A wall thickness automatic detection method and system for complex curved surface parts

Abstract

This invention belongs to the field of non-destructive testing of heat exchange equipment, specifically relating to an automatic wall thickness detection method and system for complex curved surface parts, including: acquiring complete three-dimensional point cloud data of the overall shape of the complex curved surface part to be tested, reconstructing the geometric mesh model of the curved surface part to be tested; estimating the normal vectors of all upper and lower geometric mesh patches in the geometric mesh model; selecting each measurement point p on the curved surface part to be tested. U_i And determine the geometric mesh patch N on its surface. U_j The normal vector is n U_j By searching for the surface N of the tested curved part in the thickness direction opposite to the surface where it is located. U_j Matching corresponding geometric mesh surface N D_j ; Calculate the measurement point p U_i To noodle N D_j The distance, as the measurement point p U_i The wall thickness at the location; where, prior to the matching search, the geometric mesh patch N... U_j With geometric mesh patch N D_j The normal vector direction has been consistently adjusted. The method of this invention can effectively achieve high-precision and high-efficiency extraction of the complete wall thickness of complex curved surface parts.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of nondestructive testing of heat exchange equipment, and more particularly relates to a wall thickness automatic detection method and system for complex curved surface parts. BACKGROUND

[0002] With the development of high-end equipment such as aero-engines, heat exchangers, and steam turbines towards high thrust-to-weight ratio and high thermal conductivity, the service environment of key parts thereof tends to be more and more severe and extreme, thereby greatly increasing the failure rate of the parts. Such parts usually have complex topological structure characteristics, and the uniformity of the wall thickness thereof is difficult to control. Complete precision analysis on the overall surface wall thickness of the parts can provide basic measurement data for forming process optimization, and plays a crucial role in improving the uniformity of the wall thickness of complex curved surface parts and the service safety thereof.

[0003] At present, the methods commonly used in the industrial field for detecting the uniformity of the wall thickness of complex curved surface parts mainly include traditional contact type clamp detection, three-coordinate detection, eddy current detection, ultrasonic detection, and ray detection. The first two methods are limited by the large size and complex surface of the parts, and can only detect simple features in local areas, and cannot complete the complete extraction of the overall surface data of the parts. The last three methods are limited by the complexity of the detection equipment and the high requirements on the environment, and it is usually difficult to achieve high-precision and high-efficiency extraction of the geometric morphology data of the parts. SUMMARY

[0004] In view of the defects and improvement needs of the prior art, the present application provides a wall thickness automatic detection method and system for complex curved surface parts, which aims to realize high-precision and high-efficiency complete extraction of the wall thickness of the surface of the parts.

[0005] To achieve the above-mentioned purpose, according to one aspect of the present application, a wall thickness automatic detection method for complex curved surface parts is provided, comprising:

[0006] obtaining complete three-dimensional point cloud data of the overall surface of the measured complex curved surface part; and reconstructing a geometric grid model of the measured curved surface part according to the complete three-dimensional point cloud data of the overall surface;

[0007] estimating the normal vector of all geometric grid facets in the geometric grid model; selecting each measurement point p U_i on the measured curved surface part, and determining the normal vector n U_j of the geometric grid facet N U_j on which the measurement point p U_i is located;

[0008] searching for a corresponding geometric grid facet N D_j matching the facet N U_j on the surface of the measured curved surface part opposite to the surface on which the measurement point p U_i is located in the thickness direction, and the matching search criterion is the facet NU_j The angle between the normal vector of the surface patch N D_j and the normal vector of the surface patch N U_i is less than a given threshold; the distance from the measurement point p D_j to the surface patch N U_i is calculated as the wall thickness at the measurement point p U_j ; wherein, before the matching search, the normal vector direction of the geometric grid surface patch N D_j and the normal vector direction of the geometric grid surface patch N

[0009] The beneficial effects of the present application are: the present application firstly constructs a geometric grid model of the measured curved surface part according to the complete three-dimensional point cloud data of the overall surface, and based on the principle that the normal vectors of the upper and lower surfaces in the thickness direction are approximately parallel, the relative grid plane equation of the measurement point on each surface in the thickness direction is obtained through matching search, and the wall thickness of the measurement point is calculated according to the coordinates of the measurement point and the corresponding grid plane equation, thereby effectively realizing high-precision and high-efficiency complete extraction of the wall thickness of the part surface, especially for complex curved surface parts, the present application method can also effectively realize high-precision and high-efficiency complete extraction of the wall thickness, and has high application value.

