High-speed moving target transient three-dimensional reconstruction method based on Fourier transform profilometry and binocular intersection

Abstract

The invention provides a high-speed moving target transient three-dimensional reconstruction method based on Fourier transform profilometry and binocular intersection, and the method comprises the steps: building a multi-view measurement system of a three-dimensional surface shape of a high-speed moving target, capturing a projectile through a high-resolution camera under the illumination of stripe laser, and obtaining projectile stripe distribution images at a plurality of views; for the projectile stripe distribution image, applying an FTP method to obtain original three-dimensional reconstruction phase data, and using the three-dimensional reconstruction data to obtain an original 3D image of the projectile; and for a high-speed moving target surface shape damaged part and an edge stripe overlapping region noise abnormal region, obtaining edge data of a shot projectile stripe image based on the shot projectile stripe image, fusing the edge data with original three-dimensional reconstruction data, and obtaining a projectile 3D image through fused phase data. According to the method, the three-dimensional reconstruction precision of the surface topography of the high-speed moving target is effectively improved, and a new solution is provided for the three-dimensional reconstruction problem of the surface topography of the high-speed moving target.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision / structured light three-dimensional reconstruction, and in particular relates to a method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection. Background Art

[0002] Fourier transform profilometry (FTP) is widely used for high-precision three-dimensional surface measurement, particularly in industrial inspection, biomedical imaging, and topography reconstruction of high-speed dynamic targets. For example, in the aerospace and defense sectors, FTP can be used to measure the transient three-dimensional topography of high-speed vehicles (such as missiles and bullets) to optimize their aerodynamic design. In industrial manufacturing, it can be used for non-destructive testing of precision parts and identify tiny topographical defects. In medicine, FTP can assist in the three-dimensional imaging of biological tissues, improving the accuracy of surgical navigation and medical diagnosis. Its single-frame measurement capability makes it particularly suitable for high-speed dynamic scenarios, such as projectile topography measurement and vibration analysis.

[0003] In the field of three-dimensional reconstruction of high-speed moving targets, since FTP usually uses a single perspective to obtain morphological information, it is difficult to fully capture the structure of occluded areas or large-angle curved surfaces. There are calculation noise anomalies in the surface damage and the overlapping areas of edge stripes, which makes it difficult to correctly calculate and cause errors. Summary of the Invention

[0004] In order to solve the above problems, the present invention proposes a method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

[0005] The technical solution of the present invention is: a method for transient three-dimensional reconstruction of a high-speed moving target based on Fourier transform profilometry and binocular intersection, comprising the following steps:

[0006] Step 1: Build a multi-view measurement system for the three-dimensional surface of a high-speed moving target. Capture fringe patterns on reference planes and reference inclined surfaces, plot phase-height curves for the three cameras, calculate pixelation scale parameters and phase-height coefficients for the three cameras, calibrate using a checkerboard grid, and calculate the intrinsic and extrinsic parameter matrices for the three CCD high-resolution digital cameras. Then, use the high-resolution cameras to capture the projectile under fringe laser illumination and obtain fringe distribution images of the projectile at multiple viewpoints.

[0007] Step 2: Apply the FTP method to the projectile fringe distribution image to obtain the original 3D reconstructed phase data, and use the 3D reconstructed data to obtain the original 3D image of the projectile;

[0008] Step 3: For the surface damage of the high-speed moving target and the abnormal noise area in the edge stripe overlapping area, obtain its edge data based on the captured projectile stripe image, fuse the edge data with the original three-dimensional reconstruction data, and obtain the projectile 3D image from the fused phase data.

[0009] Furthermore, a multi-view measurement system for the three-dimensional surface of a high-speed moving target is built. Fringe patterns of a reference plane and a reference inclined plane are captured, phase-height curves of the three cameras are plotted, pixelation ratio parameters and phase-height coefficients of the three cameras are calculated, and calibration is performed using a checkerboard grid. The intrinsic and extrinsic parameter matrices of the three CCD high-resolution digital cameras are calculated. Then, under stripe laser illumination, a high-resolution camera is used to capture a projectile and obtain stripe distribution images of the projectile at multiple perspectives. The steps include:

[0010] Step S11, building a multi-view measurement system for the three-dimensional shape of a high-speed moving target;

[0011] The multi-view measurement system includes three identical CCD high-resolution digital cameras, a hexahedron overall bracket, and an adapter bracket. The hexahedron overall bracket is composed of six metal rods of the same length fastened together by bolts through corner pieces to form a standard hexahedron frame. A mounting base plate is provided at the center of the three spacing surfaces. The mounting base plate is pre-machined with multiple evenly distributed mounting holes. Each CCD high-resolution digital camera is connected to the mounting base plate via an adapter bracket.

