Multiscale vision measurement coordinate system method, system, medium, and apparatus

Abstract

A multi-scale visual measurement coordinate system, including a method, system, medium, and equipment, is disclosed. The method involves using a 3D visual measurement system to perform 3D scanning of a standard part, acquiring point cloud data of a portion of the standard part's outer surface, and using a microscopic depth-of-field measurement system to measure a local area of ​​the standard part, acquiring point cloud data. The CAD model of the standard part is discretized to generate an ideal reference point cloud with geometric structure and color annotations. A high-dimensional feature vector is constructed based on the geometric and color features. An improved nearest-point search point cloud registration algorithm is used to register the acquired point cloud data of the outer surface and the local point cloud data of the standard part with the ideal reference point cloud, respectively. The rigid body transformation matrices of each relative to the ideal coordinate system are solved. The coordinate transformation relationship between the 3D visual measurement coordinate system and the microscopic depth-of-field measurement coordinate system is calculated based on the two rigid body transformation matrices. Finally, the pose of the microscopic depth-of-field measurement system in the 3D visual measurement system coordinate system is calculated.

Description

Technical Field

[0001] This invention relates to the field of high-precision structured light measurement technology, and in particular to a multi-scale visual measurement coordinate system, method, system, medium, and device. Background Technology

[0002] In the quality inspection of complex structural components, in-machine 3D measurement systems typically handle macroscopic overall positioning, while microscopic depth-of-field measurement systems are responsible for fine measurement of minute features. However, these two types of systems differ significantly in measurement mechanisms, imaging scales, optical characteristics, and clamping methods, making it difficult to directly establish a consistent, stable, and high-precision spatial mapping relationship between them. Current technologies lack a unified reference standard and coordinate system that can simultaneously adapt to the imaging characteristics of both systems and ensure no reduction in measurement accuracy, thus creating a technical bottleneck in practical engineering applications.

[0003] In-machine 3D measurement systems employ large fields of view and long-distance measurement, while microscopic depth-of-field measurement systems have a clear imaging area limited to only a few micrometers and strongly depend on the geometric relationship between the measured surface and the object's pixel size. The two types of systems differ significantly in clamping methods, coordinate axis alignment, working distance, and measurement posture. If the standard part does not consider the repeatability and posture constraints of the fixture, it may exhibit subtle but non-negligible posture deviations in both systems, leading to systematic errors in the spatial coordinate relationship matrix. Furthermore, due to the vast difference in spatial sampling density between 3D and microscopic systems, it is difficult to establish cross-scale, high-precision coordinate mapping using conventional registration methods in the absence of a unified feature structure.

[0004] Existing technologies urgently need a standard component structure and coordinate unification method that can simultaneously take into account the imaging distortion characteristics of 3D vision systems, the optical resolution of microscopic depth-of-field systems, and the consistency of clamping and positioning, in order to solve core technical problems such as the difficulty in unifying spatial coordinates, unstable imaging quality, and difficulty in jointly identifying features in cross-scale measurements.

[0005] The information disclosed in the background section is only for enhancing the understanding of the background of this invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the shortcomings, this invention provides a multi-scale visual measurement coordinate system—a method, system, medium, and device—that unifies coordinates and enables the acquisition of the pose of a microscopic depth-of-field measurement system in a three-dimensional measurement system coordinate system.

[0007] A multi-scale visual measurement coordinate system and method includes:

[0008] Step S1: Design a CAD model of a standard component for unifying the coordinate system of a large field-of-view 3D vision measurement system and a microscopic depth-of-field vision measurement system. The standard component is integrally machined from a cuboid base and a square pyramid located on top of it, wherein the bottom surface of the square pyramid has the same outline as the top surface of the cuboid and completely covers it.

[0009] Step S2: Use a 3D vision measurement system to perform a 3D scan on the standard part and collect point cloud data of part of the outer surface of the standard part. Use a microscopic depth of field measurement system to measure a part of the standard part and collect point cloud data of the part of the standard part.

