Complex component large-field-of-view measurement method based on phase matching and global mark point splicing

Abstract

The invention discloses a complex component large-view-field measurement method based on phase matching and global mark point splicing, and belongs to the field of three-dimensional measurement. Comprising a small-field-of-view high-spatial-resolution binocular fringe projection local measurement system which obtains single-field-of-view high-precision point cloud data on the surface of a component based on a heterodyne multi-frequency phase shift principle and high-dynamic measurement; the large-view-field low-spatial-resolution system is used for tracking and positioning the local measurement system and realizing point cloud splicing of the local area of the large component; realizing point cloud splicing of a plurality of local areas of the component based on rotary table global mark points calibrated by a combined laser tracker and a visual system; and on the basis of global optimization of trimmed ICP, multi-view-field point cloud precise registration with adaptive overlapping rate is realized, and accumulative errors are eliminated. According to the invention, large-field-of-view automatic measurement of high-light and weak-texture surface industrial components can be realized, and compared with a traditional method based on mark point splicing, the system can effectively avoid transmission errors caused by multi-field-of-view splicing, and the measurement precision and efficiency are greatly improved.

Description

Technical Field

[0001] The invention belongs to the technical field of three-dimensional measurement, and relates to a large-field-of-view measurement method for complex components based on phase matching and global marker point splicing. Background Art

[0002] The visual measurement system based on fringe projection has been widely used in industrial product manufacturing and inspection due to its non-contact and high efficiency advantages. However, the current commercial 3D point cloud measurement equipment is insufficient for measuring complex structures. When complex highlights appear on the surface of the component, the measurement accuracy is severely reduced or even impossible to measure. In addition, when industrial components have characteristics such as large size and weak texture, feature-based point cloud stitching cannot be used, and the traditional point cloud stitching method based on landmark points has large transmission errors, low measurement efficiency, and poor stitching accuracy.

[0003] Therefore, a large-field-of-view measurement method for complex components is urgently needed to complete the spray-free measurement of high-gloss surfaces and the flexible splicing of large-size weak-textured surfaces, realize the high-precision fusion of multi-field-of-view point cloud data in a large measurement space, and improve the measurement splicing efficiency and accuracy. Summary of the invention

[0004] In view of the above defects or improvement needs of the existing registration technology, the present invention provides a large field of view measurement method for complex components based on phase matching and global landmark point stitching, thereby solving the problems of the existing three-dimensional measurement methods in the measurement of large complex components, such as the inability to measure high-gloss surfaces and the low efficiency and poor accuracy of point cloud stitching of large-size weak-texture surfaces.

[0005] To achieve the above object, the present invention provides a method for measuring a complex component with a large field of view based on phase matching and global marker point stitching, comprising:

[0006] S1. The local measurement system with small field of view and high spatial resolution projects a single field of view 3D point cloud of multi-brightness multi-frequency grating fine stripes to reconstruct the components; the coarse stripe image is continued to be projected, and the left and right cameras and the global measurement system camera with large field of view and low spatial resolution synchronously acquire the image;

[0007] S2. By projecting the fringe projection image obtained by coarse fringe, the local system point cloud is unified to the global system coordinate system, cross-scale multi-system stitching based on phase matching is realized, and the stitching point cloud in the local area of ​​the component is obtained;

[0008] S3. Based on the global marker points of the turntable calibrated by the joint laser tracker and the visual system, the three-dimensional coordinates of the marker points of the current absolute position of the turntable are obtained according to the global system, and they are registered to the coordinate system of the global marker points of the turntable; the turntable is rotated to different absolute positions to realize the splicing of point clouds between multiple local areas;

[0009] S4. Based on the trimmed ICP algorithm, precise registration between two point clouds is performed, and multiple rounds of iterations are performed until the rigid body transformation matrix of all viewing angles is close to the unit matrix, thus achieving global optimization of the adaptive overlap rate.

