A method for constructing 4D flow field of gas-solid two-phase particle jet based on dual-camera photography

Abstract

The present invention provides a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera shooting, the construction method comprising the following steps: (1) using two calibrated high-speed cameras arranged vertically to record the gas-solid two-phase particle jet to obtain raw data; (2) the raw data is subjected to particle recognition to obtain recognition results of two high-speed cameras; based on the recognition results, particle inter-frame tracking and matching are performed to obtain particle information in two-dimensional images of two views; (3) the particle information of the first view and the second view are three-dimensionally matched, the final matching particle information is screened out, and the time-resolved 3D trajectory of the particles is obtained to complete the construction of the 4D flow field. The method of the present invention only requires two cameras to shoot and record to complete the reconstruction of the 4D process, the cost is significantly reduced, and the algorithm process is simpler and the application prospect is broader.

Description

Technical Field

[0001] The present invention relates to the technical field of gas-solid two-phase flow, and in particular to a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography. Background Art

[0002] Time-resolved three-dimensional particle flow field reconstruction is also called 4D flow field reconstruction of particle flow. The three-dimensional reconstruction of particle flow field is the key and difficulty of gas-solid two-phase particle jet research. At present, the research on particle flow field in gas-solid two-phase particle jet experiments is mainly limited to two dimensions, and the research on time-resolved three-dimensional particle flow field reconstruction needs to be broken through. Factors such as the large number of cameras required, expensive laser equipment, and complex reconstruction algorithms restrict its development.

[0003] At present, there are few studies on 4D flow field reconstruction of particle flow. Internationally, Schanz et al. proposed a STB method (see "Shake-The-Box: Lagrangian particle tracking at high particle image densities", Schanz et al., Experiments in Fluids, 2016.57(5)).

[0004] Tan et al. developed an OpenLPT method based on it (see “Introducing OpenLPT: new method of removing ghost particles and high-concentration particle shadow tracking”, Tan et al., Experiments in Fluids, 2020.61(2)).

[0005] Bourgoin et al. used ray traversal for 3D particle matching (see “Using ray-traversal for 3D particle matching in the context of particle tracking velocimetry influid mechanics”, Bourgoin et al., Review of Scientific Instruments, 2020.91(8)).

[0006] However, these methods often require 3 to 4 cameras to complete the 4D flow field reconstruction of the particle flow, and the algorithms are also relatively complex.

[0007] Therefore, it is necessary to develop a 4D flow field reconstruction method with low equipment cost and relatively simple algorithm. Summary of the invention

[0008] In view of the problems existing in the prior art, the present invention provides a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera shooting. The method uses two cameras with vertical shooting angles to shoot the gas-solid two-phase particle jet, and obtains two groups of time series of two-dimensional images of the particle flow field. The Lagrangian tracking matching of each particle in the three-dimensional particle flow field is obtained through image analysis. The equipment cost is low and the algorithm is simple, which provides a new idea for the construction of 4D flow field.

[0009] To achieve this object, the present invention adopts the following technical solutions:

[0010] The present invention provides a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography, the construction method comprising the following steps:

[0011] (1) Using two high-speed cameras with calibrated shooting angles and set vertically to record the gas-solid two-phase particle jet to obtain raw data;

[0012] (2) The raw data is subjected to particle recognition to obtain recognition results of two high-speed cameras; based on the recognition results, particle inter-frame tracking and matching are performed to obtain particle information in two-dimensional images of two views;

[0013] (3) Performing three-dimensional matching on the particle information of the first view and the second view, selecting the final matching particle information, obtaining the time-resolved 3D trajectory of the particle, and completing the construction of the 4D flow field.

[0014] The method for constructing a 4D flow field of a gas-solid two-phase particle jet provided by the present invention only requires the setting of two cameras. Compared with the prior art that requires the setting of 3 to 4 cameras, the equipment cost is significantly reduced. After obtaining the recognition result through particle recognition, duplicate particles can be deleted through three-dimensional matching, and the final matching particle information can be screened out. The obtained 4D flow field is closer to the actual flow field and has a small error.

[0015] Preferably, the two high-speed cameras are preferably respectively arranged at a front view position and a top view position, so that a front view and a top view can be obtained, which is more convenient for shooting, and the solution of the present invention can also be implemented in other views.

[0016] Preferably, the two high-speed cameras in step (1) are at equal distances from the center of the gas-solid two-phase particle jet field.