[0010] Step S3: Randomly sample the point cloud of the CAD model of the standard part to generate an ideal reference point cloud with geometric structure and color annotation. Construct a high-dimensional feature vector based on geometric and color features. Use a high-dimensional ICP algorithm improved based on geometric and annotation features to register the collected point cloud data of the outer surface and the point cloud data of the local part of the standard part with the ideal reference point cloud. Solve the rigid body transformation matrix of each relative to the ideal coordinate system. Calculate the coordinate transformation relationship between the three-dimensional vision measurement coordinate system and the microscopic depth measurement coordinate system based on the two rigid body transformation matrices. Combine the motion of the motion platform to calculate the pose of the microscopic depth measurement system in the three-dimensional vision measurement system coordinate system.

[0011] In the aforementioned multi-scale visual measurement coordinate system and method, step S1 includes,

[0012] Step S11: Determine the overall size of the standard part based on the working distance and measurement field of view of the three-dimensional vision measurement system, so as to capture the overall contour features of the standard part in the three-dimensional vision measurement system.

[0013] Step S12: Determine the size and angle of the pyramidal part of the standard component based on the equipment parameters of the microscopic depth-of-field measurement system and the reconstruction algorithm;

[0014] Step S13: Surface treatment of standard parts. The outer surface of the standard parts is grouped and colored according to the adjacent relationship between pyramids and cuboids, which is used to distinguish similar features later.

[0015] Step S14: The surface roughness of the cuboid and pyramidal portions of the standard parts is set according to the smooth surface standard g, as shown in the following formula:

[0016]

[0017] in: Let be the angle of incidence of the light ray. The incident light wavelength, denoted as surface roughness, when g < 1, the object is considered to have a smooth surface.

[0018] In the aforementioned multi-scale visual measurement coordinate system and method, S11 includes,

[0019] S111,

[0020] The overall height of the standard part is determined by the working distance and field of view of the 3D measurement system. L is the nominal working distance of the measurement system, and W and H are the field of view of the measurement system at this working distance, respectively. The corresponding horizontal outer dimension Dx and the corresponding vertical outer dimension Dy of the standard part in the field of view satisfy the following necessary constraints:

[0021]

[0022] in, For safety margin coefficient,

[0023] To ensure that the vertex of the square pyramid is located in the center of the field of view, let the projected pixel coordinates of this vertex on the image plane be... The principal point coordinates are The pixel distance between a vertex and the principal point satisfies:

[0024]

[0025] in The central region of the field of view

[0026] S12 includes,

[0027] S121, the angle of the inclined plane of the pyramid is calculated based on the camera pixel size p, optical magnification M, system depth of field DF parameter, and the ideal pixel width range n of the inclined plane in the image, as shown in the following formula:

[0028] ,

[0029] A defined slope angle ensures that the clear imaging area of ​​the slope within the system's depth of field corresponds to n pixels, thereby guaranteeing that the grayscale gradient of the pyramid's slope edge can be effectively sampled in subsequent reconstruction algorithms, meeting the target measurement accuracy requirements.

[0030] In the aforementioned multi-scale visual measurement coordinate system and method, step S2 includes:

[0031] Step S21: Clamp the standard part onto the motion platform and move the motion platform within the depth and field of view of the 3D vision measurement system.

[0032] Step S22: Adjust the exposure value of the 3D vision measurement system according to the imaging quality, and use the single-frame measurement of the 3D measuring instrument to obtain the topographic point cloud containing the standard part, wherein the acquired point cloud includes at least one side of the cuboid base and at least one inclined surface of the square pyramid structure.

[0033] Step S23: The moving platform positions the top of the pyramidal part of the standard component at the center of the field of view of the microscopic depth-of-field measurement system, ensuring a clear image of the top.

[0034] Step S24: Use a microscopic depth of field measurement system to acquire an image sequence until the clearly imaged area disappears from the field of view of the microscopic depth of field measurement system, and use a microscopic depth of field algorithm to acquire the point cloud of the pyramidal part of the standard part.

[0035] Step S25: First, use a large depth-of-field lens to acquire an image of the top of the standard part, extract the grayscale values ​​of the pixels, and then switch to a small depth-of-field lens, ensuring that the optical axes of the two lenses are the same.