[0012] Use a laser collimator to calibrate the optical axis so that the angle between the optical axes of each pair of cameras is 120°. This ensures that every point on the rotating surface to be measured falls within the overlapping area between the camera and projection lens fields of view. Run the multi-view measurement system to take pictures, and adjust the focal lengths of the three cameras until the pictures are clear.

[0013] Step S12, photographing fringe patterns of the reference plane and the reference inclined plane, plotting phase-height curves of the three cameras, and calculating pixel ratio parameters and phase-height coefficients of the three cameras;

[0014] Three CCD high-resolution digital cameras were used to capture fringe patterns of a reference plane and a reference inclined surface. The fringe patterns were sampled, cropped, and phase extracted to obtain the wrapped phase information of the reference plane and the reference inclined surface. The corresponding phase change was obtained by using the height change from the reference plane to the reference inclined surface. Phase-height curves of the three cameras were plotted, and the pixel ratio parameters of the X, Y, and Z axes, as well as the phase-height coefficients of the three cameras, were calculated.

[0015] Step S13, using a checkerboard to perform calibration, and calculating the intrinsic parameter matrix and extrinsic parameter matrix of the three CCD high-resolution digital cameras;

[0016] Using a chessboard for calibration, three CCD high-resolution digital cameras simultaneously photograph the same chessboard. By changing the shooting posture and position of the chessboard, multiple sets of calibration images are taken. The phase-height coefficient is used for corner detection and camera internal and external parameter calculation to obtain the intrinsic and external parameter matrices of the three CCD high-resolution digital cameras.

[0017] Step S14, using a high-resolution camera to capture the high-speed moving projectile under stripe laser illumination, and obtaining stripe distribution images of the projectile at multiple viewing angles;

[0018] The experiment was carried out at an outdoor shooting range. Under stripe laser illumination, a CCD high-resolution digital camera was used to capture the stripe images of the projectile passing through the field of view.

[0019] Furthermore, the FTP method is applied to the projectile fringe distribution image to obtain original three-dimensional reconstructed phase data, and the original 3D image of the projectile is obtained using the three-dimensional reconstructed data; the steps include:

[0020] Step S21, sampling and cutting the fringe images of the high-speed moving projectile captured by the three cameras to obtain three sets of fringe images;

[0021] In step S22, the FTP method is used to extract three sets of continuous phase maps reflecting the phase value of each pixel position on the projectile surface from the three sets of fringe images. The grayscale analysis of the three sets of fringe images is performed using the threshold segmentation or phase amplitude method to generate a mask image that only retains the effective area of ​​the projectile at each viewing angle. The mask image is multiplied pixel by pixel with the continuous phase map extracted at each camera viewing angle to obtain the original three-dimensional reconstructed phase data of the three surfaces of the projectile captured by the camera.

[0022] In step S23, the original 3D image of the projectile is obtained using the 3D reconstructed phase data. The phase data includes a data matrix and a mask matrix. Invalid values ​​in the data matrix are discarded based on the valid values ​​in the mask matrix, and three new surface matrices are generated. The surface matrix is ​​scaled using the pixelation scale parameter obtained in step S12, and the scaled surface matrix is ​​translated to generate a new matrix. Finally, the processed matrix is ​​plotted as a 3D grid. Each grid is rotated and translated to adjust its position and angle in 3D space. A unified coordinate axis range is set and axis labels are added to obtain the original 3D image of the projectile.

[0023] Furthermore, for the surface damage of the high-speed moving target and the abnormal noise area in the edge stripe overlapping area, the edge data is obtained based on the captured projectile stripe image, the edge data is fused with the original 3D reconstruction data, and the fused phase data is used to obtain the projectile 3D image, including the following steps:

[0024] Step S31, for the surface damage of the high-speed moving target and the noise abnormality area of ​​the edge stripe overlap area, obtain the edge data based on the captured projectile stripe image;

[0025] In step S13, the coordinate system of the three cameras is calculated. The three cameras are numbered 25, 26, and 27 respectively. For camera 26, 10 pixel points (points 1 to 10) are evenly selected along the projectile outline on the projectile streak image captured by camera 26, and their pixel coordinates are recorded.

