and posture information of the test model based on the three-dimensional world coordinates. It can be seen that the present application calibrates the high-speed camera based on the improved genetic algorithm to optimize the BP neural network, thereby meeting the high-precision and high-resolution requirements of the high-speed camera and improving the measurement accuracy of the high-speed moving test model; and the present application realizes non-contact measurement of the high-speed moving test model through a measurement system constructed based on the high-speed camera, avoiding the problem that the installation of sensors on the high-speed moving test model easily has a negative impact on the aerodynamic performance and motion characteristics of the model, improving the measurement efficiency of the high-speed moving test model, and ensuring the reliability of the test model data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0044] Figure 1 This is a flow chart of a visual measurement method for a maglev wind tunnel test model disclosed in this application;

[0045] Figure 2 A high-speed camera layout diagram disclosed in this application;

[0046] Figure 3 A horizontal field of view diagram of two high-speed cameras disclosed in this application;

[0047] Figure 4 A schematic diagram of the distance measurement principle of two high-speed cameras disclosed in this application;

[0048] Figure 5 A high-speed camera calibration flow chart disclosed in this application;

[0049] Figure 6 A specific high-speed camera calibration flow chart disclosed in this application;

[0050] Figure 7 This is a schematic structural diagram of a visual measurement device for a magnetic levitation wind tunnel test model disclosed in this application;

[0051] Figure 8 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] Due to the high speed of the suspended movers and the high measurement accuracy required in maglev wind tunnels, traditional wind tunnel visual measurement technology is not suitable for high-speed moving objects and cannot meet the requirements of high-precision measurement. To this end, this application provides a visual measurement method for maglev wind tunnel test models. It can calibrate a high-speed camera based on an improved genetic algorithm to optimize the BP neural network. It also uses a measurement system built with high-speed cameras to achieve non-contact, high-precision measurement of high-speed moving test models.

[0054] See also Figure 1 As shown, an embodiment of the present invention discloses a visual measurement method for a maglev wind tunnel test model, which is applied to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously capture the same area in the maglev wind tunnel; wherein the method includes:

[0055] Step S11, obtaining images of the test model synchronously captured by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on an improved genetic algorithm optimized BP neural network.

[0056] In this embodiment, the measurement system includes an infrared camera, two fill light sources, two high-speed cameras, and a high-speed image storage device configured for each high-speed camera, and all are placed on the inner wall of the wind tunnel pipe; among them, the infrared camera is used to achieve temperature measurement, and the two high-speed cameras are used to achieve binocular vision measurement. It should be noted that, if Figure 2 As shown in the figure, two high-speed cameras are placed at different positions on the inner wall of the wind tunnel and simultaneously shoot the same area in the maglev wind tunnel, thereby realizing binocular vision measurement. Figure 3 As shown, the focal lengths of high-speed cameras 1 and 2 are consistent, for example, 10mm; their fields of view are also consistent, and the distance between them can also be set. This allows for non-contact camera measurement, eliminating the need to install sensors on the object being measured, thus negatively impacting its aerodynamic performance and motion characteristics. This ensures objective, visual, and reliable measurement results, significantly improving measurement efficiency and accuracy. Furthermore, camera measurement offers the advantages of ease of assembly and disassembly and a compact footprint.

[0057] It should be noted that multiple measurement systems are installed throughout the wind tunnel duct, and these systems are connected via transmission lines to enable data transmission between them. Furthermore, the measurement ranges of the multiple measurement systems need to cover the entire wind tunnel duct to avoid blind spots within the duct. Simultaneously, the multiple measurement systems work collaboratively to achieve visual and temperature measurements of a test model moving at high speed within the wind tunnel duct. The test model is mounted on a magnetically levitated mover that moves at high speed within the wind tunnel duct, and the test model is pre-marked with visual feature markers. For example, the visual feature markers can be reflective markers, and the shapes of the reflective markers include circles, quadrilaterals, triangles, etc., which are not limited here. Furthermore, different arrays of visual feature markers can be set on the test model depending on the test model, and the number of visual feature markers needs to be at least three. In this way, when reflective markers are used as visual feature markers, the combination of a fill light source and reflective markers can better highlight the reflective markers, thereby improving the accuracy of subsequent visual feature marker recognition.