[0036] Step S26: Use a microscopic depth-of-field measurement system to obtain the point cloud of the pyramidal part of the standard part.

[0037] In the aforementioned multi-scale visual measurement coordinate system and method, step S23 includes,

[0038] S231, first use a three-dimensional vision measurement system to scan and reconstruct the standard part, then combine the statistical outlier removal method to perform point cloud filtering to obtain a noise-free single-area point cloud of the standard part. The statistical outlier removal method calculates the average distance between each point in the point cloud and its k nearest neighbors, thereby removing points that exceed the threshold.

[0039] S232, then based on the correspondence between the point cloud and the pixels of the 2D image acquired by the binocular camera, the ROI region of the standard part in the image is obtained, and the grayscale value of the pixels in it is extracted as one-dimensional feature information for subsequent point cloud registration. The relationship between the point cloud and the image is obtained based on the projection matrix obtained by the calibrated 3D measurement system.

[0040]

[0041] Where u and v are pixel coordinates, and X, Y, and Z are 3D coordinates. This is the projection matrix, from which the pixel set is determined:

[0042] .

[0043] In the aforementioned multi-scale visual measurement coordinate system and method, step S3 includes,

[0044] Step S31: Discretize the CAD model of the standard part to obtain a standard point cloud.

[0045] Step S32: Using a high-dimensional ICP algorithm improved based on geometric and annotation features, and taking the CAD point cloud as a reference, calculate the transformation matrix between the microscopic depth-of-field measurement point cloud and the 3D measurement point cloud. For any point P in the point cloud, construct its high-dimensional feature vector:

[0046]

[0047] Where XYZ are the three-dimensional coordinates of each point in the point cloud, and n x ,n y ,n z For each point, the normal vector is given, and I(P) represents the grayscale feature of that point. A high-dimensional distance metric function is established based on the high-dimensional feature vector. For different points P and Q in the point cloud:

[0048]

[0049] Among them, w g w n w I These are weights for geometric distance, normal consistency, and grayscale difference, respectively. A high-dimensional ICP error function is constructed based on the matching set.

[0050]

[0051] The goal is to find the rigid body transformation that minimizes the error:

[0052]

[0053] An incremental iterative model is constructed by performing a first-order Taylor expansion on the error term E. After removing the mean from the set of matching point pairs, the optimal transformation is obtained through singular value decomposition. This process is repeated iteratively until the convergence condition is met.

[0054] ;

[0055] Step S33: Using the rotation matrix R×, combined with the x, y, and z axis displacements of the motion platform moving from the three-dimensional measurement position to the microscopic measurement position, a translation matrix is ​​obtained, ultimately yielding the transformation relationship between the coordinate systems of the microscopic measurement system and the three-dimensional vision system.

[0056] In the multi-scale visual measurement coordinate system method, when the standard part is scanned and point cloud data of part of the outer surface of the standard part is collected by the three-dimensional visual measurement system, a three-dimensional point cloud containing at least one side of a cuboid and at least one inclined surface of a square pyramid is obtained, and the corresponding two-dimensional image is acquired simultaneously. The grayscale information of the image is mapped to the point cloud to form a textured three-dimensional point cloud. The correspondence between the three-dimensional coordinates of the point cloud and the pixel coordinates of the two-dimensional image is established through the calibration projection matrix of the three-dimensional visual measurement system. The pixel grayscale values ​​in the area of ​​the standard part are extracted as additional feature dimensions of the point cloud.

[0057] A system for performing the method includes:

[0058] A 3D vision measurement system is configured to acquire macroscopic point clouds of standard parts and map image textures;

[0059] A microscopic depth-of-field visual measurement system is configured to acquire point clouds of the vertex portion of a standard part's cone.

[0060] A motion platform, which can movably mount the standard components;

[0061] The processor calculates the pose of the microscopic depth-of-field measurement system in the coordinate system of the three-dimensional vision measurement system.

[0062] A computer storage medium including computer instructions that, when run on a computer, cause the computer to perform the method.