[0026] Mark the corresponding positions of the projectile streak images captured by cameras 25 and 27 with the 10 points selected by camera 26, labeled 1' to 10'. Since the angle between the three cameras is 120°, the projection position of the same projectile surface point in the images of different cameras will shift along the projectile's height. Therefore, the projectile outline edge captured by camera 26 will not correspond to the projectile outline edge in cameras 25 and 27. The projectile streak images captured by the three cameras need to be divided into six equal parts from the top to the bottom of the projectile. Since the distance from the top to the bottom of the projectile is 180°, each part is 30°. The angle difference between the three cameras is 120°. From the perspective of the projectile's upper and lower bisectors (180° range), 180° – 120° = 60°. This 60° is the adjustment angle derived from the geometric relationship between the projectile's bisector direction and the angle between the cameras. Since each part is 60°, 60° ÷ 30° = 2, so camera 25 is offset by 2 parts. The position corresponding to the upper edge of the projectile of camera 26 is two points below the upper edge of the projectile of camera 25. The same is true for camera 27. Therefore, there are 5 corresponding coordinate points between cameras 25 and 26, namely 1' to 5'; there are also 5 corresponding coordinate points between cameras 26 and 27, namely 6' to 10'.

[0027] Next, the binocular vision principle is applied to the pixel points 1 to 5 of the projectile streak image taken by camera 26, the pixel points 1' to 5' of the projectile streak image taken by camera 25, the pixel points 6 to 10 of the projectile streak image taken by camera 26, and the pixel points 6' to 10' of the projectile streak image taken by camera 27 to calculate the spatial coordinates of points 1 to 5 in the 25-26 coordinate system and points 6 to 10 in the 26-27 coordinate system.

[0028] Since points 1 to 5 of these 10 points are in the 25-26 coordinate system and points 6 to 10 are in the 26-27 coordinate system, coordinate transformation is required. The 26-27 coordinate system is converted to the world coordinate system, and then from the world coordinate system to the 25-26 coordinate system to obtain the spatial coordinates of points 6 to 10 in the 25-26 coordinate system. Finally, the spatial coordinates of points 1 to 10 in the 25-26 coordinate system are unified. These coordinates are the edge data of the projectile streak image.

[0029] Step S32: Fusing the edge data with the original 3D reconstruction data, and obtaining a 3D image of the projectile from the fused phase data.

[0030] Load the original 3D images of the three surfaces of the projectile obtained in step S23, then import the edge data solved in step S31 and use red solid dots as marks, and observe the position deviation between the projectile outline formed by the red dots and the edge outline of the original 3D image calculated by Fourier.

[0031] In order to ensure that the corresponding positions of the projectile red dot contour and the Fourier calculation results are consistent, it is necessary to unify the spatial coordinates of points 1 to 10 in step S31 in the 25-26 coordinate system to the coordinate system of the Fourier transform profilometry. To this end, three representative points in the three-dimensional coordinates of the red dot (the projectile head (one point) and the projectile tail (two points)) can be selected and marked as:

[0032]

[0033] Three points corresponding to the Fourier calculation results:

[0034]

[0035] Matching, using the minimum mean square error criterion, solves the rigid transformation matrix (rotation matrix R and translation vector t), and its mathematical model is as follows:

[0036]

[0037] The singular value decomposition (SVD) method is used to solve this problem, and finally the rigid transformation matrix is ​​obtained to align the 10 points calculated by binocular vision principle to the Fourier transform profilometry coordinate system:

[0038] X′=R·X+t#(2)

[0039] Obtain the R and t values ​​required to rigidly transform the red dot's outline to the projected coordinate system, where X represents the coordinate in the 25-26 coordinate system and X' represents the coordinate in the Fourier transform profilometry coordinate system. After this transformation, map points 1 to 10 from step S34 to the Fourier coordinate system to achieve data fusion. If there is still some deviation between the red dot and the projection after rotation and translation, repeat the above method and perform another rotation and translation until they are roughly aligned. Replace the coordinate data of the corresponding positions obtained by Fourier transform profilometry with the coordinate data of the 10 roughly aligned points to obtain the final fused phase data. Repeat step S23 to obtain a 3D image of the projectile from the fused phase data.

[0040] A method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, and steps 1 to 3 are performed respectively in three modules.