[0017] The present invention further preferably places the two high-speed cameras at equal distances from the jet center to ensure that the spatial resolution of the images is as consistent as possible.

[0018] Preferably, the calibration includes eliminating high-speed camera image aberrations and distortions.

[0019] The calibration of the present invention includes calibrating two high-speed cameras by using a signal synchronizer, so that the corresponding frames of the images captured by the two cameras are at the same time as much as possible.

[0020] Preferably, the synchronization accuracy of the signal synchronizer is within 0.1ns.

[0021] Preferably, the light source for shooting is an LED light source or a laser.

[0022] The light source in the present invention is required to be able to capture images with high differentiation between background particles and foreground particles. It is preferred to use the above light source, which has a higher differentiation.

[0023] Preferably, the original data in step (1) includes image data.

[0024] The present invention does not impose any particular limitation on the method of particle identification. It is required that the particles in the image can be clearly identified, including overlapping particles. The method can be performed using an algorithm well known to those skilled in the art, or can be selected according to specific circumstances.

[0025] Preferably, the particle identification algorithm in step (2) includes an adaptive threshold method and a Hough transform.

[0026] The present invention obtains the center coordinates of particles through particle recognition, so as to prepare for subsequent particle frame tracking and matching.

[0027] Preferably, before the particle identification, the original data is corrected according to the calibration result in step (1), and then the particle identification is performed based on the corrected data.

[0028] The present invention preferably corrects the image according to the result of camera calibration before performing particle recognition to eliminate the image aberration and distortion caused by the lens and shooting.

[0029] Preferably, the tracking method in the particle frame-to-frame tracking and matching in step (2) includes a Kalman filter method.

[0030] Preferably, the matching method in the particle inter-frame tracking matching includes the Hungarian algorithm.

[0031] Preferably, the particle inter-frame tracking and matching includes: predicting the motion trajectory using the Kalman filter method to obtain predicted particles, matching the predicted particles with particles identified in the current frame of the corrected data, and updating the motion trajectory of the predicted particles using the Kalman filter method.

[0032] Preferably, predicted particles that fail to match successfully with particles identified in multiple consecutive frames of corrected data are deleted.

[0033] The present invention can obtain the motion trajectory of each particle in the two-dimensional image through particle frame tracking and matching, that is, the time-resolved two-dimensional particle position, such as the particle trajectory of 10 consecutive frames or the particle trajectory of 100 consecutive frames.

[0034] Preferably, the particle information in step (2) includes the view where the particle is located, the particle number, the frame number where the particle is located, the two-dimensional coordinate trajectory of the particle in the time series, and the time length of the particle Lagrangian tracking.

[0035] The two-dimensional coordinate trajectory of the particle in the time series of the present invention is the Lagrangian tracking trajectory of the particle, and the time length of the Lagrangian tracking of the particle is the number of frames from the appearance to the disappearance of the particle.

[0036] Preferably, the three-dimensional matching in step (3) includes: screening out particles with the smallest error in particle information between the first view and the second view, and determining them to be the same particle.

[0037] Preferably, the particle criteria for minimizing the error of the particle information in step (3) include: the intersection lengths of the Lagrangian tracking trajectories of the particles in time are similar, the time lengths of the Lagrangian tracking trajectories of the particles are similar, the average value of the x-axis coordinate error of the Lagrangian tracking trajectories of the particles in the same time series is minimized, or the average value error of the sum of the x-axis coordinates of the Lagrangian tracking trajectories of the particles in the same time series is minimized, wherein the x-axis refers to the same coordinate axis as the first view and the second view.

[0038] The present invention adopts a combined multi-criteria method for matching, thereby further improving the accuracy of three-dimensional matching of particles.

[0039] Preferably, the step (3) of selecting particles with the smallest error in particle information between the first view and the second view specifically comprises the following steps:

[0040] (I) calculating the intersection length of particle information in time, and selecting the particle number whose intersection length is greater than a first specified value;

[0041] (II) Calculate the average x-axis coordinate error sum(Δx) of the Lagrangian tracking trajectory of the particle in the same time series;

[0042] (III) Select the particle number whose sum(Δx) is less than the second specified value, and take the particle with the smallest sum(Δx) as the valid matching particle;

[0043] (IV) calculating the time length error ΔLt of the Lagrangian tracking trajectory in the effective matching particles, screening out the particle numbers whose time length error ΔLt is less than a third specified value, and selecting the particle numbers whose Lagrangian tracking trajectories of the particles have a time intersection length greater than a fourth specified value from the particle numbers;

[0044] (V) Repeating steps (II), (III) and (IV) for the particle numbers obtained in step (IV) to screen and obtain corresponding particle numbers;

[0045] (VI) Calculate the average error of the x-axis coordinates and the Lagrangian tracking trajectory of the particles selected in step (V) on the same time series Filter out mean error The particle whose number is smaller than the fifth specified value is used as the final matching particle number.