[0010] Preferably, the local system projects fine stripes to realize the single field of view point cloud based on heterodyne multi-frequency phase shift principle and high dynamic measurement, and projects coarse stripes to trigger the local system and the global system to take pictures synchronously, specifically:

[0011] S11. A local measurement system with a small field of view and high spatial resolution projects multiple sets of multi-frequency grating fine stripe images with different brightness, and only the left and right cameras of the local system take pictures;

[0012] S12. retaining the result with the largest modulation degree in each pixel multi-brightness fringe projection map, and solving to obtain a single-view three-dimensional point cloud of the reconstructed component surface;

[0013] S13. The local system continues to project the multi-frequency grating coarse stripe image, and the local system and the global system camera with a large field of view and low spatial resolution take pictures synchronously.

[0014] Preferably, based on phase matching cross-scale multi-system stitching, a stitching point cloud in a local area of ​​a component is obtained, specifically:

[0015] S21. Based on the fringe projection image obtained by projecting coarse fringes, the phase values ​​of the component surface points under all cameras are solved, and the same-name points taken by the local system and the global system cameras are searched. For a global system with two cameras, it is expressed as: ||PP i || Φ <ε,

[0016] Where P is the absolute phase of a point on the object being measured, P i is the phase value of the image point on the four cameras, ||·|| Φ represents the Euclidean distance in the absolute phase domain, and ε is the threshold for controlling similarity;

[0017] S22. According to the principle of maximum effective phase point number and the principle of optimal binocular structure parameters, two global system cameras are selected to form the optimal global binocular system;

[0018] S23. Use the local system and the optimal global binocular system to reconstruct the three-dimensional points respectively. The coordinates of the same-name points in the two system coordinate systems are used to optimize the external parameters between the two systems. Combined with the calibrated external parameters between the cameras of the global system, the local system point cloud is unified to the global system coordinate system to achieve the splicing of the point cloud in the local area of ​​the component.

[0019] Preferably, the point clouds between multiple local areas are spliced ​​based on the global landmarks of the turntable calibrated by the combined laser tracker and the visual system, specifically:

[0020] S31. Fix the laser tracker target base evenly at different positions on the turntable, place the target ball on the target base, measure the center coordinates of the target ball by the laser tracker, obtain the coordinates of the center of all the balls on the target base in the tracker coordinate system, and construct a global reference point set;

[0021] S32. Replace the target ball with the tracker-photogrammetry conversion ball, and measure the coordinates of the center of the conversion ball and the coordinates of the marker points on the turntable surface synchronously by the global system, and align the coordinates of the center of the conversion ball to the global reference point set described in S31. Rotate the turntable and repeat the above steps to finally calibrate the global marker points of the turntable;

[0022] S33. At the current absolute position of the turntable, the global system shoots the turntable marker point, obtains the coordinates of the current marker point relative to the global system coordinate system, and aligns it to the turntable global marker point, and obtains the conversion matrix from the current global system to the turntable global marker point coordinate system;

[0023] S34. Based on the conversion matrix described in S33, transform the spliced ​​point cloud in the local area of ​​the component, rotate the turntable and repeat S33, unify all the local area point clouds into the turntable global marker point coordinate system, and realize the splicing of point clouds between multiple local areas of the component.

[0024] Preferably, the global optimization of the trimmed ICP algorithm is used to realize the precise registration of multi-field point clouds with adaptive overlap rate, specifically:

[0025] S41. In each round of iteration, starting from the first perspective, the overlap rate between the first perspective and the overall point cloud of all other perspectives is calculated, and the rigid body transformation matrix of the point cloud data of each perspective to the overall point cloud data of all other perspectives is calculated using the trimmed ICP algorithm;

[0026] S42. After the calculation is completed, the coordinate transformation of the point cloud data of the view is performed, and then the rigid body transformation matrix from the next view to all other view point cloud data is calculated, and the coordinate transformation is performed;

[0027] S43. After the first round of calculation and transformation of all perspectives is completed, the second and third rounds of calculation and transformation are continued until the rigid body transformation matrices of all perspectives are close to the unit matrix, and the overall splicing optimization is completed.