[0010] Further, the complete three-dimensional point cloud data of the overall surface of the measured complex curved surface part is obtained by using a surface structured light binocular integrated three-dimensional measurement method.

[0011] The further beneficial effects of the present application are: the present application uses a surface structured light binocular integrated three-dimensional measurement technology, generates a coded structured light pattern through a structured light projection module, and projects the coded structured light pattern on the surface of the measured part to obtain three-dimensional profile information of the part surface. Such detection means improves the technical difficulties of over-reliance on manual work in the field of industrial quality inspection from the measurement principle, and has good performance in analyzing machining errors, key dimensions and forming quality of large-size parts.

[0012] Further, the calibration method of the binocular camera is:

[0013] The same local area of the standard part is scanned to obtain image information of the local area under different cameras; the image information of the local area under different cameras is used to determine the camera internal parameters and external parameters of each camera based on the local three-dimensional feature rigid transformation consistency criterion, as the calibration parameters of the binocular camera.

[0014] The further beneficial effects of the present application are: when calibrating the binocular camera, the local three-dimensional feature rigid transformation consistency criterion is introduced to improve the solving efficiency of the calibration parameters.

[0015] Further, it also includes: evaluating the wall thickness of each measurement point, specifically:

[0016]

[0017] wherein t s represents the reference wall thickness value, t U_i is the wall thickness at the measurement point p U_i .

[0018] Further beneficial effects of the present application are that after calculating the wall thickness, the present application further proposes a wall thickness evaluation method to calculate the deviation of the wall thickness of the manufactured complex curved surface part from the reference value set before manufacturing, so as to evaluate the manufacturing process.

[0019] Further, before reconstructing the geometric grid model of the measured curved surface part, the method further comprises: down-sampling the dense sampling point cloud, specifically:

[0020] The x, y and z axes of the coordinate system in which the complete three-dimensional point cloud data is located are respectively taken as the length, width and height directions of the bounding box, wherein the complete three-dimensional point cloud data represents In the data , the maximum values x max , y max , z max and the minimum values x min , y min , z min along the x, y and z directions are searched respectively, so as to calculate the length, width and height of the bounding box;

[0021] The point cloud resolution is calculated according to the complete three-dimensional point cloud data , so as to determine the voxel grid size of the bounding box; according to the voxel grid size, the number of voxels of the bounding box on the x, y and z axes is determined; according to the number of voxels, the coordinates of each point, the voxel grid size and the minimum values x min , y min , z min , the index of all points in each voxel is calculated, so as to calculate the center of gravity of the voxel, and the center of gravity is taken as the new measurement point after down-sampling of the voxel, which is used for wall thickness calculation.

[0022] Further beneficial effects of the present application are that by down-sampling, the number of measurement points is reduced, and the detection efficiency is improved.

[0023] Further, before the matching search, after the normal vector estimation of all geometric grid patches in the geometric grid model is performed, the consistency adjustment of the normal vector directions between the geometric grid patches is performed.

[0024] Further beneficial effects of the present application are that in fact, the normal vector direction consistency adjustment can also be performed on each measurement point p U_iAfterwards, the normal vector direction between the neighborhood geometric grid patches is adjusted for consistency, but it is relatively complicated in actual implementation, and the method is convenient and fast.

[0025] Further, the normal vector direction between the geometric grid patches is adjusted for consistency in the following manner: the normal vector of each geometric grid patch is calculated in the following manner, thereby completing the adjustment of the normal vector direction between the geometric grid patches for consistency:

[0026] The center point of each geometric grid patch j is calculated In the formula, j1, j2, j K The measurement point number in the geometric grid patch j is represented;

[0027] The normal vector n of the geometric grid patch j i It is calculated by the following formula: p i Any measurement point on the geometric grid patch j is represented, and v j The normal vector of the geometric grid patch j before adjustment is represented.