[0046] The present invention further preferably adopts the above-mentioned method to judge the particle matching, and strictly follows the above-mentioned order, which can effectively avoid misjudgment and omission of particles and accurately match them.

[0047] Preferably, in step (I), when the intersection length is less than or equal to the first specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0048] Preferably, in step (IV), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0049] Preferably, in step (V), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0050] Preferably, when other criteria are not met, the particle number is deleted.

[0051] Preferably, the Voronoi concentration of particles in the gas-solid two-phase particle jet is less than 1%.

[0052] The present invention has no special limitation on the ranges of the first designated value, the second designated value, the third designated value, the fourth designated value and the fifth designated value, and they can be selected according to actual conditions.

[0053] Preferably, the fourth designated value is greater than the first designated value, which can further narrow the range.

[0054] The method of the present invention is more suitable for a jet field with a particle Voronoi concentration of less than 1%, a moderate particle concentration, a small particle recognition error in a single view, a higher reconstruction accuracy, and a smaller error.

[0055] Compared with the prior art, the present invention has at least the following beneficial effects:

[0056] (1) The method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography provided by the present invention only requires two high-speed cameras, and the equipment cost is low;

[0057] (2) The method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography provided by the present invention directly uses the captured images to calculate the true position of the particles. The algorithm is simple, which is specifically reflected in the small program size required, which is only less than 60KB, and the calculation speed is fast. For different particle concentrations and particle velocities, the longest 4D flow field construction time for completing the Lagrangian tracking of particles in the image does not exceed 100 seconds;

[0058] (3) The method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography provided by the present invention has a small error between the constructed 4D flow field and the experimental trajectory, with an average matching error within 5 pixels and an accuracy of more than 93%. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of a high-speed camera measuring device used in a specific embodiment of the present invention.

[0060] Figure 2 It is a flow chart of a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography provided in Example 1 of the present invention.

[0061] Figure 3 This is a particle size distribution diagram of particles in Example 1 of the present invention.

[0062] Figure 4 These are (a) the original data image in Example 1 of the present invention; (b) the image background of the adaptive threshold method; (c) the thresholded image; and (d) the particle identification mark image.

[0063] Figure 5 It is the particle tracking trajectory in the two-dimensional image obtained by particle inter-frame tracking matching in Example 1 of the present invention.

[0064] Figure 6 This is the time-resolved 3D trajectory of the particle obtained in Example 1 of the present invention.

[0065] Figure 7 It is the spatial position of particles in a certain frame in embodiment 1 of the present invention.

[0066] Figure 8 It is a projection view of the spatial position of particles in a certain frame in the first view in embodiment 1 of the present invention.

[0067] Fig. 9 It is a projection view of the spatial position of particles in a certain frame in the second view in embodiment 1 of the present invention.

[0068] Fig.10 This is the matching situation of particles in a certain frame in the first view according to the first embodiment of the present invention.

[0069] Fig.11 This is the matching situation of particles in a certain frame in the second view according to the first embodiment of the present invention.

[0070] In the figure: 1-feeder; 2-injector; 3-measuring box; 4-first high-speed camera; 5-second high-speed camera; 6-light source; 7-signal synchronizer; 8-computer. DETAILED DESCRIPTION

[0071] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0072] The present invention is further described in detail below. However, the following examples are only simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

[0073] As a specific embodiment of the present invention, the specific embodiment provides a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography. In this specific embodiment, the schematic diagram of the construction device layout used in the construction method is as follows: Figure 1 shown.

[0074] from Figure 1 It can be seen that the construction device includes a first high-speed camera 4 , a second high-speed camera 5 , a signal synchronizer 7 , a computer 8 , a light source 6 , an ejector 2 , a measuring box 3 and a feeder 1 .

