A 3D measurement method and system for aircraft skin patterns based on phase matching

Abstract

The present invention discloses a three-dimensional measurement method and system for aircraft skin patterns based on phase matching, belonging to the field of three-dimensional measurement. It includes calculating the camera phase map based on the grating fringe image of multi-frequency phase shift; constructing a contour extraction model based on gray gradient and phase gradient based on the camera phase map and the camera gray image obtained under ambient light to extract the sub-pixel skin pattern contour; realizing the corresponding matching of the contour points of the left and right camera images based on the constraints such as the phase value and gray gradient value corresponding to the extracted sub-pixel skin pattern contour. Compared with the traditional method of contour three-dimensional reconstruction based on passive binocular stereo vision, it can not only directly match all contour edge points within the image range to improve the measurement speed, but also reduce the influence of factors such as contour extraction deviation caused by camera viewing angle, inconsistent contour extraction of left and right cameras, and too many epipolar matching points at the upper and lower edges resulting in matching errors, and obtain a more accurate and complete three-dimensional contour point cloud.

Description

Technical Field

[0001] The present invention belongs to the field of three-dimensional measurement, and more specifically, relates to a three-dimensional measurement method and system for aircraft skin patterns based on phase matching. Background Art

[0002] In the processing of large thin-walled parts such as aircraft skins, laser scribing and other methods are often used to scribe the edges of the areas to be milled, and then the protective film is removed accordingly for etching processing to mill out weight-reducing grooves or stiffeners, which is one of the essential processing methods for aircraft part processing. Among them, the accuracy of the pattern obtained by laser scribing is an important guarantee for the accuracy of subsequent milling area etching. Traditional methods generally use manual comparison with processing templates, which is inefficient and difficult to ensure accuracy, and cannot meet the measurement efficiency and accuracy requirements of actual parts.

[0003] For this problem, common automated three-dimensional measurement methods such as coordinate measuring machines, structured light methods, and passive binocular stereo vision are all difficult to achieve good results. Among them, the coordinate measuring machine is difficult to directly locate the contour of the processed pattern and cannot perform measurements; the structured light method will reconstruct the entire surface, and it is difficult to extract the contour point cloud for comparison; the passive binocular stereo vision method can directly reconstruct the contour point cloud. The accuracy problems of traditional passive binocular contour three-dimensional measurement mainly appear in two aspects: First, affected by the perspectives of the binocular cameras and illumination, there are certain differences in multiple aspects such as gray level and shape between the corresponding contours in the left and right camera images, and it is difficult to ensure the consistent extraction of the left and right image contours; Second, when performing stereo matching at the upper and lower edges of the pattern contour, there are multiple similar matching point pairs, and it is difficult to meet the precise pairing only relying on methods such as epipolar constraint and feature matching, and the reconstruction result is difficult to meet the requirements of actual applications.

[0004] Therefore, there is an urgent need for a method that can ensure the accurate extraction and matching of aircraft skin patterns, realize the stable and accurate three-dimensional reconstruction of aircraft skin patterns, improve the detection efficiency and accuracy, and provide a strong guarantee for the processing quality of large thin-walled parts such as aircraft skins. Summary of the Invention

[0005] In view of the above-mentioned defects or improvement requirements of the prior art, the present invention provides a three-dimensional measurement method and system for aircraft skin patterns based on phase matching, thereby solving the technical problem that the inaccurate reconstruction point cloud caused by the low accuracy of contour extraction and matching in the existing passive binocular stereo vision method.

[0006] To achieve the above object, on the one hand, the present invention provides a three-dimensional measurement method for aircraft skin patterns based on phase matching, including:

[0007] S1. Project a set of multi - frequency grating fringe images onto the surface of parts such as the skin of the aircraft to be measured. Synchronously acquire the fringe projection images formed on the surface of the workpiece to be measured through the left and right cameras, perform phase - unwrapping to obtain the phase map; and directly acquire the grayscale image of the skin of the aircraft to be measured through the left and right cameras under the ambient light without projecting the multi - frequency grating fringe images.