[0041] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

[0042] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

[0043] Compared with the existing technology, the present invention has the following significant advantages: 1) By integrating Fourier transform profilometry and binocular vision three-dimensional measurement technology, the spatial reconstruction accuracy and overall morphological integrity of complex curved surface targets are significantly improved. 2) A small number of key spatial points are used for coordinate system integration, and the fusion algorithm is efficient and concise, which is convenient for engineering implementation and has good versatility and stability. 3) Binocular vision has strong dynamic response capabilities, which effectively enhances the system's adaptability and measurement robustness to moving or partially occluded targets. 4) The reconstruction results have high fidelity in spatial positioning and morphological restoration, meeting the needs of subsequent high-precision engineering applications such as finite element simulation and ballistic analysis. 5) The fusion mechanism of multi-source data gives full play to the complementary advantages of various sensors in accuracy and field of view, providing solid technical support for high-precision three-dimensional imaging systems. This method improves the accuracy of feature point detection of high-speed moving targets from 70% to 90%. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a system flow chart of the present invention

[0045] Figure 2 This is the measurement system layout diagram of the present invention

[0046] Figure 3 This is the checkerboard calibration diagram of the present invention

[0047] Figure 4 Picture of the projectile captured by the CCD camera

[0048] Figure 5 Projectile mask image

[0049] Figure 6 Reconstructing graphs for MATLAB programs DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific implementation process described here is only used to explain this application and is not intended to limit this application.

[0051] like Figure 1 As shown, the present invention provides a method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, the steps of which are as follows:

[0052] Step 1: Use a high-speed moving projectile, such as a projectile fired from a 12.7mm ballistic gun, as the experimental carrier, build a multi-view measurement system for the three-dimensional surface of a high-speed moving target, shoot the fringe patterns of the reference plane and reference inclined plane, draw the phase-height curves of the three cameras, calculate the pixel ratio parameters and the phase-height coefficients of the three cameras, calibrate using a checkerboard, calculate the intrinsic and extrinsic parameter matrices of the three CCD high-resolution digital cameras, and then use a high-resolution camera to capture the projectile under stripe laser illumination to obtain the projectile stripe distribution images at multiple perspectives. The overall layout is as follows: Figure 2 The specific operations are as follows:

[0053] Tests were conducted at an outdoor shooting range, and a multi-view measurement system for the three-dimensional surface of a high-speed moving target was constructed. The multi-view measurement system consists of three identical CCD high-resolution digital cameras, a regular hexahedron overall bracket, and an adapter bracket. The regular hexahedron overall bracket is composed of six metal rods of uniform length connected by corner pieces and bolted together to form a standard regular hexahedron frame. A mounting base plate is located at the center of the three spacing surfaces. The mounting base plate has multiple evenly distributed mounting holes pre-machined on it. Each CCD high-resolution digital camera is connected to the mounting base plate via an adapter bracket.

[0054] Use a laser collimator to calibrate the optical axis so that the angle between the optical axes of each pair of cameras is 120°. This ensures that every point on the rotating surface to be measured falls within the overlapping area between the camera and projection lens fields of view. Run the multi-view measurement system to take pictures, and adjust the focal lengths of the three cameras until the pictures are clear.

[0055] Capture the fringe patterns of the reference plane and reference inclined surface, plot the phase-height curves of the three cameras, and calculate the pixelation ratio parameters and phase-height coefficients of the three cameras. Use three CCD high-resolution digital cameras to capture the fringe patterns of the reference plane and reference inclined surface. Sample and crop the fringe patterns (take the middle portion of the image row direction after sampling, the image size is 2048*64), and extract the phase to obtain the wrapped phase information of the reference plane and reference inclined surface. Use the height change from the reference plane to the reference inclined surface to determine the corresponding phase change. Plot the phase-height curves of the three cameras, calculate the pixelation ratio parameters of the XY and Z axes, the phase-height coefficients of the three cameras, and the coordinate system of the three camera combinations.

[0056] Calibration was performed using a single 25mm x 25mm checkerboard. Three high-resolution CCD digital cameras simultaneously photographed the same checkerboard, with each pair capturing the same square. By varying the shooting posture and position of the checkerboard, a total of 18 calibration images were captured, as shown in the figure. The phase-height coefficient was used for corner detection and camera internal and external parameter calculations, resulting in the intrinsic and extrinsic parameter matrices for the three high-resolution CCD digital cameras.