[0058] In this embodiment, using any measurement system within a wind tunnel as an example, before the measurement system uses two calibrated high-speed cameras to capture the same area within the wind tunnel, the coordinates of the infrared camera and the two high-speed cameras must be aligned to establish a world coordinate system. Furthermore, the high-speed cameras must be calibrated with high precision using a BP (Back Propagation) neural network optimized using an improved genetic algorithm (GA). This GA-optimized BP neural network calibration of the high-speed cameras meets the high-precision and high-resolution requirements for high-speed cameras, improving the measurement accuracy of high-speed motion test models.

[0059] Furthermore, the measurement system uses two calibrated high-speed cameras to synchronously capture the same area within the wind tunnel, generating images of the test model. These images are then stored in separate high-speed image storage devices via a Camera Link data cable. Camera Link is a standard interface protocol for high-speed image data transmission, and can be configured in Full mode, achieving a maximum throughput of 680MB / s, meeting the image processing speed requirements of maglev wind tunnel testing.

[0060] Step S12: extracting feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points, and determining a plurality of groups of feature points having corresponding relationships based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicate that the feature points in the same group correspond to the same spatial point on the test model; the test model is a model mounted on a magnetically levitated mover that moves at high speed within the wind tunnel duct.

[0061] In this embodiment, to overcome image noise caused by low-light conditions, the test model images are preprocessed before feature point extraction. For example, each test model image is Gaussian filtered to produce a processed model image. Next, a number of feature markers are identified from each processed model image. Feature points are extracted from each feature marker to obtain a number of feature points and their pixel coordinates. This embodiment, combined with Gaussian filtering and supplemental light sources, can adapt to the low-light conditions of a maglev wind tunnel and achieve real-time calculation of the pixel coordinates of feature points.

[0062] Among them, the extraction of feature points is specifically to extract the center point of each feature marker based on the grayscale centroid method to obtain several feature points and the pixel coordinates of each feature point. Specifically, all pixel coordinates in each feature marker and the grayscale values ​​corresponding to all pixel coordinates are determined respectively; then, based on all pixel coordinates in each feature marker and the grayscale values ​​corresponding to all pixel coordinates, the pixel coordinates of the center point in each feature marker are calculated to obtain the pixel coordinates of several feature points, that is, the pixel coordinates of the center point in the feature marker are the pixel coordinates of the feature point. The formulas involved are as follows:

[0063] ;

[0064] Where Ω represents a feature marker; G(x,y) represents the grayscale value of the pixel at coordinate (x,y) within the feature marker Ω; It represents the pixel coordinates of the center point within the feature mark Ω, that is, the pixel coordinates of the feature point within the feature mark Ω.

[0065] In this embodiment, after obtaining the pixel coordinates of several feature points, the relative position information of each feature point in the corresponding test model image is determined, and based on the internal and external parameters of the two high-speed cameras and the relative position information of each feature point in the corresponding test model image, several groups of feature points with corresponding relationships are determined; wherein the corresponding relationship indicates that the feature points in the same group correspond to the same spatial point on the test model.

[0066] Step S13: Calculate the corresponding three-dimensional world coordinates of the test model based on the pixel coordinates of the plurality of groups of corresponding feature points using the parallax principle and the triangulation principle, and determine the speed information and posture information of the test model based on the three-dimensional world coordinates.

[0067] In this embodiment, a pre-established world coordinate calculation formula based on the principles of parallax and triangulation is used to calculate the pixel coordinates of any set of corresponding feature points to obtain the three-dimensional world coordinates of the test model. Velocity and posture information of the test model are then determined based on the three-dimensional world coordinates. It should be noted that after obtaining the velocity and posture information, vibration and acceleration information of the test model can be further extracted. This allows the two high-speed cameras in the measurement system to collect a variety of information about the test model.