[0028] The technical solution provided by the present invention has the following beneficial effects:

[0029] In the present invention, a method for measuring complex components with a large field of view based on phase matching and global marker point stitching is proposed. First, a local system with a small field of view and high spatial resolution projects multi-brightness fine stripes, and based on the heterodyne multi-frequency phase shift principle and high dynamic measurement, single-field high-precision point cloud data acquisition of the component surface is realized, and coarse stripes are projected to trigger the local system and the global system to synchronously take pictures. Using the coarse stripe projection image, cross-scale multi-system stitching based on phase matching is realized to obtain the stitching point cloud in the local area of ​​the component. Then, based on the global marker points of the turntable calibrated by the joint laser tracker and the visual system, the stitching of point clouds between multiple local areas is realized, and the global optimization based on the trimmed ICP algorithm is used to realize the precise registration of multi-field point clouds with adaptive overlap rate, and obtain a complete high-precision point cloud. Various experiments have proved that compared with the traditional point cloud measurement method based on stripe projection and the point cloud stitching method based on feature points, the algorithm of the present invention can realize the spray-free measurement of the high-gloss surface of industrial components, as well as the flexible stitching of large-size weak-textured surfaces, realize the high-precision fusion of multi-field point cloud data in a large measurement space, and improve the measurement stitching efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic flow chart of a method for measuring a large field of view of a complex component based on phase matching and global marker point splicing provided by the present invention.

[0031] Figure 2 This is a schematic diagram of the coarse fringe projection image captured synchronously by the local system and the global system in the present invention.

[0032] Figure 3 Schematic diagram of searching for the same-name points in the local system and global system camera fringe projection images in the present invention.

[0033] Figure 4 The diagram is a schematic diagram of a complex component large field of view measurement system based on phase matching and global marker point stitching in the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose and technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0035] refer to Figure 1 The embodiment of the present invention provides a method for measuring a complex component with a large field of view based on phase matching and global marker point stitching, which specifically includes the following steps:

[0036] S1. The local measurement system with small field of view and high spatial resolution projects a single-field 3D point cloud of multi-brightness and multi-frequency grating fine stripes to reconstruct components; the coarse stripe image is continued to be projected, and the left and right cameras and the global measurement system camera with large field of view and low spatial resolution synchronously acquire the image.

[0037] Specifically, the local system projects fine stripes to realize the single field of view point cloud based on the heterodyne multi-frequency phase shift principle and high dynamic measurement, and projects coarse stripes to trigger the local system and the global system to take pictures synchronously. The specific process is as follows:

[0038] S11. The local measurement system with small field of view and high spatial resolution projects multiple sets of multi-frequency grating fine stripe images with different brightness from dark to bright, and only the left and right cameras of the local system take pictures.

[0039] S12. Calculate the modulation degree in the projection map of the fringe with different brightness for each pixel, expressed as:

[0040]

[0041] Where k is the brightness level, I″ k (x, y) is the modulation degree at the k brightness level, i is the phase shift number, and the value is 0, 1, 2, 3, is the brightness of the fringe pattern collected at brightness level k.

[0042] The result with the largest modulation degree of the multi-brightness fringe projection image of each pixel is retained, the phase-shifted fringe image is synthesized, and the single-view three-dimensional point cloud of the reconstructed component surface is obtained by solving the solution.

[0043] S13. The local system continues to project the multi-frequency grating coarse stripe image, and the local system and the global system camera with large field of view and low spatial resolution take pictures synchronously, such as Figure 2 shown.

[0044] S2. By projecting the fringe projection image obtained by coarse stripes, the local system point cloud is unified to the global system coordinate system, cross-scale multi-system stitching based on phase matching is realized, and the stitching point cloud in the local area of ​​the component is obtained.