[0057] A timing controller is used to control the timing, and the software triggers the laser to emit laser light. Under the illumination of the stripe laser, a high-resolution camera is used to capture the high-speed moving projectile and obtain stripe distribution images of the projectile from multiple perspectives.

[0058] Step 2: Apply the FTP method to the projectile fringe distribution image to obtain the original 3D reconstructed phase data. Use the 3D reconstructed data to obtain the original 3D image of the projectile. The specific operations are as follows:

[0059] The stripe images of the projectile taken by the three cameras are sampled, and the coordinates of the starting points of the three measured object images that need to be cropped are recorded and calculated using a drawing board. The three measured object images and the reference plane fringe image are cropped at the same position of the original sampled image to obtain three sets of images (image size is 512*512). The FTP method is used to extract three sets of continuous phase images reflecting the phase value of each pixel position on the projectile surface from these three sets of fringe images. The three sets of cropped fringe images are subjected to grayscale analysis using threshold segmentation or phase amplitude method. A mask image that only retains the effective area of ​​the projectile at each viewing angle is generated and multiplied pixel by pixel with the continuous phase image extracted from each camera viewing angle to obtain the original three-dimensional reconstructed phase data of the three surfaces of the projectile taken by the camera.

[0060] The original 3D image of the projectile is obtained using 3D reconstructed phase data. The phase data includes a data matrix and a mask matrix. Based on the valid values ​​in the mask matrix, invalid values ​​in the data matrix are discarded, and three new surface matrices are generated. The surface matrix is ​​scaled using the pixelation scale parameter and translated to generate new matrices. Finally, the processed matrix is ​​plotted as a 3D grid. Each grid is rotated and translated to adjust its position and angle in 3D space. A uniform coordinate axis range is set and axis labels are added to obtain the original 3D image of the projectile.

[0061] Step 3: For the surface damage of the high-speed moving target and the abnormal noise area in the edge stripe overlapping area, obtain its edge data based on the captured projectile stripe image, fuse the edge data with the original 3D reconstruction data, and obtain the projectile 3D image from the fused phase data. The specific operations are as follows:

[0062] The three cameras are numbered 25, 26, and 27 respectively. For camera 26, 10 pixel points, i.e. points 1 to 10, are evenly selected along the outline of the projectile on the projectile streak image taken by camera 26, and their pixel coordinates are recorded.

[0063] Mark the corresponding positions of the projectile streak images captured by cameras 25 and 27 with the 10 points selected by camera 26, labeled 1' to 10'. Since the angle between the three cameras is 120°, the projection position of the same projectile surface point in the images of different cameras will shift along the projectile's height. Therefore, the projectile outline edge captured by camera 26 will not correspond to the projectile outline edge in cameras 25 and 27. The projectile streak images captured by the three cameras need to be divided into six equal parts from the top to the bottom of the projectile. Since the distance from the top to the bottom of the projectile is 180°, each part is 30°. The angle difference between the three cameras is 120°. From the perspective of the projectile's upper and lower bisectors (180° range), 180° – 120° = 60°. This 60° is the adjustment angle derived from the geometric relationship between the projectile's bisector direction and the angle between the cameras. Since each part is 60°, 60° ÷ 30° = 2, so camera 25 is offset by 2 parts. The position corresponding to the upper edge of the projectile of camera 26 is two points below the upper edge of the projectile of camera 25. The same is true for camera 27. Therefore, there are 5 corresponding coordinate points between cameras 25 and 26, namely 1' to 5'; there are also 5 corresponding coordinate points between cameras 26 and 27, namely 6' to 10'.

[0064] Next, the binocular vision principle is applied to the pixel points 1 to 5 of the projectile streak image taken by camera 26, the pixel points 1' to 5' of the projectile streak image taken by camera 25, the pixel points 6 to 10 of the projectile streak image taken by camera 26, and the pixel points 6' to 10' of the projectile streak image taken by camera 27 to calculate the spatial coordinates of points 1 to 5 in the 25-26 coordinate system and points 6 to 10 in the 26-27 coordinate system.

[0065] Since points 1 to 5 of these 10 points are in the 25-26 coordinate system and points 6 to 10 are in the 26-27 coordinate system, coordinate transformation is required. The 26-27 coordinate system is converted to the world coordinate system, and then from the world coordinate system to the 25-26 coordinate system to obtain the spatial coordinates of points 6 to 10 in the 25-26 coordinate system. Finally, the spatial coordinates of points 1 to 10 in the 25-26 coordinate system are unified. These coordinates are the edge data of the projectile streak image.