[0068] Among them, for the construction of the world coordinate calculation formula, such as Figure 4As shown, obtain the focal length f of the high-speed camera and the distance d between the optical centers of the two high-speed cameras; it should be noted that the focal length and field of view of the two high-speed cameras are the same. Using the parallax principle and triangulation principle and based on the focal length f and distance d, a world coordinate calculation formula is constructed. Specifically, for a certain spatial point Q (x, y, z) on the test model, the point imaged by its projection on the two high-speed cameras is and In the image physical coordinate system, they are expressed as and , we can get the following formula from similar triangles:

[0069] ;

[0070] After stereo calibration, the imaging planes of the two high-speed cameras are coplanar and of equal height, so , and finally the three-dimensional world coordinates of the spatial point Q can be solved as shown below:

[0071] ;

[0072] That is, first, the pixel coordinates of any group of corresponding feature points are transformed to obtain the image physical coordinates of any group of corresponding feature points, and then the image physical coordinates of any group of corresponding feature points are substituted into the above-mentioned three-dimensional world coordinate calculation formula for calculation, so as to obtain the three-dimensional world coordinates of a certain spatial point on the test model. Further, based on the three-dimensional world coordinates, the speed information and posture information of the test model can be determined.

[0073] In this embodiment, in addition to using two high-speed cameras to perform binocular vision measurement to determine the speed information and posture information of the test model, temperature measurement can also be performed through an infrared camera to determine the temperature data of the test model. Considering that for a surface source target such as a test model, a simple change in distance will not affect the temperature measurement result, but the angle between the moving test model and the infrared camera changes at all times, which can easily cause the temperature measurement of the infrared camera to be offset. Therefore, it is necessary to establish a model for the influence of distance and angle on the accuracy of the infrared camera, and to correct and compensate the temperature data based on high-precision position data, so as to meet the needs of high-precision measurement. For a surface source target such as a test model, the distance and angle of the windward surface relative to the measurement system change at all times. When the test model has a certain angle with the infrared camera, the temperature value output by the infrared camera is linearly related to the cosine value of the angle, that is:

[0074] ;

[0075] Among them, I R(T) is the temperature function of the infrared camera, C is a constant, T is the original temperature measured by the infrared camera, n is a coefficient related to the wavelength of electromagnetic radiation, and θ is the angle between the infrared camera and the test model. Specifically, the original temperature of the test model measured by the infrared camera is obtained, and the angle between the infrared camera and the test model is determined. The angle and the original temperature of the test model measured by the infrared camera are input into the above formula to adjust the original temperature and obtain the temperature data of the test model.

[0076] It can be seen that the present application calibrates the high-speed camera based on the BP neural network optimized by the improved genetic algorithm, thereby meeting the high-precision and high-resolution requirements of the high-speed camera and improving the measurement accuracy of the high-speed motion test model; and, the present application realizes the contactless measurement of the high-speed motion test model through the measurement system constructed based on the high-speed camera, avoiding the problem of installing sensors on the high-speed motion test model, which easily has a negative impact on the aerodynamic performance and motion characteristics of the model, improving the measurement efficiency of the high-speed motion test model, and ensuring the reliability of the test model data.

[0077] Based on the previous embodiment, this embodiment will further elaborate on how to perform high-precision calibration of a high-speed camera based on an improved genetic algorithm to optimize a BP neural network. Figure 5 As shown, an embodiment of the present invention discloses a high-speed camera calibration process, including:

[0078] Step S21 , obtaining a plurality of checkerboard images taken from different angles, and preprocessing each of the checkerboard images to obtain a preprocessed image; the preprocessing includes grayscale transformation, Gaussian filtering, and image reduction processing based on a multi-grid algorithm.

[0079] In this embodiment, Figure 6 As shown in the figure, based on the Zhang calibration method, several checkerboard images taken from different angles are obtained. Each checkerboard image is then subjected to grayscale transformation, Gaussian smoothing denoising, and image reduction processing based on a multi-grid algorithm to obtain the pre-processed images. In this way, the grayscale transformation can improve the image contrast and make the image display clearer, and the Gaussian filter can remove image noise.

[0080] Step S22: performing corner point detection on each of the pre-processed images, and performing matching of the detected corner points with the same name based on grid motion statistics to obtain a plurality of corner points with the same name.

[0081] In this embodiment, Figure 6As shown, a corner detection algorithm is used to detect corners in each preprocessed image to obtain a number of corner points. The corner detection algorithm can employ, for example, the Trajkovic corner detection algorithm. These detected corner points are then coarsely matched to obtain a number of matching corner points with the same name. These corner points are located at the same position in different checkerboard images. Considering that the coarse matching of corner points with the same name may result in mismatches, these mismatches are removed from the matching corner points with the same name based on grid motion statistics, resulting in the final number of matching corner points with the same name.