[0008] S2. Perform distortion correction and epipolar correction on the grayscale image and the phase map using the pre - calibrated camera parameters. Construct a composite gradient map based on the grayscale image and the phase to achieve adaptive preliminary contour extraction, obtain the preliminary contour, and then perform morphological processing and remove the incorrect contours to obtain the local area where the contour is located.

[0009] S3. Based on the preliminary contour region positioning, combine the grayscale image and the phase map to construct a sub - pixel - level precision contour extraction model based on the grayscale gradient and the phase gradient, achieve high - precision contour extraction of the left and right camera images, and obtain the contour map.

[0010] S4. Based on the left - camera contour map and the corresponding phase, find the corresponding contour on the right camera, and find the matching point pairs with the originally extracted right - camera contour according to the range constraint; construct a quality evaluation function, and perform interpolation according to the quality of the matching point pairs on the left and right camera images to generate new contour reconstruction points on the right - camera image.

[0011] S5. According to the new contour reconstruction points on the right - camera image, perform phase matching to reconstruct the three - dimensional contour point cloud.

[0012] Preferably, constructing a composite gradient map based on the grayscale image and the phase to achieve adaptive preliminary contour extraction, obtaining the preliminary contour, specifically:

[0013] S21. Construct horizontal and vertical Gaussian gradient operators, calculate the horizontal - direction and vertical - direction gradient maps based on the grayscale image, and construct the grayscale gradient map.

[0014] S22. Segment the grayscale gradient map based on the histogram, and calculate the upper and lower thresholds of the canny operator according to the grayscale gradient of the segmented regions.

[0015] S23. Perform contour extraction according to the upper and lower thresholds of the canny operator to obtain the initial edge region.

[0016] S24. Construct horizontal and vertical Gaussian gradient operators, calculate the horizontal - direction and vertical - direction gradient maps based on the phase image, and construct the phase gradient map.

[0017] S25. Segment the phase gradient map based on the histogram, obtain the regions with large phase - gradient changes, intersect with the initial edge region, and obtain the preliminary contour.

[0018] Preferably, the sub - pixel - level precision contour extraction model based on the grayscale gradient and the phase gradient is specifically:

[0019] S31. Set the phase gradient and grayscale value as constraints based on the local area where the initial contour is located, perform Steger centerline extraction, and obtain the initial centerline of the contour.

[0020] S32. Based on the initial centerline and the original grayscale image, calculate the normal vector of each point on the contour centerline, perform extension calculation of the local area width, set the size k and σ value during Gaussian filtering of the Steger algorithm, extract the centerline, and obtain the grayscale centerline of the contour.

[0021] The size k and σ value during Gaussian filtering are:

[0022]

[0023] where CEILING means rounding up to the nearest integer.

[0024] S33. Use phase information instead of grayscale information to calculate the center point coordinates. Then, combine the center point calculated in S32 to calculate the accurate sub-pixel coordinates, expressed as:

[0025]

[0026] where P represents the calculated center point coordinates, P g , P p respectively represent the sub-pixel point coordinates calculated using the grayscale centerline coordinates and phase information of the contour, w g , w p respectively represent the corresponding weight values.

[0027] Preferably, based on the left camera contour map and the corresponding phase, find the corresponding contour on the right camera, and find the matching point pairs with the originally extracted right camera contour according to the range constraint. Specifically:

[0028] S41. Find the matching pixel coordinates on the corresponding row (for convenient calculation at the pixel level) of the right camera phase map based on the phase corresponding to the left camera contour points;

[0029] S42. Perform sub-pixel interpolation according to the matching pixel coordinates and the phases of adjacent pixels to calculate the pixel coordinates of the actual matching phase;

[0030] S43. Search for the nearest neighbor matching points within the set neighborhood range of the corresponding points on the right camera contour map according to the pixel coordinates of the actual matching phase.