[0066] In the next step, the edge data is fused with the original 3D reconstruction data, and the 3D image of the projectile is obtained from the fused phase data.

[0067] Load the original 3D images of the three sides of the projectile, then import the calculated edge data and use red solid dots as marks, and observe the position deviation between the projectile outline formed by the red dots and the edge outline of the original 3D image calculated by Fourier.

[0068] In order to ensure that the corresponding positions of the projectile red dot outline and the Fourier calculation results are consistent, it is necessary to unify the spatial coordinates of points 1 to 10 calculated using the binocular vision principle in the 25-26 coordinate system to the coordinate system of the Fourier transform profilometry. To this end, three representative points in the three-dimensional coordinates of the red dot (the projectile head (one point) and the projectile tail (two points)) can be selected and marked as:

[0069]

[0070] Three points corresponding to the Fourier calculation results:

[0071]

[0072] Matching, using the minimum mean square error criterion, solve the rigid body transformation matrix (rotation matrix R and translation vector t), its mathematical model is as follows:

[0073]

[0074] The singular value decomposition (SVD) method is used to solve this problem, and finally the rigid transformation matrix that aligns the 10 points calculated by binocular vision principle to the FTP point cloud is obtained:

[0075] X′=R·X+t#(2)

[0076] Obtain the R and t values ​​required to rigidly transform the red dot's outline into the projected coordinate system, where X represents the coordinate in the 25-26 coordinate system and X' represents the coordinate in the Fourier transform profilometry coordinate system. After this transformation, map points 1 to 10, calculated using binocular vision principles, to the Fourier coordinate system to achieve data fusion. If there is still some deviation between the red dot and the projection after rotation and translation, repeat the above method and perform another rotation and translation until they are roughly aligned. Replace the coordinate data of the corresponding positions obtained by Fourier transform profilometry with the coordinate data of the 10 roughly aligned points to obtain the final fused phase data. Repeat the steps in step 2 to obtain a 3D image of the projectile using the phase data to obtain the final 3D image of the projectile.

[0077] The present invention also proposes a method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection. The method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, and steps 1 to 3 are executed respectively in three modules.

[0078] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

[0079] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

[0080] In summary, the high-speed projectile detection model constructed by the present invention achieves a good balance between the accuracy, real-time performance and computational efficiency of target reconstruction, and provides a solution to the calculation noise anomalies existing in the surface damage areas and edge stripe overlapping areas of high-speed moving targets.

[0081] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0082] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, characterized in that: The following steps are involved: Step 1: Build a multi-view measurement system for the three-dimensional surface of a high-speed moving target. Capture fringe patterns on reference planes and reference inclined surfaces, plot phase-height curves for the three cameras, calculate pixelation scale parameters and phase-height coefficients for the three cameras, calibrate using a checkerboard grid, and calculate the intrinsic and extrinsic parameter matrices for the three CCD high-resolution digital cameras. Then, use the high-resolution cameras to capture the projectile under fringe laser illumination and obtain fringe distribution images of the projectile at multiple viewpoints. Step 2: Apply the FTP method to the projectile fringe distribution image to obtain the original 3D reconstructed phase data, and use the 3D reconstructed data to obtain the original 3D image of the projectile; Step 3: For the surface damage of the high-speed moving target and the abnormal noise area in the edge stripe overlapping area, obtain its edge data based on the captured projectile stripe image, fuse the edge data with the original three-dimensional reconstruction data, and obtain the projectile 3D image from the fused phase data.