[0063] An electronic device, the electronic device comprising:

[0064] Memory, processor, and computer programs stored in memory and executable on the processor, wherein,

[0065] The processor implements the method when executing the program.

[0066] Compared with existing technologies, this invention has the following advantages: This invention designs a standard component for unifying the coordinate system of large field-of-view 3D vision measurement equipment and the coordinate system of microscopic depth-of-field vision measurement equipment, which can better adapt to the characteristics of the two vision systems. Based on an improved 2D-3D hybrid high-dimensional ICP algorithm using geometric and annotation features, it combines geometric and color features to calculate the transformation matrix for point cloud registration, and reduces the design complexity and manufacturing difficulty of the standard component from an algorithmic perspective. By combining the self-designed standard component and the improved ICP algorithm, a unified coordinate system is achieved for two measurement systems with different field-of-view sizes. Attached Figure Description

[0067] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0068] In the attached diagram:

[0069] Figure 1 This is a flowchart of the present invention;

[0070] Figure 2 This is a model drawing of a standard part designed in this invention;

[0071] Figure 3This is a schematic diagram of the field of view of the three-dimensional measurement system upon which the design of the standard parts of this invention is based;

[0072] Figure 4 A schematic diagram of the clearly imaged area of ​​the microscopic depth-of-field measurement system upon which the design of the standard parts of this invention is based;

[0073] Figure 5 This is a schematic diagram of the 2D-3D hybrid high-dimensional ICP algorithm based on geometric and color features improved in this invention.

[0074] Figure 6 This is a schematic diagram of the device platform structure used in this invention;

[0075] Figure 7 This is a diagram showing the effect of microstructure positioning and measurement achieved using this method in the present invention.

[0076] The present invention will be further explained below with reference to the accompanying drawings and embodiments. Detailed Implementation

[0077] Specific embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While specific embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0078] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0079] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0080] like Figures 1 to 7 As shown, the multi-scale visual measurement coordinate system method includes the following steps:

[0081] Step S1: Design a CAD model of a standard component for unifying the coordinate system of a large field-of-view 3D vision measurement system and a microscopic depth-of-field vision measurement system. The standard component is integrally machined from a cuboid base and a square pyramid located on top of it, wherein the bottom surface of the square pyramid has the same outline as the top surface of the cuboid and completely covers it.

[0082] Step S2: Use a 3D vision measurement system to perform a 3D scan on the standard part and collect point cloud data of part of the outer surface of the standard part. Use a microscopic depth of field measurement system to measure a part of the standard part and collect point cloud data of the part of the standard part.

[0083] Step S3: Randomly sample the point cloud of the CAD model of the standard part to generate an ideal reference point cloud with geometric structure and color annotation. Construct a high-dimensional feature vector based on geometric and color features. Use a high-dimensional ICP algorithm improved based on geometric and annotation features to register the collected point cloud data of the outer surface and the point cloud data of the local part of the standard part with the ideal reference point cloud. Solve the rigid body transformation matrix of each relative to the ideal coordinate system. Calculate the coordinate transformation relationship between the three-dimensional vision measurement coordinate system and the microscopic depth measurement coordinate system based on the two rigid body transformation matrices. Combine the motion of the motion platform to calculate the pose of the microscopic depth measurement system in the three-dimensional vision measurement system coordinate system.

[0084] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, step S1 includes:

[0085] Step S11: Determine the overall size of the standard part based on the working distance and measurement field of view of the three-dimensional vision measurement system, so as to capture the overall contour features of the standard part in the three-dimensional vision measurement system.

[0086] Step S12: Determine the angle of the inclined plane of the pyramidal part of the standard part based on the camera pixel size, optical magnification, system depth of field, and ideal pixel width range of the inclined plane in the imaging of the microscopic depth of field measurement system.

[0087] Step S13: Surface treatment of standard parts. The outer surface of the standard parts is grouped and colored according to the adjacent relationship between pyramids and cuboids, which is used to distinguish similar features later.