[0082] Step S23 , using the least squares method and based on the binocular vision model, converting the two-dimensional pixel coordinates of each of the corner points with the same name into corresponding three-dimensional world coordinates, so as to obtain coordinate information corresponding to each of the corner points with the same name.

[0083] Step S24: construct a corner point set based on the coordinate information corresponding to each corner point with the same name, and divide the corner point set into a training set and a test set.

[0084] In this embodiment, Figure 6 As shown, using the least squares method and a conventional binocular vision model, the 2D pixel coordinates of each matching corner point with the same name are converted to corresponding 3D world coordinates to obtain the coordinate information corresponding to each corner point with the same name. A corner point set is then constructed based on the coordinate information corresponding to each corner point with the same name, and the set is divided into a training set and a test set according to a preset ratio.

[0085] Step S25: Utilize the training set and perform optimization training on the BP neural network based on the improved genetic algorithm to obtain a BP neural network optimized based on the improved genetic algorithm.

[0086] In this embodiment, Figure 6As shown, the network topology is determined, and the weight threshold length of the initial BP neural network is initialized. The initial values ​​are encoded using an improved genetic algorithm (GA) to obtain a population of BP neural networks with different structures and random weights. Each BP neural network in the population is trained using a training set, and the resulting error is used as the fitness value of each BP neural network, triggering operations to improve the adaptive crossover and mutation probabilities. Specifically, a BP neural network is selected from the population based on a selection operation, and the selected BP neural network is crossovered based on a crossover operation to generate a new BP neural network. The new BP neural network is then mutated based on a mutation operation, for example, by changing its structure or adjusting its weights to obtain a mutated BP neural network. The fitness value of the mutated BP neural network is calculated, and the population is updated based on the mutated BP neural network. A first preset termination condition is then determined, which includes the number of iterations reaching a preset threshold. If the first preset termination condition is not met, the operations to improve the adaptive crossover and mutation probabilities are retriggered. If the first preset termination condition is met, a BP neural network with an optimal weight threshold is obtained from the population, and the error of the BP neural network with the optimal weight threshold is calculated to update the weight threshold to obtain an updated weight threshold. Then, it is determined whether the second preset termination condition is met. The second preset termination condition includes the number of iterations reaching a preset number threshold. If the second preset termination condition is not met, the error is recalculated based on the updated weight threshold, and the weight threshold is re-updated based on the error. If the second preset termination condition is met, a BP neural network optimized based on an improved genetic algorithm, i.e., a GA-BP neural network, is obtained.

[0087] Step S26: Use the test set to test the BP neural network optimized based on the improved genetic algorithm to obtain optimized world coordinates corresponding to the corner points with the same name in the test set, and calibrate the high-speed camera based on the optimized world coordinates.

[0088] In this embodiment, Figure 6 As shown in the figure, after obtaining a BP neural network optimized by an improved genetic algorithm, also known as an improved GA-BP neural network, the test set is normalized and input into the BP neural network optimized by the improved genetic algorithm to obtain the output world coordinates corresponding to the corner points of the same name in the normalized test set. After denormalizing the output world coordinates, the optimized world coordinates corresponding to the corner points of the same name are obtained, and high-precision calibration of the high-speed camera is performed based on the optimized world coordinates. In this way, by calibrating the high-speed camera using the BP neural network optimized by the improved genetic algorithm, submillimeter displacement measurement accuracy can be achieved.

[0089] Thus, it can be seen that this embodiment is based on checkerboard calibration, the calibration process is simple, the calibration results are relatively accurate and reliable, and it has good flexibility. Furthermore, this embodiment constructs a corner point set based on the coordinate information of the corner points of the same name in the checkerboard image, so as to optimize the training and testing of the BP neural network according to the improved genetic algorithm, thereby calibrating the high-speed camera based on the BP neural network optimized by the improved genetic algorithm, which can meet the high-precision and high-resolution requirements of the high-speed camera and improve the measurement accuracy of the high-speed motion test model.