[0075] The outlet of the feeder 1 is connected to the inlet of the injector 2, and the outlet of the injector 2 is connected to the inlet of the measuring box 3. The first high-speed camera 4 is preferably arranged directly above the measuring box 3, and the shooting direction is consistent with the top-down direction of the measuring box 3; the second high-speed camera 5 is preferably arranged on the side of the measuring box 3, and its shooting direction is at a 90° angle to the shooting direction of the first high-speed camera 4; a light source 6 is arranged in front of the first high-speed camera 4 and the second high-speed camera 5, and the first high-speed camera 4 and the second high-speed camera 5 are jointly connected to a signal synchronizer 7 for synchronizing and transmitting image signals, and the signal synchronizer 7 is connected to the computer 8 for transmitting the signal to the computer 8 for processing.

[0076] In this specific embodiment, the light source is preferably an LED light source. The feeder can be any feeder, for example, a screw feeder.

[0077] In this specific embodiment, the construction method includes the following steps:

[0078] (1) Two high-speed cameras with calibrated shooting angles and set vertically were used to record the gas-solid two-phase particle jet. Figure 1The device in the apparatus is used to record and obtain the original data;

[0079] (2) The raw data is corrected according to the camera calibration results to eliminate the image aberration and distortion caused by the lens and the shooting, and then the particles are identified to obtain the identification results of the two high-speed cameras; based on the identification results, the particle inter-frame tracking and matching is performed to obtain the particle information in the two-dimensional images of the two views;

[0080] (3) Performing three-dimensional matching on the particle information of the first view and the second view, selecting the final matching particle information, obtaining the time-resolved 3D trajectory of the particle, and completing the construction of the 4D flow field.

[0081] In a preferred solution, the particle recognition method in step (2) is an adaptive threshold method and Hough transform, and particles that overlap in the view can be recognized during the recognition process.

[0082] In a preferred embodiment, the method for tracking and matching particles between frames in step (2) includes: using the Kalman filter method to predict the motion trajectory to obtain predicted particles, and matching the predicted particles with the particles identified in the current frame of the corrected data, and simultaneously using the Kalman filter method to update the motion trajectory of the predicted particles, thereby obtaining the motion trajectory of each particle in the two-dimensional image.

[0083] The specific particle frame tracking and matching process is based on the idea of ​​sufficiency proof. The Hungarian algorithm is used to continuously search for augmented paths, traverse each matching object, and find the particle that best matches the motion trajectory predicted by the Kalman filter method to achieve the most matching of particles.

[0084] In a preferred embodiment, the process of three-dimensional matching of particles in step (3) comprises the following steps:

[0085] (I) calculating the intersection length of particle information in time, and selecting the particle number whose intersection length is greater than a first specified value;

[0086] (II) Calculate the average x-axis coordinate error sum(Δx) of the Lagrangian tracking trajectory of the particle in the same time series;

[0087] (III) Select the particle number whose sum(Δx) is less than the second specified value, and take the particle with the smallest sum(Δx) as the valid matching particle;

[0088] (IV) calculating the time length error ΔLt of the Lagrangian tracking trajectory in the effective matching particles, screening out the particle numbers whose time length error ΔLt is less than a third specified value, and selecting the particle numbers whose Lagrangian tracking trajectories of the particles have a time intersection length greater than a fourth specified value from the particle numbers;

[0089] (V) Repeating steps (II), (III) and (IV) for the particle numbers obtained in step (IV) to screen and obtain corresponding particle numbers;

[0090] (VI) Calculate the average error of the x-axis coordinates and the Lagrangian tracking trajectory of the particles selected in step (V) on the same time series Filter out mean error The particles smaller than the fifth specified value are used as the final matching particle numbers;

[0091] In step (I), when the intersection length is less than or equal to the first specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0092] In step (IV), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0093] In step (V), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

[0094] When other criteria are not met, the particle number is deleted.

[0095] This specific implementation method uses a further preferred criterion to determine that the particles with the smallest particle information error in the two views are the same particle, thereby constructing the 3D trajectory of the particle.