[0088] Step S14: The surface roughness of the standard cuboid and pyramidal parts is set according to the smooth surface standard g, as shown in the following formula:

[0089]

[0090] in: Let be the angle of incidence of the light ray. The incident light wavelength (in a three-dimensional measurement system, it is the light emitted by the projector; in a microscopic measurement system, it is the coaxial light used in the measurement system). Surface roughness. The generally accepted standard for a smooth surface is that an object is considered to have a smooth surface when g < 1.

[0091] Three-dimensional measurement systems rely on diffuse reflection to ensure the stability of structured light fringes / features, while specular reflection can cause fringe breakage and affect the acquisition of local feature information. Microscopic depth-of-field measurement systems rely on local gradient sharpness to ensure sharpness, while excessive surface roughness can lead to too many local high-reflectivity spots, affecting accuracy. Therefore, the surfaces of different parts of the standard parts are machined according to the smooth surface standard to ensure that the surface quality requirements of the standard parts for acquiring features by different systems are met.

[0092] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, S11 includes,

[0093] S111, the overall height of the standard part is determined by the working distance and field of view of the 3D measurement system. L is the nominal working distance of the measurement system, and W and H are the field of view of the measurement system at this working distance, respectively. The corresponding horizontal outer dimension Dx and the corresponding vertical outer dimension Dy of the standard part in the field of view satisfy the following necessary constraints:

[0094]

[0095] in, For safety margin coefficient,

[0096] To ensure that the vertex of the square pyramid is located in the center of the field of view, let the projected pixel coordinates of this vertex on the image plane be... The principal point coordinates are The pixel distance between a vertex and the principal point satisfies:

[0097]

[0098] in The central region of the field of view

[0099] S12 includes,

[0100] S121, the angle of the inclined plane of the pyramid is calculated based on the camera pixel size p, optical magnification M, system depth of field DF parameter, and the ideal pixel width range n of the inclined plane in the image, as shown in the following formula:

[0101] ,

[0102] A defined slope angle ensures that the clear imaging area of ​​the slope within the system's depth of field corresponds to n pixels, thereby guaranteeing that the grayscale gradient of the pyramid's slope edge can be effectively sampled in subsequent reconstruction algorithms, meeting the target measurement accuracy requirements.

[0103] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, step S2 includes:

[0104] Step S21: Clamp the standard part onto the motion platform and move the motion platform within the depth of field and field of view of the three-dimensional vision measurement system;

[0105] Step S22: Adjust the exposure value of the three-dimensional vision measurement system according to the imaging quality, and adjust the pose of the standard parts clamped on the motion platform to ensure that the three-dimensional vision measurement field of view includes at least two adjacent sides of the cuboid base and two adjacent inclined surfaces of the quadrangular pyramid structure.

[0106] Step S23: The standard part is scanned and reconstructed using a 3D measurement system. Then, point cloud filtering is performed using a statistical outlier removal method to obtain a noise-free single-area point cloud of the standard part. Based on the correspondence between the point cloud and the pixels of the 2D image obtained by the binocular camera, the ROI region of the standard part in the image is obtained, and the grayscale values ​​of the pixels are extracted as one-dimensional feature information for subsequent point cloud registration.

[0107] Step S24: The moving platform places the top of the pyramidal part of the standard part at the center of the field of view of the microscopic depth-of-field measurement system, ensuring a clear image of the top, and records the displacements along the x, y, and z axes respectively.

[0108] Step S25: First, use a large depth-of-field lens to acquire an image of the top of the standard part and extract the grayscale values ​​of the pixels. Then, replace it with a small depth-of-field lens and make the optical axes of the two lenses the same.

[0109] Step S26: Obtain the point cloud of the pyramidal part of the standard part using a microscopic depth-of-field measurement system. Since the point cloud is obtained using the conversion relationship between the pixel size of the image and the actual size in the microscopic measurement system, each point in the point cloud corresponds one-to-one with each pixel in the image. Therefore, the pixel values ​​obtained in step S25 can be linked to each point in the point cloud for use in subsequent algorithms.