[0090] See also Figure 7 As shown, an embodiment of the present invention discloses a visual measurement device for a maglev wind tunnel test model, which is applied to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously capture the same area in the maglev wind tunnel; wherein the device includes:

[0091] An image acquisition module 11 is used to acquire images of the test model taken synchronously by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on a BP neural network optimized by an improved genetic algorithm;

[0092] a correspondence determination module 12 for extracting feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points, and determining a plurality of groups of feature points having corresponding relationships based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicating that the feature points in the same group correspond to the same spatial point on the test model; the test model being a model mounted on a magnetically suspended mover that moves at high speed within the wind tunnel duct;

[0093] The information determination module 13 is configured to calculate the three-dimensional world coordinates of the test model based on the pixel coordinates of the plurality of groups of corresponding feature points by utilizing the parallax principle and the triangulation principle, and to determine the velocity information and posture information of the test model based on the three-dimensional world coordinates.

[0094] It can be seen that the present application calibrates the high-speed camera based on the BP neural network optimized by the improved genetic algorithm, thereby meeting the high-precision and high-resolution requirements of the high-speed camera and improving the measurement accuracy of the high-speed motion test model; and, the present application realizes the contactless measurement of the high-speed motion test model through the measurement system constructed based on the high-speed camera, avoiding the problem of installing sensors on the high-speed motion test model, which easily has a negative impact on the aerodynamic performance and motion characteristics of the model, improving the measurement efficiency of the high-speed motion test model, and ensuring the reliability of the test model data.

[0095] In some specific embodiments, the magnetic levitation wind tunnel test model visual measurement device further includes:

[0096] A coordinate alignment unit, configured to align the coordinates of the infrared camera and the two high-speed cameras to establish a world coordinate system;

[0097] Accordingly, the magnetic levitation wind tunnel test model visual measurement device further includes:

[0098] The image storage unit is used to store the calibrated test model images synchronously photographed by the two high-speed cameras in the corresponding storage device.

[0099] In some specific embodiments, the magnetic levitation wind tunnel test model visual measurement device further includes:

[0100] An image preprocessing unit, configured to obtain a plurality of checkerboard images taken from different angles, and preprocess each of the checkerboard images to obtain a preprocessed image; the preprocessing includes grayscale transformation, Gaussian filtering, and image reduction processing based on a multi-grid algorithm;

[0101] A same-name corner point matching unit is used to perform corner point detection on each of the preprocessed images, and perform same-name corner point matching on a plurality of detected corner points based on grid motion statistics to obtain a plurality of same-name corner points;

[0102] A coordinate conversion unit, configured to convert the two-dimensional pixel coordinates of each of the corner points of the same name into corresponding three-dimensional world coordinates using a least squares method and based on a binocular vision model, so as to obtain coordinate information corresponding to each of the corner points of the same name;

[0103] A corner point set construction unit, configured to construct a corner point set based on coordinate information corresponding to each corner point with the same name, and divide the corner point set into a training set and a test set;

[0104] A neural network training unit is used to optimize and train the BP neural network using the training set and based on an improved genetic algorithm to obtain a BP neural network optimized based on the improved genetic algorithm;

[0105] A camera calibration unit is used to test the BP neural network optimized based on the improved genetic algorithm using the test set to obtain optimized world coordinates corresponding to the corner points of the same name in the test set, and calibrate the high-speed camera based on the optimized world coordinates.

[0106] In some specific embodiments, the correspondence relationship determination module 12 includes:

[0107] a feature marker recognition unit, configured to perform Gaussian filtering on each of the test model images to obtain each processed model image, and to recognize a plurality of feature markers from each of the processed model images;

[0108] The feature point extraction submodule is used to extract feature points from each of the feature marks to obtain pixel coordinates of several feature points.

[0109] In some specific embodiments, the feature point extraction submodule includes:

[0110] a grayscale value determining unit, configured to respectively determine all pixel coordinates within each of the feature marks, and grayscale values ​​corresponding to all the pixel coordinates;

[0111] The pixel coordinate acquisition unit is used to calculate the pixel coordinates of the center point in each of the feature marks based on all the pixel coordinates in each of the feature marks and the grayscale values ​​corresponding to all the pixel coordinates, so as to obtain the pixel coordinates of several feature points.