[0028] The further beneficial effect of the present application is to simplify the algorithm complexity.

[0029] The present application also provides a computer readable storage medium, which comprises a stored computer program, wherein the computer program controls the device where the storage medium is located to execute the wall thickness automatic detection method for complex curved surface parts as described above when the computer program is run by a processor.

[0030] Overall, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0031] The present application proposes that a geometric grid model of the measured curved surface part is first constructed according to the overall complete three-dimensional point cloud data, and based on the principle that the normal vectors of the upper and lower surfaces in the thickness direction are approximately parallel, the relative grid plane equation of the measurement point on each surface in the thickness direction is obtained by matching search according to the coordinate of each measurement point and the corresponding grid plane equation, and the wall thickness of the measurement point is calculated according to the coordinate of each measurement point and the corresponding grid plane equation, thereby effectively realizing high-precision and high-efficiency complete extraction of the part surface wall thickness, especially for complex curved surface parts, the present application method can also effectively realize high-precision and high-efficiency complete extraction of the wall thickness, and has high application value. Further, the preferred schemes of the face structure light binocular integrated three-dimensional measurement method, point cloud downsampling, and normal vector consistency adjustment are proposed, thereby realizing more efficient nondestructive detection. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 A wall thickness automatic detection method for complex curved surface parts is provided for the embodiment of the present application.

[0033] Figure 2 A schematic diagram of a matching triangular mesh search provided in an embodiment of the present invention;

[0034] Figure 3 This is a simplified structural diagram of an automatic wall thickness detection device for complex curved surface parts, provided as an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0036] Example 1

[0037] An automatic wall thickness detection method for complex curved surface parts, such as... Figure 1 As shown, it includes:

[0038] Acquire complete 3D point cloud data of the overall shape of the complex curved surface part under test; reconstruct the geometric mesh model of the curved surface part under test based on the complete 3D point cloud data of the overall shape.

[0039] Normal vector estimation is performed on all geometric mesh patches in the upper and lower geometric mesh model; measurement points p on the measured curved surface part are selected. U_i And determine the geometric mesh patch N on its surface. U_j The normal vector is n U_j ;

[0040] By searching for the surface N of the tested curved surface part in the thickness direction opposite to the surface where it is located. U_j Matching corresponding geometric mesh surface N D_j The matching search criterion is face N. U_j With sheet N D_j The angle between the normal vectors is less than a given threshold; calculate the measurement point p. U_i To noodle N D_j The distance, as the measurement point p U_i The wall thickness at the location; where, prior to the matching search, the geometric mesh patch N... U_j With geometric mesh patch N D_j The direction of the normal vector has been adjusted for consistency.

[0041] It should be noted that the above-mentioned threshold values ​​are set according to accuracy requirements.

[0042] To illustrate the method more specifically, take a triangular mesh model as an example for illustration, as shown in Figure 2 P u_i represents an arbitrary measurement point on the upper surface, and the vector n is its corresponding normal vector. Then, according to the nearest neighbor search matching algorithm, by giving a search neighborhood radius r, the lower surface P L and the point P u_i set of nearest neighbor points are extracted. Further, in the XOY plane, the lower surface and the point P u_i with the same index P l_i is taken as the center, the matching point set is divided into 6 equal parts by counterclockwise rotation along the OY axis, and then the plane reconstruction algorithm is used to fit and reconstruct each triangular mesh plane, so as to obtain the corresponding normal vector n t (t = 1, 2…6). Finally, by calculating the angle between the normal vector n t and n, and comparing it with the angle reference threshold, the corresponding triangular mesh N u_i matching the point P D_j is searched, and thus the distance between the point P u_i and the matching mesh N D_j is calculated to obtain the measurement thickness value at the point P u_i .

[0043] The embodiment proposes that first, according to the complete three-dimensional point cloud data of the overall surface of the measured curved surface part, a geometric mesh model of the measured curved surface part is constructed, according to the geometric mesh model, based on the principle that the normal vectors of the upper and lower surfaces in the thickness direction are approximately parallel, through matching search, the grid plane equation of the measurement point on each surface in the thickness direction is obtained, according to the coordinates of each measurement point and the corresponding grid plane equation, the wall thickness of the measurement point is calculated, which effectively realizes high-precision and high-efficiency complete extraction of the wall thickness of the part surface, especially for complex curved surface parts, the method of the present application can also effectively realize high-precision and high-efficiency complete extraction of the wall thickness, which has high application value.