[0110] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, step S22 includes,

[0111] S231, first use a three-dimensional vision measurement system to scan and reconstruct the standard part, then combine the statistical outlier removal method to perform point cloud filtering to obtain a noise-free single-area point cloud of the standard part. The statistical outlier removal method calculates the average distance between each point in the point cloud and its k nearest neighbors, thereby removing points that exceed the threshold.

[0112] S232, then based on the correspondence between the point cloud and the pixels of the 2D image acquired by the binocular camera, the ROI region of the standard part in the image is obtained, and the grayscale value of the pixels is extracted as one-dimensional feature information for subsequent point cloud registration. The relationship between the point cloud and the image is obtained based on the projection matrix obtained by the calibrated 3D measurement system.

[0113]

[0114] Where uv represents pixel coordinates, and X, Y, and Z represent 3D coordinates. This is the projection matrix, from which the pixel set is determined based on the projection results. :

[0115] .

[0116] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, step S3 includes,

[0117] Step S31: Discretize the CAD model of the standard part to obtain a standard point cloud.

[0118] Step S32: A high-dimensional ICP algorithm based on geometric and annotation features is used. This algorithm calculates the transformation matrix between the microscopic depth-of-field measurement point cloud and the 3D measurement point cloud, using the CAD point cloud as a reference. Specifically, for any point P in the point cloud, its high-dimensional feature vector f(P) is constructed:

[0119]

[0120] The geometric features include: XYZ, i.e., the three-dimensional coordinates of each point in the point cloud, n x ,n y ,n z， That is, the normal vector of each point is labeled with the feature I(P), which is the gray-level feature of that point. A high-dimensional distance metric function D(P,Q) is established based on the high-dimensional feature vector. For different points P and Q in the point cloud:

[0121]

[0122] Among them, w g w n w I These are weights for geometric distance, normal consistency, and grayscale difference, respectively. A high-dimensional ICP error function E is constructed based on the matching set.

[0123]

[0124] The goal is to find the rigid body transformation that minimizes the error:

[0125]

[0126] An incremental iterative model is constructed by performing a first-order Taylor expansion on the error term E. After removing the mean from the set of matching point pairs, the optimal transformation is obtained through singular value decomposition. This process is repeated iteratively until the convergence condition is met.

[0127] ;

[0128] In step S33, the rotation matrix R× obtained in S32 is used in conjunction with the x, y, and z axis displacements of the motion platform from the three-dimensional measurement position to the microscopic measurement position in step 23 to obtain the translation matrix, and finally the transformation relationship between the coordinate systems of the microscopic measurement system and the three-dimensional vision system is obtained.

[0129] In a preferred embodiment of the multi-scale visual measurement coordinate system and method, when the standard part is scanned and point cloud data of part of the outer surface of the standard part is collected by the three-dimensional visual measurement system, a three-dimensional point cloud containing at least one side of a cuboid and at least one inclined surface of a pyramid is obtained, and the corresponding two-dimensional image is acquired simultaneously. The grayscale information of the image is mapped to the point cloud to form a textured three-dimensional point cloud. The correspondence between the three-dimensional coordinates of the point cloud and the pixel coordinates of the two-dimensional image is established through the calibration projection matrix of the three-dimensional visual measurement system, and the pixel grayscale values ​​in the area of ​​the standard part are extracted as additional feature dimensions of the point cloud.

[0130] A system for performing the method includes:

[0131] A 3D vision measurement system is configured to acquire macroscopic point clouds of standard parts and map image textures;

[0132] A microscopic depth-of-field visual measurement system is configured to acquire point clouds of the vertex portion of a standard part's cone.

[0133] A motion platform, which can movably mount the standard components;

[0134] The processor calculates the pose of the microscopic depth-of-field measurement system in the coordinate system of the three-dimensional vision measurement system.

[0135] A computer storage medium including computer instructions that, when run on a computer, cause the computer to perform the method.

[0136] An electronic device, the electronic device comprising:

[0137] Memory, processor, and computer programs stored in memory and executable on the processor, wherein,

[0138] The processor implements the method when executing the program.