2. The method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection according to claim 1, characterized in that: Step 1: Build a multi-view measurement system for the three-dimensional surface of a high-speed moving target, capture fringe patterns on a reference plane and a reference inclined plane, plot phase-height curves for the three cameras, calculate pixelation ratio parameters and phase-height coefficients for the three cameras, calibrate using a checkerboard grid, calculate the intrinsic and extrinsic parameter matrices for the three CCD high-resolution digital cameras, and then use high-resolution cameras to capture a high-speed moving projectile under fringe laser illumination to obtain fringe distribution images from multiple perspectives of the projectile. This includes the following steps: Step S11, building a multi-view measurement system for the three-dimensional shape of a high-speed moving target; The multi-view measurement system includes three identical CCD high-resolution digital cameras, a hexahedron overall bracket, and an adapter bracket. The hexahedron overall bracket is composed of six metal rods of the same length fastened together by bolts through corner pieces to form a standard hexahedron frame. A mounting base plate is provided at the center of the three spacing surfaces. The mounting base plate is pre-machined with multiple evenly distributed mounting holes. Each CCD high-resolution digital camera is connected to the mounting base plate via an adapter bracket. Use a laser collimator to calibrate the optical axis so that the angle between the optical axes of each pair of cameras is 120°. This ensures that every point on the rotating surface to be measured falls within the overlapping area between the camera and projection lens fields of view. Run the multi-view measurement system to take pictures, and adjust the focal lengths of the three cameras until the pictures are clear. Step S12, photographing fringe patterns of the reference plane and the reference inclined plane, plotting phase-height curves of the three cameras, and calculating pixel ratio parameters and phase-height coefficients of the three cameras; Three CCD high-resolution digital cameras were used to capture fringe patterns of a reference plane and a reference inclined surface. The fringe patterns were sampled, cropped, and phase extracted to obtain the wrapped phase information of the reference plane and the reference inclined surface. The corresponding phase change was obtained by using the height change from the reference plane to the reference inclined surface. Phase-height curves of the three cameras were plotted, and the pixel ratio parameters of the X, Y, and Z axes, as well as the phase-height coefficients of the three cameras, were calculated. Step S13, using a checkerboard to perform calibration, and calculating the intrinsic parameter matrix and extrinsic parameter matrix of the three CCD high-resolution digital cameras; Using a chessboard for calibration, three CCD high-resolution digital cameras simultaneously photograph the same chessboard. By changing the shooting posture and position of the chessboard, multiple sets of calibration images are taken. The phase-height coefficient is used for corner detection and camera internal and external parameter calculation to obtain the intrinsic and external parameter matrices of the three CCD high-resolution digital cameras. Step S14, using a high-resolution camera to capture the high-speed moving projectile under stripe laser illumination, and obtaining stripe distribution images of the projectile at multiple viewing angles; The experiment was carried out at an outdoor shooting range. Under stripe laser illumination, a CCD high-resolution digital camera was used to capture the stripe images of the projectile passing through the field of view.

3. The method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection according to claim 1, characterized in that: Step 2: Apply the FTP method to the projectile fringe distribution image to obtain the original three-dimensional reconstructed phase data, and use the three-dimensional reconstructed data to obtain the original 3D image of the projectile, including the following steps: Step S21, sampling and cutting the fringe images of the high-speed moving projectile captured by the three cameras to obtain three sets of fringe images; Step S22: Using the FTP method, the three sets of fringe images are first extracted to produce three sets of continuous phase maps reflecting the phase value of each pixel position on the projectile surface. A grayscale analysis is performed on the three sets of fringe images using a threshold segmentation or phase amplitude method to generate a mask image that retains only the effective area of ​​the projectile at each viewing angle. The mask image is then multiplied pixel by pixel with the continuous phase map extracted at each camera viewing angle to obtain the original three-dimensional reconstructed phase data of the three surfaces of the projectile captured by the camera. Step S23, using the three-dimensional reconstructed phase data to obtain the original 3D image of the projectile, the phase data contains a data matrix and a mask matrix, according to the valid values ​​in the mask matrix, the invalid values ​​in the data matrix are discarded, and three new surface matrices are generated, the surface matrix is ​​scaled using the pixelation ratio parameter, and the scaled surface matrix is ​​translated to generate a new matrix, the processed matrix is ​​drawn as a 3D grid graph, each grid graph is rotated and translated to adjust its position and angle in the 3D space, a unified coordinate axis range is set and coordinate axis labels are added to obtain the original 3D image of the projectile.