[0112] In some specific embodiments, the information determination module 13 includes:

[0113] A world coordinate calculation unit, configured to calculate the pixel coordinates of any set of corresponding feature points using a pre-established world coordinate calculation formula based on the parallax principle and the triangulation principle, so as to obtain the three-dimensional world coordinates of the test model;

[0114] The magnetic levitation wind tunnel test model visual measurement device further comprises:

[0115] A camera parameter acquisition unit, configured to acquire the focal length of the high-speed camera and the distance between the optical centers of the two high-speed cameras; the focal lengths and field angles of the two high-speed cameras are the same;

[0116] A formula construction unit is used to construct the world coordinate calculation formula based on the focal length and the distance by utilizing the parallax principle and the triangulation principle.

[0117] In some specific embodiments, the magnetic levitation wind tunnel test model visual measurement device further includes:

[0118] a temperature acquisition unit, configured to acquire the original temperature of the test model measured by the infrared camera;

[0119] A temperature adjustment unit is used to determine an angle between the infrared camera and the test model, and adjust the original temperature based on the angle and a coefficient related to the wavelength of electromagnetic radiation to obtain temperature data of the test model.

[0120] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram should not be considered as any limitation to the scope of application of the present application.

[0121] Figure 8 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps of the visual measurement method for a maglev wind tunnel test model disclosed in any of the aforementioned embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0122] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0123] In addition, the memory 22 as a carrier for resource storage can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0124] The operating system 221 is used to manage and control the hardware devices on the electronic device 20 and the computer program 222. The operating system 221 can be Windows Server, NetWare, Unix, Linux, etc. In addition to including a computer program capable of implementing the visual measurement method of the magnetic levitation wind tunnel test model performed by the electronic device 20 as disclosed in any of the aforementioned embodiments, the computer program 222 can further include computer programs capable of performing other specific tasks.

[0125] Furthermore, this application discloses a computer-readable storage medium for storing a computer program. When executed by a processor, the computer program implements the aforementioned method for visually measuring a maglev wind tunnel test model. The specific steps of this method can be found in the aforementioned embodiments and are not further detailed here.

[0126] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0127] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0128] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0129] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0130] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A visual measurement method for a maglev wind tunnel test model, characterized in that: Applied to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously photograph the same area in the maglev wind tunnel; wherein the method includes: Acquire images of the test model taken synchronously by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on an improved genetic algorithm to optimize a BP neural network; Feature points are extracted from each of the test model images to obtain pixel coordinates of a plurality of feature points, and a plurality of groups of feature points having corresponding relationships are determined based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicate that the feature points in the same group correspond to the same spatial point on the test model; the test model is a model mounted on a magnetically suspended mover that moves at high speed within the wind tunnel duct; The three-dimensional world coordinates of the test model are calculated based on the pixel coordinates of the groups of corresponding feature points using the parallax principle and the triangulation principle, and the speed information and posture information of the test model are determined based on the three-dimensional world coordinates.

2. The visual measurement method of the maglev wind tunnel test model according to claim 1 is characterized in that: Before obtaining the calibrated test model images synchronously photographed by the two high-speed cameras, the method further includes: Aligning the coordinates of the infrared camera and the two high-speed cameras to establish a world coordinate system; Accordingly, after obtaining the calibrated test model images synchronously photographed by the two high-speed cameras, the method further includes: The test model images synchronously photographed by the two calibrated high-speed cameras are stored in the corresponding storage device.

3. The visual measurement method of the maglev wind tunnel test model according to claim 1 is characterized in that: Before obtaining the calibrated test model images synchronously photographed by the two high-speed cameras, the method further includes: Acquire a plurality of checkerboard images taken from different angles, and preprocess each of the checkerboard images to obtain a preprocessed image; the preprocessing includes grayscale transformation, Gaussian filtering, and image reduction processing based on a multi-grid algorithm; Performing corner point detection on each of the preprocessed images, and matching the detected corner points with the same name based on grid motion statistics to obtain a plurality of corner points with the same name; Converting the two-dimensional pixel coordinates of each corner point with the same name into corresponding three-dimensional world coordinates using the least squares method and based on a binocular vision model to obtain coordinate information corresponding to each corner point with the same name; Constructing a corner point set based on coordinate information corresponding to each of the corner points with the same name, and dividing the corner point set into a training set and a test set; Utilizing the training set and based on the improved genetic algorithm to perform optimization training on the BP neural network, so as to obtain a BP neural network optimized based on the improved genetic algorithm; The BP neural network optimized based on the improved genetic algorithm is tested using the test set to obtain optimized world coordinates corresponding to the corner points with the same name in the test set, and the high-speed camera is calibrated based on the optimized world coordinates.