[0031] Preferably, construct a quality evaluation function. Specifically: Construct a quality evaluation function based on the second-order grayscale gradient and the second-order phase gradient, expressed as:

[0032] Q = w g *gradg +w p *grad p +w w *s

[0033] Among them, Q represents quality, and grad g , grad p , s respectively represent the second-order gray gradient, the second-order phase gradient, and the contour smoothness. w g , w p , w w respectively represent the corresponding weight values. The larger Q is, the better the quality is.

[0034] Preferably, interpolation is performed according to the quality of the matching point pairs on the left and right camera images to generate new contour reconstruction points on the right camera image, specifically:

[0035]

[0036] Among them, P new represents the newly generated contour reconstruction point obtained by interpolation, P1 and P2 represent the matching point pairs, and Q1 and Q2 represent the corresponding qualities of the points.

[0037] On the other hand, the present invention provides a three-dimensional measurement system for aircraft skin patterns based on phase matching, including: a computer-readable storage medium and a processor;

[0038] The computer-readable storage medium is used to store executable instructions;

[0039] The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the above-mentioned three-dimensional measurement method for aircraft skin patterns based on phase matching.

[0040] Generally speaking, compared with the prior art through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0041] 1. Affected by factors such as camera angle, lighting, and contour extraction parameters, there is often a certain degree of mismatch between the contours directly extracted from the left and right camera images, resulting in the inability to find or misfinding the corresponding points during stereo matching. The present invention reconstructs the real contour by combining the contours extracted from the binocular camera images, which can limitedly reduce the problem of contour extraction errors caused by factors such as camera perspective and lighting. At the same time, based on phase matching, the stereo matching errors caused by inconsistent contour extraction of the left and right camera images can be reduced, ensuring the correctness of the reconstructed point cloud.

[0042] 2. When using traditional passive binocular vision for reconstruction, only information such as gray gradients is relied on for contour extraction. It is greatly affected by stains, oil stains, etc. on the surface of the object to be measured, and it is difficult to ensure the integrity and accuracy of contour extraction. When the present invention performs contour extraction, it refers to both gray gradients and phase gradients, which can further ensure the integrity of contour extraction and at the same time can further improve the accuracy of contour extraction.

[0043] 3. When using passive binocular vision to extract contours for stereo matching, when there are multiple pixel points to be matched in each row at the upper and lower contour edges, the matching algorithm is difficult to ensure the correctness of the matching, which will lead to a large missing of the reconstructed point cloud at the upper and lower contours. This method directly uses the phase for contour matching, which can not only ensure the correctness of the matching points, but also further improve the accuracy of the matching, reduce the contour matching errors caused by the different perspectives of the binocular cameras, and improve the accuracy and integrity of the reconstructed point cloud. Brief Description of the Drawings

[0044] Figure 1 It is a schematic flow chart of the three-dimensional measurement method for aircraft skin patterns based on phase matching provided by the present invention;

[0045] Figure 2 It is an experimental diagram of contour extraction of the left and right cameras provided by the present invention. (a) From left to right are the images of the left camera after calibration and the contour extraction results, and (b) are the images of the right camera after calibration and the contour extraction results. It can be seen that the stable extraction of the contour center line of the aircraft skin pattern can be achieved;

[0046] Figure 3 It is a schematic diagram of inconsistent contour extraction of the left and right cameras provided by the present invention. (a) is the contour map of the right camera image and the phase matching points of the left camera contour on the right camera image, and (b) is the local position magnification of the rectangular frame in (a);

[0047] Figure 4 It is the three-dimensional measurement result of the aircraft skin based on structured light;

[0048] Figure 5 It is the three-dimensional measurement result of the aircraft skin contour based on passive binocular vision. (a) is the reconstructed contour point cloud, and (b) is the local point cloud of the rectangular frame in (a) magnified at different angles;

[0049] Figure 6 It is the result of three-dimensional measurement by directly using the right camera to extract the contour for phase matching. (a) is the reconstructed contour point cloud, and (b) is the local point cloud of the rectangular frame in (a) magnified at different angles;

[0050] Figure 7Results of three-dimensional measurement by directly using the left and right cameras to extract contours for phase matching respectively. (a) shows the reconstructed contour point clouds by two methods. The gray one is the contour reconstructed point cloud of the left camera, and the black one is the contour reconstructed point cloud of the right camera; (b) shows the enlarged views of the local point clouds in the rectangular frame in (a) at different angles.