[0096] Example 1

[0097] This embodiment provides a method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography. The overall process is as follows: Figure 2 As shown, the construction method comprises the following steps:

[0098] (1) Two high-speed cameras with calibrated shooting angles were set vertically to observe the gas-solid two-phase particle jet (where the Voronoi concentration (volume fraction, the dynamic change of particle concentration during the jetting process) was 10 -5 ~10 -2 , the jet outlet gas velocity is 25m / s as an example, where the particle size distribution diagram is as follows Figure 3 As shown, Figure 3 Where dp represents the particle size, the average particle size dp (mean) is 708.2 μm, and dp (rms) is 32.33 μm) is recorded to obtain the original data. The original data is as follows: Figure 4 As shown in (a);

[0099] (2) The original data is corrected according to the camera calibration results to eliminate the image distortion caused by the lens and shooting, and then the recognition results of the two high-speed cameras are obtained through particle recognition. The recognition results are as follows: Figure 4 As shown, from Figure 4 The identified particles can be clearly seen in the figure; the particle identification method is adaptive threshold method and Hough transformation, the specific process is:

[0100] Calculate the adaptive threshold background: Due to the influence of lighting, the background brightness of the image may be unevenly distributed. It is difficult to segment the target image area by directly performing global threshold segmentation. In order to solve this problem, this paper uses an adaptive threshold method to process images. The so-called adaptive threshold method is to calculate the threshold of each area of ​​the image. Compared with the global threshold method, for different areas of the image with uneven brightness distribution, the adaptive threshold method can calculate the threshold of different areas, and its image processing accuracy is higher. The present invention calculates the local threshold based on the local mean intensity of each pixel neighborhood. The advantage is that it takes into account the neighboring pixels on all sides of each pixel, rather than calculating the average value of the pixels in a segmented area of ​​the image. This image processing accuracy is higher, and the image background obtained by the adaptive threshold method is as follows: Figure 4 As shown in (b).

[0101] Then the image is thresholded and the resulting image is Figure 4 As shown in (c).

[0102] The particles are identified by Hough transform, and the center coordinates of the particles are obtained, thereby obtaining an image of the particle identification mark, such as Figure 4 As shown in (d).

[0103] Based on the recognition result, particle inter-frame tracking and matching are performed to obtain particle information in two-dimensional images of two views;

[0104] The particle frame tracking and matching method includes: using the Kalman filter method to predict the motion trajectory to obtain the predicted particles, and matching the predicted particles with the particles identified in the current frame of the corrected data, and using the Kalman filter method to update the motion trajectory of the predicted particles, thereby obtaining the motion trajectory of each particle in the two-dimensional image. Figure 5 As shown, Figure 5 Figure a shows the particle trajectories of 10 frames in the side view, Figure b shows the particle trajectories of 100 frames in the side view, Figure c shows the particle trajectories of 10 frames in the top view, and Figure d shows the particle trajectories of 100 frames in the top view.

[0105] (3) performing three-dimensional matching on the particle information of the first view and the second view, selecting the final matching particle information, obtaining the time-resolved 3D trajectory of the particle, and completing the construction of the 4D flow field;

[0106] The process of three-dimensional matching of particles includes the following steps: first, the particle position in the first view is marked as the x, z coordinates of the particle, and the particle position in the second view is marked as the x, y coordinates of the particle. Both have an x-axis, and three-dimensional matching is performed based on the x-axis coordinates of the particle.

[0107] In order to improve the accuracy of particle 3D matching, first construct particle information for each particle. The particle information of each particle includes the view where the particle is located, the particle number, the frame number where the particle is located, the 2D coordinate trajectory of the particle in the time series, and the time length of the particle Lagrangian tracking. For example, the view where particle a is located is the side view (second view), and the particle number is i. front , the frame number set of particles is A = [m, m*], and the particle Lagrangian tracking trajectory set is X front =[x frontm , x frontm* ], the time length of the particle's Lagrangian tracking trajectory is Lt front , then the particle information of particle a is a={front, i front , A, X front , Lt front Similarly, the particle information of particle b in the top view (first view) is a={top, i top , B, X top , Lt top When the information error between the particle in the first view and the particle in the second view is the smallest, it is determined that they are the same particle, which specifically includes the following steps:

[0108] (I) Calculating the time intersection length Lints of the particle information; for example, the frame number set where particle a appears in the second view is A = [m, m*], and the frame number set where particle b appears in the first view is B = [n, n*], then the common frame set of the two particles is C = A∩B = [M, N], assuming that C is a non-empty set, the time intersection length of the Lagrangian tracking trajectories of the two particles is obtained to be Lints = N-M+1, and the particle number whose intersection length Lints is greater than the first specified value is selected;