[0045] Specifically, based on phase matching, cross-scale multi-system stitching is used to obtain the stitching point cloud in the local area of ​​the component. The specific process is as follows:

[0046] S21. Based on the fringe projection image obtained by projecting the coarse fringe, the phase values ​​of the component surface points under all cameras are solved, and the same-name points captured by the local system and the global system cameras are searched, such as Figure 3 As shown. For a global system with two cameras, the solution for the same-name point is expressed as:

[0047]

[0048] Where P is the absolute phase of a point on the object being measured, P i is the phase value of the image point on the four cameras, ||·|| Φ represents the Euclidean distance in the absolute phase domain, and ε is the threshold for controlling similarity.

[0049] S22. According to the principle of maximum effective phase point number and the principle of optimal binocular structure parameters, two global system cameras are selected to form the optimal global binocular system.

[0050] Specifically, the camera field of view of the global positioning network is large, while the projection field of view of the local system is small. Its projected fringe image only occupies the local area of ​​the global camera. Therefore, the best tracking camera is determined by counting the number of points of the global camera that fall into the value range of the local camera. Taking the left camera of the local system as the reference camera, firstly open its phase mask to remove small outliers; then, within the preprocessed mask range, search for phase extreme values ​​Φ xMAX , Φ xMIN , Φ yMAX , Φ yMIN Finally, the number of effective phase pixels n is counted Φ , where the effective phase must satisfy the following formula,

[0051] Φ xi ∈(Φ xMIN , Φ xMAX )∩Φ yi ∈(Φ yMIN , Φ yMAX )

[0052] For each camera n Φ The cameras with more effective phase pixels have a higher weight as the optimal tracking camera. After determining the left camera of the optimal tracking camera pair, when there are multiple cameras with similar effective phase points that meet the matching requirements, the camera is selected so that the binocular system structure parameters formed by it and the optimal tracking left camera are more accurate.

[0053] S23. Use the local system and the optimal global binocular system to reconstruct the three-dimensional points respectively. The coordinates of the same-name points in the two system coordinate systems are optimized to obtain the external parameters between the two systems. The optimization function is:

[0054]

[0055] Among them, C1 and C2 are the point clouds reconstructed by the optimal global binocular system and the local system, respectively. 1 i is a point in the optimal global binocular system, P 2 i YesP 1 i At points of the same name in the local system, and is the rotation matrix and translation vector from the local system to the optimal global binocular system.

[0056] The initial value of the optimization function can be obtained using the quaternion method, and then the position and posture combination is further optimized through the ICP algorithm. Finally, the external parameters between the cameras of the global system are calibrated, and the local system point cloud is unified to the global system coordinate system to achieve the splicing of the point cloud in the local area of ​​the component.

[0057] S3. Based on the global marker points of the turntable calibrated by the joint laser tracker and the visual system, the three-dimensional coordinates of the marker points of the current absolute position of the turntable are obtained according to the global system, and they are aligned to the coordinate system of the global marker points of the turntable; the turntable is rotated to different absolute positions to realize the stitching of point clouds between multiple local areas.

[0058] Specifically, the global landmarks of the turntable calibrated by the combined laser tracker and the visual system are used to realize the splicing of point clouds between multiple local areas. The specific process is as follows:

[0059] S31. Fix the laser tracker target base evenly at different positions on the turntable, place the target ball on the target base, measure the center coordinates of the target ball with the laser tracker, obtain the coordinates of the center of all balls on the target base in the tracker coordinate system, and construct a global reference point set.

[0060] S32. Replace the target ball with the tracker-photogrammetry conversion ball, and measure the three-dimensional coordinates of the center of the marker point on the conversion ball by the global system. At this time, the coordinates of the center of the conversion ball marker point are consistent with the coordinates of the target ball center. Synchronously shoot the marker points on the turntable surface, solve the coordinates of the center of the marker point, align the coordinates of the center of the conversion ball to the global reference point set described in S31, and synchronously convert the marker points on the turntable surface to the global reference point set coordinate system. Turn the turntable and repeat the above steps to finally achieve the calibration of the turntable global marker points.