[0051] Figure 8 Results of three-dimensional measurement of the aircraft skin pattern based on phase matching provided by the present invention. (a) shows the reconstructed contour point cloud; (b) shows the enlarged views of the local point clouds in the rectangular frame in (a) at different angles. Detailed implementation manners

[0052] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0053] An embodiment of the present invention provides a three-dimensional measurement method for an aircraft skin pattern based on phase matching, as Figure 1 shown, including:

[0054] S1. Project a set of multi-frequency grating stripe images onto the surface of parts such as the aircraft skin to be measured. Synchronously acquire the stripe projection images formed on the surface of the workpiece to be measured through the left and right cameras, perform phase demodulation, and obtain the phase map; the projector does not project the pattern, and directly acquire the gray-scale image of the workpiece to be measured through the camera under ambient light.

[0055] S2. Perform distortion correction and epipolar correction on the gray-scale images and phase maps of the left and right cameras through pre-calibrated camera parameters; construct a gradient map based on the gray-scale image and the phase map to achieve adaptive preliminary contour extraction, obtain the preliminary contour, and then perform morphological processing and remove the incorrect contours to obtain the local area where the contour is located.

[0056] Specifically, constructing a gradient map based on the gray-scale image and the phase map to achieve adaptive preliminary contour extraction and obtain the preliminary contour, the specific process is as follows:

[0057] S21. Construct horizontal and vertical Gaussian gradient operators, calculate the horizontal and vertical gradient maps based on the gray-scale image, and construct the gray-scale gradient map;

[0058] S22. Segment the gray-scale gradient map based on the histogram, and calculate the upper and lower thresholds of the canny operator according to the gray-scale gradient of the segmented region;

[0059] S23. Perform contour extraction according to the upper and lower thresholds of the canny operator to obtain the initial edge region;

[0060] S24. Construct horizontal and vertical Gaussian gradient operators, calculate the gradient maps in the horizontal and vertical directions based on the phase image, and construct a phase gradient map;

[0061] S25. Segment the phase gradient map based on the histogram, obtain the regions with large phase gradient changes, intersect with the initial edge region, and obtain a preliminary contour.

[0062] Among them, the reason for combining the gray gradient and the phase gradient is mainly to reduce the interference caused by stains and the like to the traditional gray-based contour extraction method, and lay a foundation for removing the wrong contours in the next step.

[0063] Among them, the method of removing wrong contours based on morphological processing is to extract connected components from the preliminary contour image, and then remove the noise contours according to parameters such as the size of the connected components.

[0064] S3. Based on the initial contour region positioning, combine the workpiece gray image and the phase map, construct a sub-pixel precision contour extraction model based on the gray gradient and the phase gradient, realize the high-precision contour extraction of the left and right camera images, and obtain a contour map, as Figure 2 shown.

[0065] S31. Specifically, the sub-pixel precision contour extraction model based on Steger and the phase gradient is: based on the local region where the initial contour is located, set the phase gradient and the gray value as constraints, perform Steger centerline extraction, and obtain the initial centerline of the contour.

[0066] Among them, the constraint conditions are specifically that the phase gradient of the center point needs to be greater than the set threshold, and the gray value needs to be greater than the set threshold, so as to remove the interference of darker scratches, oil stains and other influencing factors and improve the stability of the centerline extraction.

[0067] Among them, the basic process of Steger centerline extraction is to first perform Gaussian filtering, then calculate the Hessian matrix, calculate the eigen values to obtain the normal direction, and then calculate the center point.