[0044] As a preferred embodiment, a face structure light binocular integrated three-dimensional measurement method is adopted to obtain complete three-dimensional point cloud data of the overall surface of the measured complex curved surface part, which mainly includes the following steps: using a calibrated binocular camera to capture the projection pattern of the face structure light projector on the surface of the complex curved surface part; determining the pixel point matching relationship between the projection patterns captured by the left and right cameras according to the calibration parameters of the binocular camera; decoding the projection patterns captured by the left and right cameras according to the pixel point matching relationship, and obtaining complete three-dimensional point cloud data of the overall surface of the measured complex curved surface part by using the optical triangulation principle.

[0045] In the device configuration, as Figure 3As shown, can include: image acquisition module, logic motion control module, and image processing module; wherein the image acquisition module is arranged on the logic motion control module, for capturing the projection pattern of the surface structure light projector on the surface of the measured curved part under the motion control of the logic motion control module; the image processing module is used for determining the pixel point matching relationship between the projection patterns corresponding to the left and right cameras according to the calibration parameters of the binocular camera; according to the matching relationship, the projection patterns corresponding to the left and right cameras are decoded, and the complete three-dimensional point cloud data of the measured curved part is obtained.

[0046] The above image acquisition module is composed of a binocular camera integrated surface structure light scanner on the structural main body. In implementation, first, the speckle pattern is projected on the surface of the part by the surface structure light projector, and the image acquisition is performed by the binocular camera, so as to complete the local image acquisition under single view angle; the above logic motion control module drags the surface structure light scanner fixed on the moving platform, and the complete contour image of the part surface is obtained by continuously adjusting and transforming the spatial measurement pose of the scanner; the above image processing module obtains the complete three-dimensional appearance data of the measured part according to the internal and external parameters obtained by the binocular camera calibration, and then uses the three-dimensional reconstruction algorithm to complete the automatic evaluation of the part wall thickness uniformity based on the digital model comparison.

[0047] The preferred scheme adopts the surface structure light binocular integrated three-dimensional measurement technology, generates the coded structure light pattern through the structure light projection module, and projects the pattern on the surface of the measured part to obtain the three-dimensional contour information of the part surface. Such detection means improves the technical difficulties in the industrial quality inspection field which relies too much on manual work in terms of measurement principle, and has good performance in analyzing the machining error, key size and forming quality of large-size parts.

[0048] As a preferred embodiment, the calibration method of the above binocular camera is:

[0049] Scanning the same local area of the standard part to obtain image information of the local area under different cameras; using the image information of the local area under different cameras, determining the camera internal parameters and external parameters of each camera based on the local three-dimensional feature rigidity transformation consistency criterion, as the calibration parameters of the binocular camera.

[0050] The specific way of determining the calibration parameters of the binocular camera based on the local three-dimensional feature rigidity transformation consistency criterion is:

[0051] S1, assuming that there is a point P(X w ,Y w ,Z w ) on the local area in space, and the imaging point of P on the image plane of one of the cameras is p(u,v), then according to the camera pinhole imaging model, we have:

[0052]

[0053] where s is a size factor; u0, v0 represent the principal point position; f x , f y are the focal length in horizontal and vertical direction of the image coordinate system respectively; A is the camera intrinsic parameter matrix; R, T are the rotation matrix and translation matrix of the extrinsic parameter respectively; γ is the skew factor;

[0054] S12, assuming the calibration board is located in the X w O w Y w plane of the world coordinate system, then Z w = 0, formula (1) is transformed into:

[0055]

[0056] where r1, r2, r3 are column vectors of the rotation matrix R, and have the property of being pairwise orthogonal;

[0057] Let H = λA[r1 r2 T] = [h1 h2 h3], then:

[0058]

[0059] Let B = A -T A -1 , we have:

[0060]