[0139] In one embodiment, S1, the standard component design steps for a coordinate system for a multi-scale measurement system include:

[0140] S11, such as Figure 2 As shown, the standard component for coordinate system one of the two measurement systems consists of a cuboid and a square pyramid.

[0141] S111, the overall height of the standard part is determined by the field-of-view parameters of the 3D measurement system, and the field-of-view size relationship is as follows: Figure 3 As shown. Where L is the nominal working distance of the measurement system, and W and H are the field of view sizes of the measurement system at that working distance. The outer dimensions of the standard part in the field of view. (Corresponding to the horizontal direction) and (Corresponding to the vertical direction) The following necessary constraints should be met:

[0142]

[0143] in, This is a safety margin factor used to compensate for factors such as clamping errors, imaging distortion, and shooting offset.

[0144] Furthermore, to ensure that the vertex of the square pyramid is located in the central region of the field of view (low distortion region), let the projected pixel coordinates of this vertex on the image plane be... The principal point coordinates are The pixel distance between a vertex and the principal vertex is specified to satisfy:

[0145]

[0146] in This is the central region of the field of view.

[0147] S112, the base portion of the standard part is designed as a cuboid structure. On one hand, the cuboid's shape possesses regular planar geometry, and the orthogonal relationship between its six faces provides a clear spatial reference for the 3D vision system, improving the geometric stability of point cloud coordinates and the determinism of attitude calculation. On the other hand, the large, flat surfaces of the cuboid can generate high-quality, continuous point cloud data in the 3D vision system, avoiding sparse points or feature instability problems caused by curved surface structures, thereby improving the accuracy of subsequent point cloud matching and model fitting. Based on its regular shape, the cuboid structure also facilitates stable clamping using conventional mechanical fixtures, which helps reduce vibration disturbances and attitude deviations during measurement, improving the overall repeatability and reliability of the system.

[0148] S12, the size of the pyramidal part of the standard part is determined according to the field parameters of the microscopic depth of field measurement system.

[0149] S121, the clear imaging area of ​​the microscopic depth-of-field measurement system is as follows: Figure 4As shown; the angle of the pyramid's slope is theoretically calculated based on the camera pixel size p, optical magnification M, system depth of field DF parameter of the microscopic depth of field measurement system, and the ideal pixel width range n that the slope needs to remain sharp in imaging, as shown in the following formula:

[0150]

[0151] The slope angle determined by the above relationship ensures that the clear imaging area of ​​the slope within the system's depth of field corresponds to n pixels, thereby guaranteeing that the gray-level gradient of the pyramid's slope edge can be effectively sampled in subsequent reconstruction algorithms, satisfying the target measurement accuracy.

[0152] S122, for ease of manufacturing, features a quadrangular pyramid shape for the pyramidal portion, maintaining the same base area as the cuboid portion to reduce manufacturing complexity. This also facilitates surface grouping for subsequent color processing and provides features for point cloud data generation.

[0153] S13, standard parts with surface treatment, such as Figure 2 As shown, the outer surfaces of standard parts are grouped and colored according to the adjacency relationship between pyramids and cuboids. Each pyramidal face and the cuboid side directly connected to it are considered as a group. Among them, two adjacent groups of faces use the same color as the first color scheme, and the other two groups use a second color scheme that is different from the first color scheme, thus forming two groups of face groups with clear color distinction, in order to solve the problem of multiple solutions for point cloud pairing caused by similar geometric features.

[0154] S14. The surface roughness of the cuboid and pyramidal parts of the design standard part is below a certain threshold to ensure that the surface quality requirements for obtaining high-precision features of the standard part can be met by different systems.

[0155] S2, the steps for acquiring multi-feature point clouds based on 2D-3D hybridization include:

[0156] S21, clamp the standard part onto the motion platform and move the displacement platform within the depth of field and field of view of the three-dimensional measurement system.

[0157] S22, adjust the exposure value of the three-dimensional vision measurement system according to the imaging quality, and adjust the pose of the standard parts clamped on the motion platform to ensure that the three-dimensional vision measurement field of view includes at least two adjacent sides of the cuboid base and two adjacent inclined surfaces of the quadrangular pyramid structure, so as to ensure that the on-machine system can extract multi-faceted geometric features for spatial positioning.