[0068] Among them, the sub-pixel coordinates obtained by the Steger algorithm are:

[0069] (P sx ,P sy )=(P x ,P y )+(tn x ,tn y )

[0070] Among them, (P sx ,P sy ) represents the sub-pixel coordinates, (P x ,P y ) represents the pixel coordinates, (nx , n y ) is the gradient direction, and at the same time

[0071]

[0072] r x , r y , r xx , r yy , r xy is the second-order partial derivative of the image. In addition, if the pixel point taken is to be a point on the center line, it must satisfy

[0073] S32. Based on the initial center line and the original grayscale image, calculate the normal vector of each point on the contour center line, perform extension calculation on the local area width, set the size k and σ values for Gaussian filtering in the Steger algorithm, extract the center line, and obtain the grayscale center line of the contour.

[0074] where the size k and σ values for Gaussian filtering

[0075]

[0076] where CEILING means rounding up to the nearest integer.

[0077] S33. Use phase information instead of grayscale information to calculate the center point coordinates. Then, combine the center points calculated in S32 to calculate the accurate sub-pixel coordinates, which are expressed as:

[0078]

[0079] where P represents the calculated center point coordinates, P g , P p respectively represent the sub-pixel point coordinates calculated using grayscale information and phase information, w g , w p respectively represent the corresponding weight values, which are set to 0.7 and 0.3 respectively.

[0080] S4. Based on the left camera contour map and the corresponding phase, find the corresponding contour on the right camera, and find the matching point pairs with the originally extracted right camera contour according to the range constraint; construct a quality evaluation function, and perform interpolation based on the quality of the matching point pairs on the left and right camera images to generate new contour reconstruction points on the right camera image.

[0081] Specifically, based on the left camera contour map and the corresponding phase, find the corresponding contour on the right camera, and find the matching point pairs with the originally extracted right camera contour according to the range constraint, specifically:

[0082] S41. Search for the matching pixel coordinates on the corresponding row (for convenient pixel-level calculation) of the phase diagram of the right camera based on the phase corresponding to the contour points of the left camera. As Figure 3 shown, it can be seen that there are certain differences between the matching points of the left camera contour points on the right camera image and the contour points on the right camera. In the figure, a value of 0 represents a non-contour area, a value of 1 represents only the contour extraction of the right camera image (the contour extracted from the left camera image does not match it), a value of 2 represents only the pixel points on the right camera based on the phase matching of the contour extracted from the left camera image (the contour extracted from the right camera image does not match it), and a value of 3 represents that the contour extractions from both the left and right camera images are consistent, that is, they can be mutually matched based on the phase.

[0083] S42. Perform sub-pixel interpolation based on the phase of the most matching pixel coordinates and adjacent pixels to calculate the pixel coordinates of the actual matching phase.

[0084] S43. Search for the nearest neighbor matching points within the set neighborhood range of the corresponding points on the right camera contour map according to the matching phase coordinates.

[0085] Specifically, construct a quality evaluation function, specifically: construct a quality evaluation function based on the second-order gray gradient and the second-order phase gradient, expressed as:

[0086] Q = w g *grad g +w p *grad p +w w *s

[0087] Among them, Q represents the quality, grad g , grad p , s respectively represent the second-order gray gradient, the second-order phase gradient, and the contour smoothness, and w g , w p , w w respectively represent the corresponding weight values. The larger Q is, the better the quality.

[0088] Specifically, perform interpolation based on the quality of the matching point pairs on the left and right camera images to generate new contour reconstruction points on the right camera image, specifically:

[0089]

[0090] Among them, P new represents the newly generated contour reconstruction point obtained by interpolation, P1, P2 represent the matching point pairs, and Q1, Q2 represent the corresponding qualities of the points.

[0091] S5. Based on the new contour reconstruction points on the right camera image, perform phase matching to reconstruct the three-dimensional contour point cloud.