[0061] B is a symmetric matrix, which can be further expressed as a six-dimensional vector:

[0062] b = [B 11 B 12 B 22 B 13 B 23 B 33 ] (5)

[0063] Let the i-th column of the matrix H be h i = [h i1 h i2 h i3 ] T , we have:

[0064]

[0065] Substitute formula (6) into formula (3), then we have:

[0066]

[0067] That is:

[0068] Vb = 0 (8)

[0069] In the formula, V is the set matrix of v;

[0070] S3, the same area of the same calibration board is photographed by the left and right cameras in the binocular camera model respectively, and n images are obtained by multi-pose and multi-angle shooting, and n pixel points are obtained, and the corresponding formula (8) is 2n relationship formulas; then the value of matrix b is obtained according to the consistency criterion of three-dimensional space rigid transformation, that is, there is a consistent corresponding constraint relationship between the left and right camera measurement data sets; further substituting it into formula (5) can obtain the corresponding internal parameters u0, v0, λ, f x , f y , γ; and external parameters R, T.

[0071] When calibrating the binocular camera, the local three-dimensional feature rigid transformation consistency criterion is introduced to improve the solving efficiency of the calibration parameters.

[0072] From camera calibration to part three-dimensional point cloud data acquisition, the process can be summarized as follows:

[0073] S1, according to the topological structure of the measured part, the feature size, and the like, the surface structure light projector and the surface structure light scanner integrated with the binocular camera are cascaded on the three-dimensional moving platform, and the three-dimensional moving platform moves under the control of the logical motion control module;

[0074] S2, by extracting the local features of the same standard part under different viewing angles by the surface structure light scanner, the three-dimensional space rigid transformation consistency criterion is used to realize the global calibration of the camera internal and external parameters;

[0075] S3, the projection pattern of the surface structure light projector on the part surface is captured by the binocular camera, and the scanner is dragged from the three-dimensional space by the motion cascade mechanism to complete the complete acquisition of the three-dimensional image of the overall appearance of the measured part;

[0076] S4, the image is decoded and analyzed by the data processing algorithm to extract the complete three-dimensional point cloud data of the part. On this basis, the three-dimensional model reconstruction of the scattered point cloud data is completed, and the part wall thickness uniformity is automatically evaluated according to the numerical model comparison result.

[0077] As a preferred embodiment, it also includes: evaluating the wall thickness of each measurement point, specifically:

[0078]

[0079] In the formula, t s represents the reference wall thickness value, t U_i is the wall thickness of the measurement point p U_i .

[0080] The method further proposes a wall thickness evaluation method after calculating the wall thickness, to calculate the deviation of the wall thickness of the manufactured complex curved surface part from the reference value set before manufacturing, so as to evaluate the manufacturing process.

[0081] As a preferred embodiment, before reconstructing the geometric grid model of the measured curved surface part, the method further comprises: down-sampling the dense sampling point cloud, specifically:

[0082] The x, y and z axes of the coordinate system in which the complete three-dimensional point cloud data is located are respectively taken as the length, width and height directions of the bounding box, wherein the complete three-dimensional point cloud data represents In the data , the maximum values x max , y max , z max and the minimum values x min , y min , z min along the x, y and z directions are searched respectively to calculate the length, width and height of the bounding box;

[0083] The point cloud resolution is calculated according to the complete three-dimensional point cloud data to determine the voxel grid size of the bounding box; according to the voxel grid size, the number of voxels of the bounding box on the x, y and z axes is determined; according to the number of voxels, the coordinates of each point, the voxel grid size and the minimum values x min , y min , z min , the index of all points in each voxel is calculated to calculate the center of gravity of the voxel, and the center of gravity is taken as the new measurement point of the voxel after down-sampling, which is used for wall thickness calculation.

[0084] The method reduces the number of measurement points through down-sampling, and improves the detection efficiency.

[0085] As a preferred embodiment, before the matching search, after the normal vector estimation of all geometric grid patches in the geometric grid model is performed, the consistency adjustment of the normal vector directions between the geometric grid patches is performed.

[0086] In fact, the consistency adjustment of the normal vector directions can also be performed after the selection of each measurement point p U_i on the measured curved surface part, but it is more cumbersome in actual implementation, and the method is convenient and fast.