4. The method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection according to claim 1, wherein step 3: for the surface damage of the high-speed moving target and the noise abnormality area in the edge stripe overlapping area, edge data is obtained based on the captured projectile stripe image, the edge data is fused with the original 3D reconstruction data, and a 3D image of the projectile is obtained from the fused phase data, comprising the following steps: Step S31, for the surface damage of the high-speed moving target and the noise abnormality area of ​​the edge stripe overlap area, obtain the edge data based on the captured projectile stripe image; The three cameras are numbered 25, 26, and 27 respectively. For camera 26, 10 pixel points, i.e. points 1 to 10, are evenly selected along the outline of the projectile on the projectile streak image taken by camera 26, and their pixel coordinates are recorded. The corresponding positions of the projectile stripe images captured by cameras 25 and 27 are marked with 10 corresponding points selected by camera 26, and are recorded as 1' to 10'. Since the angle between the three cameras is 120 degrees, the projection position of the same projectile surface point in the images of different cameras will be offset along the height direction of the projectile. Therefore, the projectile contour edge of camera 26 does not correspond to the projectile contour edge in cameras 25 and 27. The projectile stripe images captured by the three cameras need to be divided into 6 parts along the height direction of the projectile. Since the angle from the upper edge to the lower edge of the projectile is 180 degrees, each part is 30 degrees. The angle between the three cameras is 120 degrees. The angle difference is 120°. From the perspective of the projectile's upper and lower bisection directions, 180° – 120° = 60°. This 60° is the adjustment angle derived from the geometric relationship between the projectile's bisection direction and the camera's angle. Since each section is 60°, 60° ÷ 30° = 2. Therefore, camera 25 is offset by 2 sections. The position corresponding to the upper edge of the projectile of camera 26 is two sections below the upper edge of the projectile of camera 25. The same is true for camera 27. Therefore, camera 25 has 5 coordinate points corresponding to camera 26, namely 1' to 5'; camera 27 also has 5 coordinate points corresponding to camera 26, namely 6' to 10'. Apply binocular vision principle to calculate the spatial coordinates of pixel points 1 to 5 of the projectile streak image taken by camera 26, pixel points 1' to 5' of the projectile streak image taken by camera 25, pixel points 6 to 10 of the projectile streak image taken by camera 26, and pixel points 6' to 10' of the projectile streak image taken by camera 27 in the 25-26 coordinate system and the 26-27 coordinate system respectively; Since points 1 to 5 are in the 25-26 coordinate system and points 6 to 10 are in the 26-27 coordinate system, coordinate transformation is required. The 26-27 coordinate system is transformed into the world coordinate system, and then the world coordinate system is transformed into the 25-26 coordinate system. The spatial coordinates of points 6 to 10 in the 25-26 coordinate system are obtained, and finally the spatial coordinates of points 1 to 10 in the 25-26 coordinate system are obtained. These coordinates are the edge data of the projectile fringe image. Step S32, fusing the edge data with the original 3D reconstructed data, and obtaining a 3D image of the projectile from the fused phase data; Load the original 3D images of the three sides of the projectile, then import the edge data calculated in step S31 and mark them with red solid dots. Observe the position deviation between the projectile outline formed by the red dots and the edge outline of the original 3D image calculated by Fourier transform. In order to ensure that the corresponding positions of the projectile red dot contour and the Fourier calculation results are consistent, it is necessary to unify the spatial coordinates of points 1 to 10 in step S31 in the 25-26 coordinate system to the coordinate system of the Fourier transform profilometry. To this end, three representative points in the three-dimensional coordinates of the red dot are selected, namely one point at the projectile head and two points at the projectile tail, and marked as: Three points corresponding to the Fourier calculation results: Matching, using the minimum mean square error criterion, solves the rigid body transformation matrix, including the rotation matrix R and translation vector t. Its mathematical model is as follows: The singular value decomposition method is used to solve this problem, and finally the rigid transformation matrix is ​​obtained to align points 1 to 10 calculated using the binocular vision principle to the Fourier transform profilometry coordinate system: X′=R·X+t#(2) The above formula is used to obtain the R and t required for the rigid transformation of the red dot outline to the projection coordinate system, where X represents the coordinate in the 25-26 coordinate system and X' represents the coordinate in the Fourier transform profilometry coordinate system. Points 1 to 10 are mapped to the Fourier coordinate system after this transformation to achieve data fusion. If there is still some deviation between the red dot and the projection after rotation and translation, the above method can be repeated for another rotation and translation until they are roughly aligned. The coordinate data of the 10 roughly aligned points are used to replace the coordinate data of the corresponding positions obtained by Fourier transform profilometry to obtain the final fused phase data.

5. A method for transient 3D reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, characterized in that: Implement the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection as described in any one of claims 1-4 to realize transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection, and perform steps 1 to 3 in three modules respectively.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection according to any one of claims 1 to 4 is implemented to achieve transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

7. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection as described in any one of claims 1 to 4, thereby realizing transient three-dimensional reconstruction of high-speed moving targets based on Fourier transform profilometry and binocular intersection.

Images