[0092] Specifically, based on the absolute phase corresponding to the right camera contour, corresponding points on the absolute phase map of the left camera are found to achieve phase point matching, and three-dimensional point clouds are reconstructed according to the principle of triangulation. Figures 4 - 8 They are, in sequence, structured light reconstructed point clouds, passive binocular reconstructed point clouds, right camera image contour reconstructed point clouds based on phase matching, comparison of left and right camera image contour reconstructed point clouds based on phase matching, and point clouds reconstructed by this method. Figure 4 This is the three-dimensional measurement result of the aircraft skin based on structured light provided by the present invention. It can be seen that all point clouds are reconstructed, and it is difficult to directly obtain the accurate value of the aircraft skin contour. Figure 5 This is the three-dimensional measurement result of the aircraft skin contour based on passive binocular vision. (a) is the reconstructed contour point cloud, and (b) is the enlarged view of the local point cloud in the rectangular frame in (a) at different angles. It can be seen that although the result of the reconstructed point cloud is good in the front-facing direction, affected by the inconsistent extraction of the left and right camera images, the point cloud is uneven in the depth direction and the calculation error is large. Figure 6 This is the result of three-dimensional measurement by directly using the right camera to extract the contour for phase matching. (a) is the reconstructed contour point cloud, and (b) is the enlarged view of the local point cloud in the rectangular frame in (a) at different angles. It can be seen that the reconstructed point cloud not only has a good result in the front-facing direction but also seems to have a good result in the depth direction, and the point cloud is relatively smooth. However, it should be noted that as Figure 6 shown by the point cloud at the elliptical frame in (b) of , although the point cloud seems smooth at some positions, the contour is uneven and does not conform to the actual result. The main reason for this situation is that although phase matching can ensure the consistency of left and right camera contour matching, affected by factors such as camera perspective and lighting, the contour extracted from a certain camera image may have objects, and direct matching will produce incorrect results. Figure 7 This is the result of three-dimensional measurement by directly using the left and right cameras to extract the contour for phase matching respectively. (a) is the reconstructed contour point cloud by the two methods. The gray one is the reconstructed point cloud of the left camera contour, and the black one is the reconstructed point cloud of the right camera contour; (b) is the enlarged view of the local point cloud in the rectangular frame in (a) at different angles. It can be seen that the point clouds reconstructed by the two methods look very smooth, but there are obvious deviations in the results, which corroborates the inconsistent contour extraction in . Therefore, there may be large errors no matter which camera contour is directly used for phase matching. Figure 3 Figure 8 This is the three-dimensional measurement result of the aircraft skin pattern based on phase matching provided by the present invention. (a) is the reconstructed contour point cloud; (b) is the enlarged view of the local point cloud in the rectangular frame in (a) at different angles. It can be seen that the method proposed by the present invention can reconstruct more accurate three-dimensional point cloud data on the basis of extracting contours from the left and right cameras to reduce the problem of local contour extraction errors in a certain camera image caused by factors such as camera perspective and lighting. It can be seen that compared with Figure 4 ​For the structured light reconstructed point cloud, this method can directly obtain the contour; compared with Figure 5 the passive binocular reconstructed point cloud, the point cloud reconstructed by this method has better quality; compared with Figure 6 , Figure 7 the single-camera reconstructed contour point cloud, the reconstruction by this method has better stability.

[0093] Among them, let the right camera contour map be E R (the contour is set to 1, and the area outside the contour is set to 0), and the left camera phase map be P R , then the phase map corresponding to the camera contour is E R * P R , and based on this, directly find the matching points on the right camera phase map P L .