[0087] As a preferred implementation, the method for consistently adjusting the normal vector direction between geometric mesh patches is to uniformly calculate the normal vector of each geometric mesh patch in the following way, thereby achieving consistent adjustment of the normal vector direction between the geometric mesh patches:

[0088] Calculate the center point of each geometric mesh patch j In the formula, j1, j2, j K Indicates the measurement point number within the geometric mesh patch j;

[0089] The normal vector n of the geometric mesh patch j i Calculated using the following formula: p i v represents any measurement point on the geometric mesh patch j. j This represents the normal vector of the geometric mesh patch j before adjustment.

[0090] This method simplifies the algorithm, reduces its complexity, and facilitates execution. It can be used to perform consistency adjustments on the normal vector directions between geometric mesh patches after estimating the normal vectors of all upper and lower geometric mesh patches in the geometric mesh model before matching and searching; it can also be used to select measurement points p on the measured curved surface part. U_i Next, the normal vector directions between neighboring geometric mesh patches are adjusted for consistency.

[0091] An example of using the method of the present invention is given below:

[0092] (1) Using the calibrated binocular camera's internal and external parameters, the depth information of each pixel is calculated by the optical triangulation principle. Based on the matching result of "pixel map + depth map", the point cloud data of the overall shape of the measured part is effectively extracted.

[0093] (2) Using the surface reconstruction algorithm, the point cloud data without topological connection relationship is reconstructed to obtain a triangular mesh model;

[0094] (3) Further solve for the normal vectors of each triangular mesh on the upper and lower surfaces to complete the adjustment of the normal vector direction consistency. At this time, the normal vector at the measurement point is approximately equal to the normal vector of the triangular mesh in which it is located;

[0095] (4) A measurement point p on the above surface U_i (x i ,y i ,z i Taking the wall thickness extraction at point N as an example: Assume that the point is located in the triangular mesh surface N. U_i The normal vector is n U_i By searching in the lower surface for the patch N U_i The corresponding triangular mesh face N of the nearest matchingD_i (assuming its plane equation is: a i x+b i y+c i z+d i = 0), the corresponding normal vector is n D_i , the matching search criterion is based on:

[0096]

[0097] Further, the wall thickness t U_i at point p U_i can be obtained by the formula , and the corresponding wall thickness uniformity information can be judged by the formula ; in the formula, U and D represent the upper and lower surfaces; i represents the corresponding serial number; alpha represents the included angle between the normal vectors n U_i and n D_i ; t s represents the reference wall thickness value.

[0098] (5) Repeat step (4) until the complete extraction of the forming wall thickness information of the overall shape surface of the part to be measured.

[0099] In this method, the non-contact automatic detection is implemented in combination with the structured light three-dimensional measurement technology. The detection process mainly includes the following steps: a surface structured light scanner is formed by cascading a surface structured light projector and a binocular camera, complete three-dimensional data of the overall shape surface of the part is obtained by analyzing the part topography image obtained by the binocular camera; then a three-dimensional reconstruction algorithm is used to obtain a triangular mesh model of the part, the normal vector information of the point cloud is solved, and on this basis, the corresponding matching triangular mesh is sought according to the normal vector information, so that the corresponding wall thickness uniformity information is solved according to the point-to-plane distance. This embodiment effectively improves the technical short board of low detection efficiency, inability to achieve all inspection and full inspection when the traditional detection method measures some machining errors, key dimensions and forming quality information of complex curved surface parts, and further improves the production efficiency of enterprises and the operation safety of equipment. In summary, the present application realizes online non-destructive detection of the wall thickness uniformity of complex curved surface parts, is not affected by the complex topography of the measured object, has high measurement accuracy and strong reliability of the measured results.

[0100] Embodiment Two

[0101] A computer readable storage medium comprising a stored computer program, wherein the computer program, when executed by a processor, controls the device in which the storage medium is located to perform a wall thickness automatic detection method for complex curved surface parts as described above.

[0102] The related technical solutions are the same as those in Embodiment One, which will not be repeated here.