4. The visual measurement method of a maglev wind tunnel test model according to claim 1, characterized in that: The extraction of feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points includes: Performing Gaussian filtering on each of the test model images to obtain each processed model image, and identifying a plurality of feature markers from each of the processed model images; Feature points are extracted from each of the feature marks to obtain pixel coordinates of several feature points.

5. The visual measurement method of the maglev wind tunnel test model according to claim 4 is characterized in that: The extracting of feature points from each of the feature markers to obtain pixel coordinates of a plurality of feature points includes: respectively determining all pixel coordinates within each of the feature marks and the grayscale values ​​corresponding to all the pixel coordinates; Based on all pixel coordinates in each of the feature marks and the grayscale values ​​respectively corresponding to all of the pixel coordinates, the pixel coordinates of the center point in each of the feature marks are calculated to obtain the pixel coordinates of several feature points.

6. The visual measurement method of a maglev wind tunnel test model according to claim 1, characterized in that: The method of calculating the corresponding three-dimensional world coordinates of the test model by utilizing the parallax principle and the triangulation principle and based on the pixel coordinates of the plurality of groups of corresponding feature points includes: Utilizing a pre-established world coordinate calculation formula based on the parallax principle and the triangulation principle, the pixel coordinates of any set of corresponding feature points are calculated to obtain the three-dimensional world coordinates of the test model; Before calculating the pixel coordinates of any set of corresponding feature points using the pre-constructed world coordinate calculation formula based on the parallax principle and the triangulation principle, the method further includes: Obtaining the focal length of the high-speed camera and the distance between the optical centers of the two high-speed cameras; the focal length and field of view angle of the two high-speed cameras are the same; The world coordinate calculation formula is constructed based on the focal length and the distance by utilizing the parallax principle and the triangulation principle.

7. The visual measurement method for a maglev wind tunnel test model according to any one of claims 1 to 6, characterized in that: Also includes: Obtaining the original temperature of the test model measured by the infrared camera; An angle between the infrared camera and the test model is determined, and the original temperature is adjusted based on the angle and a coefficient related to the wavelength of electromagnetic radiation to obtain temperature data of the test model.

8. A visual measurement device for a magnetic levitation wind tunnel test model, characterized in that: Applicable to a measurement system; the measurement system includes an infrared camera, a fill light source, two high-speed cameras, and a storage device configured for each high-speed camera; the two high-speed cameras are placed at different positions on the inner wall of the wind tunnel pipe and synchronously photograph the same area in the maglev wind tunnel; wherein the device includes: An image acquisition module is used to acquire images of the test model synchronously captured by the two calibrated high-speed cameras; the two calibrated high-speed cameras are both cameras calibrated based on a BP neural network optimized by an improved genetic algorithm; a correspondence determination module for extracting feature points from each of the test model images to obtain pixel coordinates of a plurality of feature points, and determining a plurality of groups of feature points having corresponding relationships based on relative position information of the feature points in the corresponding test model images; the corresponding relationships indicating that the feature points in the same group correspond to the same spatial point on the test model; the test model being a model mounted on a magnetically suspended mover that moves at high speed within the wind tunnel duct; An information determination module is configured to calculate the three-dimensional world coordinates of the test model based on the pixel coordinates of the plurality of groups of corresponding feature points using the parallax principle and the triangulation principle, and to determine the velocity information and posture information of the test model based on the three-dimensional world coordinates.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the visual measurement method for a maglev wind tunnel test model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Used to store a computer program, which, when executed by a processor, implements the visual measurement method for a magnetic levitation wind tunnel test model according to any one of claims 1 to 7.

Images