[0094] It is easy for those skilled in the art to understand that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A three-dimensional measurement method for aircraft skin patterns based on phase matching, characterized in that, Including: S1. Project a set of multi-frequency grating fringe images onto the surface of the aircraft skin to be measured. Synchronously obtain the fringe projection images formed on the surface of the workpiece to be measured through the left and right cameras, perform phase demodulation to obtain the phase map; and directly obtain the grayscale image of the aircraft skin to be measured through the left and right cameras under the ambient light without projecting the multi-frequency grating fringe images. S2. Perform distortion correction and epipolar correction on the grayscale image and the phase map through the pre-calibrated camera parameters. Then, construct a composite gradient map based on grayscale and phase to achieve adaptive preliminary contour extraction, obtain the preliminary contour, and then perform morphological processing and remove the incorrect contours to obtain the local area where the contour is located. S3. Based on the preliminary contour region localization, combine the grayscale image and the phase map to construct a sub-pixel accuracy contour extraction model based on grayscale gradient and phase gradient to achieve high-precision contour extraction of the left and right camera images and obtain the contour map. Specifically including: S31. Based on the local area where the initial contour is located, set the phase gradient and grayscale value as constraints, perform Steger centerline extraction to obtain the initial centerline of the contour. S32. Based on the initial centerline and the original grayscale image, calculate the normal vector of each point on the contour centerline, perform extension calculation of the local area width, and perform Steger centerline extraction to obtain the grayscale centerline of the contour. S33. Comprehensively consider the phase information and the grayscale information to calculate the accurate sub-pixel coordinates of the skin contour, expressed as: Among them, P represents the calculated central point coordinates, P g , P p respectively represent the sub-pixel point coordinates calculated using the gray center line coordinates and phase information of the contour, w g , w p respectively represent the corresponding weight values; S4. Based on the left camera contour map and the corresponding phase, find the corresponding contour on the right camera, and find the matching point pairs with the originally extracted right camera contour according to the range constraint; construct a quality evaluation function, and perform interpolation according to the quality of the matching point pairs on the left and right camera images to generate new contour reconstruction points on the right camera image. S5. According to the new contour reconstruction points on the right camera image, perform phase matching to reconstruct the three-dimensional contour point cloud.

2. The method according to claim 1, characterized in that, Construct a composite gradient map based on grayscale and phase to achieve adaptive preliminary contour extraction, specifically including: S21. Construct horizontal and vertical Gaussian gradient operators, calculate the horizontal and vertical direction gradient maps based on the grayscale image, and construct the grayscale gradient map. S22. Segment the grayscale gradient map based on the histogram, and calculate the upper and lower thresholds of the Canny operator according to the grayscale gradient of the segmented region. S23. Perform contour extraction according to the upper and lower thresholds of the Canny operator to obtain the initial edge region. S24. Construct horizontal and vertical Gaussian gradient operators, calculate the horizontal and vertical direction gradient maps based on the phase image, and construct the phase gradient map. S25. Segment the phase gradient map based on the histogram, obtain the region with large phase gradient changes, and take the intersection with the initial edge region to obtain the preliminary contour.

3. The method according to claim 1, characterized in that, The method of finding the corresponding contour on the right camera based on the left camera contour map and the corresponding phase, and finding the matching point pairs with the originally extracted right camera contour according to the range constraint specifically includes: S41. Find the matching pixel coordinates on the corresponding row of the right camera phase map based on the phase corresponding to the left camera contour points. S42. Perform sub-pixel interpolation according to the matching pixel coordinates and the phases of the adjacent pixels to calculate the pixel coordinates of the actually matched phase. S43. Search for the nearest neighbor matching point within the set neighborhood range of the corresponding point on the right camera contour map according to the pixel coordinates of the actually matched phase.

4. The method according to claim 3, characterized in that Comprehensively consider three constraints: second-order gray gradient, second-order phase gradient, and contour smoothness, which is expressed as: Q = w g * grad g + w p * grad p + w w * s Among them, Q represents quality, grad g , grad p , s respectively represent the second-order gray gradient, the second-order phase gradient, and the contour smoothness, w g , w p , w w respectively represent the corresponding weight values.

5. The method according to claim 4, wherein Interpolate comprehensively according to the contours extracted from the left and right camera images to improve the stability of contour extraction, which is expressed as: Among them, P new represents the newly interpolated contour reconstruction point, P1 and P2 represent the matching point pairs, and Q1 and Q2 represent the corresponding qualities of the points.

6. A three-dimensional measurement system for aircraft skin patterns based on phase matching, characterized in that It includes: A computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium and execute the three-dimensional measurement method for the aircraft skin pattern based on phase matching according to any one of claims 1 to 5.

Images