[0061] S33. During the measurement process, at the current absolute position of the turntable, the global system photographs the turntable marker point, obtains the coordinates of the current marker point relative to the global system coordinate system, and aligns it to the turntable global marker point, and obtains the transformation matrix from the current global system to the turntable global marker point coordinate system.

[0062] S34. Based on the transformation matrix described in S33, transform the spliced ​​point cloud in the local area of ​​the component, rotate the turntable and repeat S33, such as Figure 4 As shown, all local area point clouds are unified into the turntable global landmark point coordinate system to realize the splicing of point clouds between multiple local areas of the component.

[0063] S4. Based on the trimmed ICP algorithm, precise registration between two point clouds is performed, and multiple rounds of iterations are performed until the rigid body transformation matrix of all viewing angles is close to the unit matrix, thus achieving global optimization of the adaptive overlap rate.

[0064] Specifically, the global optimization of the trimmed ICP algorithm is used to achieve the multi-field point cloud precise registration with adaptive overlap rate. The specific process is as follows:

[0065] S41. In each iteration, starting from the first view, set a distance threshold associated with the average point spacing, find the nearest neighbor point in the point cloud of this view in the overall point cloud of all other viewpoints, and count the proportion of points whose distances are within the distance threshold to the total number of points as the overlap rate. Use the overlap rate as the proportion of matching point pairs to be retained, perform point-to-surface ICP on the point cloud of this viewpoint and the overall point cloud of all other viewpoints, and calculate the rigid body transformation matrix of the two. The point-to-surface ICP optimization objective function is:

[0066]

[0067] Among them, p i is a point in the point cloud of the viewpoint, q i is the p in the overall point cloud of all other perspectives i matching points, n i For point q i The normal vector of , R is the required rotation matrix, and t is the required translation vector.

[0068] S42. After the calculation is completed, the coordinate transformation of the point cloud data of the viewing angle is performed, and then the rigid body transformation matrix from the next viewing angle to all other viewing angle point cloud data is calculated and the coordinate transformation is performed.

[0069] S43. After the first round of calculation and transformation of all perspectives is completed, the second and third rounds of calculation and transformation are continued until the rigid body transformation matrices of all perspectives are close to the unit matrix, and the overall splicing optimization is completed.

Claims

1. A method for measuring complex components with a large field of view based on phase matching and global landmark stitching, characterized in that: include: S1. The local measurement system with small field of view and high spatial resolution projects a single field of view 3D point cloud of multi-brightness multi-frequency grating fine stripes to reconstruct the components; the coarse stripe image is continued to be projected, and the left and right cameras and the global measurement system camera with large field of view and low spatial resolution synchronously acquire the image; S2. By projecting the fringe projection image obtained by coarse fringe, the local system point cloud is unified to the global system coordinate system, cross-scale multi-system stitching based on phase matching is realized, and the stitching point cloud in the local area of ​​the component is obtained; S3. Based on the global marker points of the turntable calibrated by the joint laser tracker and the visual system, the three-dimensional coordinates of the marker points of the current absolute position of the turntable are obtained according to the global system, and they are registered to the coordinate system of the global marker points of the turntable; the turntable is rotated to different absolute positions to realize the splicing of point clouds between multiple local areas; S4. Based on the trimmed ICP algorithm, precise registration between two point clouds is performed, and multiple rounds of iterations are performed until the rigid body transformation matrix of all viewing angles is close to the unit matrix, thus achieving global optimization of the adaptive overlap rate.