[0109] (II) Calculate the average x-axis coordinate error sum(Δx) of the Lagrangian tracking trajectory of the particle in the same time series, and the calculation formula is as described in the following formula (1):

[0110]

[0111] (III) Select the particle number whose sum(Δx) is less than the second specified value, and then select the particles with similar Lagrangian tracking trajectories, and take the particle with the smallest sum(Δx) as the effective matching particle;

[0112] (IV) Calculate the time length error ΔLt of the Lagrangian tracking trajectory of the particle, and the calculation formula is shown in formula (2);

[0113] ΔLt=Lt front -Lt top Formula (2)

[0114] Screening out particle numbers whose time length errors ΔLt are less than a third specified value, and selecting particle numbers whose Lagrangian tracking trajectories have a time intersection length greater than a fourth specified value from the particle numbers;

[0115] (V) repeating steps (II), (III) and (IV) for the particle numbers obtained in step (IV) to screen and obtain corresponding particle numbers; wherein the third designated value and the fourth designated value in the repeated process may be different from those in the first process;

[0116] (VI) Calculate the average error of the x-axis coordinates and the Lagrangian tracking trajectory of the particles selected in step (V) on the same time series The calculation formula is shown in formula (3), and the average error is screened out The particles smaller than the fifth specified value are used as the final matching particle numbers;

[0117]

[0118] In the figure, A indicates: the next particle in the first view is matched with all particles in the second view. In step (I), when the intersection length is less than or equal to the first specified value, the next particle in the first view is matched with all particles in the second view; in step (IV), when the time length error ΔLt is greater than or equal to the third specified value, the next particle in the first view is matched with all particles in the second view; in step (V), when the time length error ΔLt is greater than or equal to the third specified value, the next particle in the first view is matched with all particles in the second view. When other criteria are not met, the particle number is deleted.

[0119] The motion trajectory of particles in the three-dimensional flow field reconstructed in this embodiment is as follows: Figure 6 As shown, the 3D trajectory of a particle in a certain frame is as follows Figure 7 As shown, the direction of the arrow indicates the direction of particle movement, and the length of the arrow indicates the speed of particle movement. Figures 8-9 They respectively represent the trajectory of a frame of the reconstructed three-dimensional flow field projected onto the two-dimensional trajectory in the first view and the second view, wherein the direction of the arrow also represents the direction of particle movement, and the length of the arrow represents the speed of particle movement.

[0120] will with Figures 8-9The two-dimensional particle position after reconstruction is matched with the actual particle position, and the matching image of a frame in the first view and the second view is selected respectively as follows Fig.10 and Fig.11 As shown, from Figures 10-11 It can be seen that, whether it is the first view or the second view, the matching is good, only a few particles are not matched (white particles are successfully matched particles, and gray particles are unmatched particles), and the accuracy can reach more than 93%. In addition, the particle information calculated in this embodiment is reprojected into the original image, and the reprojection error of the particle trajectory is calculated. The average matching error is within 5 pixels.

[0121] In summary, the method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography provided by the present invention directly uses the captured images to calculate the true position of the particles. The algorithm is simple and accurate, which makes up for the defects of the existing reconstruction technology, such as high equipment cost and complex algorithm.

[0122] The present invention illustrates the detailed structural features of the present invention through the above embodiments, but the present invention is not limited to the above detailed structural features, that is, it does not mean that the present invention must rely on the above detailed structural features to be implemented. Those skilled in the art should understand that any improvement of the present invention, equivalent replacement of the components selected by the present invention, addition of auxiliary components, selection of specific methods, etc., all fall within the protection scope and disclosure scope of the present invention.