[0103] Those skilled in the art can easily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A wall thickness automatic detection method for a complex curved surface part, characterized in that, The method comprises: acquiring overall surface complete three-dimensional point cloud data of a measured complex curved surface part; reconstructing a geometric grid model of the measured curved surface part according to the overall surface complete three-dimensional point cloud data; The normal vector of all geometric grid surface patches in the geometric grid model is estimated; each measuring point on the measured curved surface part is selected , and the geometric grid surface patch N on the surface where it is located is determined U_j , and the normal vector of the geometric grid surface patch N is n U_j ; By searching for the surface patch N in the thickness direction of the tested curved surface part opposite to the surface where it is located. U_j Matching corresponding geometric mesh surface N D_j The matching search criterion is face N. U_j With sheet N D_j The angle between the normal vectors is less than a given threshold; calculate the measurement point. To noodle N D_j The distance, as the measurement point The wall thickness at the location; where, prior to the matching search, the geometric mesh patch N... U_j With geometric mesh patch N D_j The direction of the normal vector has been consistently adjusted; wherein the manner of adjusting the normal vector directions between the geometric grid patches uniformly is as follows: uniformly calculating the normal vector of each geometric grid patch to complete the adjustment of the normal vector directions between the geometric grid patches. Calculate each geometric mesh face j center point : In the formula, , , Represents geometric mesh patches j The measurement point number within; The geometric mesh surface j normal vector Calculated using the following formula: , Represents geometric mesh patches j Any measurement point on, Represents geometric mesh patches j The normal vector before adjustment.

2. The wall thickness automatic detection method according to claim 1, characterized by, The overall surface complete three-dimensional point cloud data of the measured complex curved surface part is acquired by using a surface structured light binocular integrated three-dimensional measurement method, specifically, a calibrated binocular camera is used to capture the projection pattern of a surface structured light projector on the surface of the complex curved surface part; the pixel point matching relationship between the projection patterns captured by the left and right cameras is determined according to the calibration parameters of the binocular camera; the projection patterns captured by the left and right cameras are decoded by using the optical triangulation principle according to the pixel point matching relationship, so as to acquire the overall surface complete three-dimensional point cloud data of the measured complex curved surface part.

3. The wall thickness automatic detection method according to claim 2, wherein The calibration method of the binocular camera is as follows: the same local area of a standard part is scanned to obtain image information of the local area under different cameras; the image information of the local area under different cameras is used to determine the camera internal parameters and external parameters of each camera based on the local three-dimensional feature rigid transformation consistency criterion, which are used as the calibration parameters of the binocular camera.

4. The wall thickness automatic detection method according to claim 1, characterized by, The method further comprises: evaluating the wall thickness of each measurement point, specifically: ； In the formula, denotes the reference wall thickness value, is the wall thickness at the measurement point .

5. The wall thickness automatic detection method according to claim 1, wherein Before reconstructing the geometric grid model of the measured curved surface part, the method further comprises: down-sampling the dense sampling point cloud, specifically: The coordinate system of the complete 3D point cloud data , , The axes represent the length, width, and height of the bounding box, respectively, where the complete 3D point cloud data represents... In the data Search along the middle , , Maximum value in three directions 、 、 and minimum value , , To calculate the length, width, and height of the bounding box; According to the complete three-dimensional point cloud data , the point cloud resolution is calculated to determine the voxel grid size of the bounding box; according to the voxel grid size, the number of voxels of the bounding box on 、 、 three axes is determined; according to the number of voxels, the coordinates of each point, the voxel grid size and the minimum value 、 、 , the index of all points in each voxel is calculated to calculate the center of gravity of the voxel, and the center of gravity is taken as the new measurement point of the voxel after down-sampling for wall thickness calculation.

6. The wall thickness automatic detection method according to claim 1, wherein Before the matching search, after the normal vector estimation of all the geometric grid patches in the geometric grid model, the adjustment of the normal vector directions between the geometric grid patches is performed.

7. A computer readable storage medium characterized in that, The computer readable storage medium comprises a stored computer program, wherein when the computer program is run by a processor, the device where the storage medium is located is controlled to perform the wall thickness automatic detection method for a complex curved surface part according to any one of claims 1 to 6.

Images