2. The method according to claim 1, characterized in that: As described in S1, a local measurement system with a small field of view and high spatial resolution projects a multi-brightness fine stripe image to obtain a component single field of view three-dimensional point cloud, and continues to project a coarse stripe image to trigger the local system to synchronously take pictures with a global system camera with a large field of view and low spatial resolution, specifically including: S11. A local measurement system with a small field of view and high spatial resolution projects multiple sets of multi-frequency grating fine stripe images with different brightness, and only the left and right cameras of the local system take pictures; S12. retaining the result with the largest modulation degree in each pixel multi-brightness fringe projection map, and solving to obtain a single-view three-dimensional point cloud of the reconstructed component surface; S13. The local system continues to project the multi-frequency grating coarse stripe image, and the local system and the global system camera with a large field of view and low spatial resolution take pictures synchronously.

3. The method according to claim 1, characterized in that The cross-scale local and global system point cloud stitching described in S2 is realized based on phase matching to obtain the stitching point cloud in the local area of ​​the component, specifically including: S21. Based on the fringe projection image obtained by projecting the coarse fringe, the phase values ​​of the component surface points under all cameras are solved, and the same-name points taken by the local system and the global system cameras are searched. For a global system with two cameras, it is expressed as: Where P is the absolute phase of a point on the object being measured, P i is the phase value of the image point on the four cameras, ||·|| Φ represents the Euclidean distance in the absolute phase domain, and ε is the threshold for controlling similarity; S22. According to the principle of maximum effective phase point number and the principle of optimal binocular structure parameters, two global system cameras are selected to form the optimal global binocular system; S23. Use the local system and the optimal global binocular system to reconstruct the three-dimensional points respectively. The coordinates of the same-name points in the two system coordinate systems are used to optimize the external parameters between the two systems. Combined with the calibrated external parameters between the cameras of the global system, the local system point cloud is unified to the global system coordinate system to achieve the splicing of the point cloud in the local area of ​​the component.

4. The method according to claim 1, characterized in that: The global marker points of the turntable calibrated by the combined laser tracker and the visual system described in S3 realize the splicing of point clouds between multiple local areas, specifically including: S31. Fix the laser tracker target base evenly at different positions on the turntable, place the target ball on the target base, measure the center coordinates of the target ball by the laser tracker, obtain the coordinates of the center of all the balls on the target base in the tracker coordinate system, and construct a global reference point set; S32. Replace the target ball with the tracker-photogrammetry conversion ball, and measure the coordinates of the center of the conversion ball and the coordinates of the marker points on the turntable surface synchronously by the global system, and align the coordinates of the center of the conversion ball to the global reference point set described in S31. Rotate the turntable and repeat the above steps to finally calibrate the global marker points of the turntable; S33. At the current absolute position of the turntable, the global system shoots the turntable marker point, obtains the coordinates of the current marker point relative to the global system coordinate system, and aligns it to the turntable global marker point, and obtains the conversion matrix from the current global system to the turntable global marker point coordinate system; S34. Based on the conversion matrix described in S33, transform the spliced ​​point cloud in the local area of ​​the component, rotate the turntable and repeat S33, unify all the local area point clouds into the turntable global marker point coordinate system, and realize the splicing of point clouds between multiple local areas of the component.

5. The method according to claim 1, characterized in that The global optimization based on the trimmed ICP algorithm described in S4 realizes the precise registration of multi-view point clouds with adaptive overlap rate, specifically including: S41. In each round of iteration, starting from the first perspective, the overlap rate between the first perspective and the overall point cloud of all other perspectives is calculated, and the rigid body transformation matrix of the point cloud data of each perspective to the overall point cloud data of all other perspectives is calculated using the trimmed ICP algorithm; S42. After the calculation is completed, the coordinate transformation of the point cloud data of the view is performed, and then the rigid body transformation matrix from the next view to all other view point cloud data is calculated, and the coordinate transformation is performed; S43. After the first round of calculation and transformation of all perspectives is completed, the second and third rounds of calculation and transformation are continued until the rigid body transformation matrices of all perspectives are close to the unit matrix, and the overall splicing optimization is completed.

Images