Claims

1. A method for constructing a 4D flow field of a gas-solid two-phase particle jet based on dual-camera photography, characterized in that: The construction method comprises the following steps: (1) Using two high-speed cameras with calibrated shooting angles and set vertically to record the gas-solid two-phase particle jet to obtain raw data; (2) The raw data is subjected to particle recognition to obtain recognition results of two high-speed cameras; based on the recognition results, particle frame tracking and matching are performed to obtain particle information in two-dimensional images of two views; the particle information includes the view where the particle is located, the particle number, the frame number where the particle is located, the two-dimensional coordinate trajectory of the particle in the time series, and the time length of the particle Lagrangian tracking; wherein the particle information of particle a is a={front, i front , A, X front , Lt front }, the particle information of particle b is b = {top, i top , B, X top , Lt top }; (3) performing three-dimensional matching on the particle information of the first view and the second view, selecting the final matching particle information, obtaining the time-resolved 3D trajectory of the particle, and completing the construction of the 4D flow field; The process of three-dimensional matching of particles includes the following steps: firstly, the particle position in the first view is marked as the x and z coordinates of the particle, and the particle position in the second view is marked as the x and y coordinates of the particle, both of which have an x-axis, and three-dimensional matching is performed according to the x-axis coordinates of the particle; First, the particle information of each particle is constructed. The particle information of each particle includes the view where the particle is located, the particle number, the frame number where the particle is located, the two-dimensional coordinate trajectory of the particle in the time series, and the time length of the particle Lagrangian tracking. When the error between the particle information in the first view and the particle information in the second view is the smallest, it is determined to be the same particle. Specifically, the following steps are included: (I) Calculating the time intersection length Lints of the particle information; wherein the frame number set where particle a appears in the second view is A = [m, m*], and the frame number set where particle b appears in the first view is B = [n, n*], then the common frame set of the two particles is C = A∩B = [M, N], assuming that C is a non-empty set, the time intersection length of the Lagrangian tracking trajectories of the two particles is obtained to be Lints = N-M+1, and the particle number whose intersection length Lints is greater than the first specified value is selected; (II) Calculate the average x-axis coordinate error sum(Δx) of the Lagrangian tracking trajectory of the particle in the same time series, and the calculation formula is as described in the following formula (1): (III) Select the particle number whose sum(Δx) is less than the second specified value, and then select the particles with similar Lagrangian tracking trajectories, and take the particle with the smallest sum(Δx) as the effective matching particle; (IV) Calculate the time length error ΔLt of the Lagrangian tracking trajectory of the particle, and the calculation formula is shown in formula (2); ΔLt=Lt front -Lt top Formula (2) Screening out particle numbers whose time length errors ΔLt are less than a third specified value, and selecting particle numbers whose Lagrangian tracking trajectories have a time intersection length greater than a fourth specified value from the particle numbers; (V) repeating step (II), step (III) and step (IV) for the particle number obtained in step (IV) to screen and obtain the corresponding particle number; wherein the third designated value and the fourth designated value in the repeated process may be different from those in the first process; (VI) Calculate the average error of the x-axis coordinates and the Lagrangian tracking trajectory of the particles selected in step (V) on the same time series The calculation formula is shown in formula (3), and the average error is screened out The particles smaller than the fifth specified value are used as the final matching particle numbers; 2. The construction method according to claim 1, characterized in that: The two high-speed cameras described in step (1) are at equal distances from the center of the gas-solid two-phase particle jet field.

3. The construction method according to claim 1, characterized in that: The calibration includes eliminating high-speed camera image aberrations and distortions.

4. The construction method according to claim 1 or 2, characterized in that: The original data in step (1) includes image data.

5. The construction method according to claim 1, characterized in that: The particle recognition algorithm in step (2) includes an adaptive threshold method and a Hough transform.

6. The construction method according to claim 1, characterized in that: Before the particle identification, the original data is corrected according to the result of the calibration in step (1), and then the particle identification is performed based on the corrected data.

7. The construction method according to claim 1, characterized in that: The tracking method in the particle frame-to-frame tracking matching in step (2) includes a Kalman filter method.

8. The construction method according to claim 1, characterized in that: The matching method in the particle inter-frame tracking matching includes the Hungarian algorithm.

9. The construction method according to claim 1, characterized in that: The particle frame tracking and matching includes: using the Kalman filter method to predict the motion trajectory to obtain predicted particles, and matching the predicted particles with particles identified in the current frame of the corrected data, and using the Kalman filter method to update the motion trajectory of the predicted particles.

10. The construction method according to claim 1, characterized in that: Delete the predicted particles that have not been successfully matched with the particles identified in the corrected data for multiple consecutive frames.

11. The construction method according to claim 1, characterized in that: In step (I), when the intersection length is less than or equal to the first specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

12. The construction method according to claim 1, characterized in that: In step (IV), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

13. The construction method according to claim 1, characterized in that: In step (V), when the time length error ΔLt is greater than or equal to the third specified value, the next matching calculation is performed on the next particle in the first view and all particles in the second view.

Images