[0158] S23, for three-dimensional measurement systems, such as Figure 5 As shown on the far left, the effective information in the image corresponding to the point cloud of the object under test only accounts for a portion of all pixels in the image.

[0159] S231, In this method, a three-dimensional measurement system is first used to scan and reconstruct the standard part, and then a statistical outlier removal method is used to filter the point cloud to obtain a noise-free single-area point cloud of the standard part. The statistical outlier removal method calculates the average distance between each point in the point cloud and its k nearest neighbors, thereby removing points that exceed the threshold.

[0160] S232, then based on the correspondence between the point cloud and the pixels of the two-dimensional image obtained by the stereo camera, the standard part ROI region in the image is obtained, and the gray value of the pixels in it is extracted as one-dimensional feature information for subsequent point cloud registration.

[0161] The relationship between point clouds and images can be obtained from the projection matrix obtained by calibrating the 3D measurement system:

[0162]

[0163] Where u and v are pixel coordinates, and X, Y, and Z are 3D coordinates. This is the projection matrix, which allows us to determine the pixel set based on the projection results:

[0164]

[0165] S24, the moving motion platform places the top of the pyramid of the standard part at the center of the microscopic measurement field of view, and makes the top of the part clearly imaged.

[0166] S25, for microscopic depth-of-field measurement systems, such as Figure 5 As shown on the right, because the system has a small field of view, a large depth-of-field lens is used to first acquire the image of the top of the standard part, extract the grayscale values ​​of the pixels, and then use a subsequent pixel-point cloud mapping algorithm to add dimensional information to the point cloud.

[0167] To reduce the impact of noise and measurement errors, a weighted average is used to process the grayscale values ​​of the pixels:

[0168]

[0169] Where N(u,v) is the pixel neighborhood, and the weights are... Take a custom interpolation distance, and use this grayscale value as a one-dimensional component of the high-dimensional description of the point cloud.

[0170] Then, replace the lens with a shallow depth of field lens and make the optical axes of the two lenses the same.

[0171] Step S26: Obtain the point cloud of the pyramidal part of the standard part using a microscopic depth-of-field measurement system. Since the point cloud is obtained using the conversion relationship between the pixel size of the image and the actual size in the microscopic measurement system, each point in the point cloud corresponds one-to-one with each pixel in the image. Therefore, the pixel values ​​obtained in step S25 can be linked to each point in the point cloud for use in subsequent algorithms.

[0172] S3, the steps of the high-dimensional ICP method based on geometric and labeled features include:

[0173] S31, such as Figure 5 As shown, after acquiring the point cloud of a standard part using two vision systems, it is necessary to discretize the standard CAD model into a point cloud and map color information to the standard point cloud according to the design model. This step can be achieved using existing software.

[0174] S32, For any point P in the point cloud, construct its high-dimensional feature vector:

[0175]

[0176] Where X, Y, and Z are the three-dimensional coordinates of each point in the point cloud, and n x ,n y ,n z For each point, the normal vector is given, and I(P) represents the grayscale feature of that point. A high-dimensional distance metric function is then established based on the high-dimensional feature vectors, for different points P and Q in the point cloud:

[0177]

[0178] Among them, w g w n w I These are the geometric distance weight, the normal uniformity weight, and the grayscale difference weight, respectively.

[0179] S33, Construct a high-dimensional ICP error function based on the matching set:

[0180]

[0181] The goal is to find the rigid body transformation that minimizes the error:

[0182]

[0183] An incremental iterative model is constructed by performing a first-order Taylor expansion on the error term E. After removing the mean from the set of matching point pairs, the optimal transformation is obtained through singular value decomposition. This process is repeated iteratively until the convergence condition is met.

[0184]

[0185] S34, through combination Figure 6 The inverse kinematics model shown analyzes the displacement of the platform during rotation and translation along each axis, thereby accurately calculating the spatial pose of the microscopic measurement system in the coordinate system of the three-dimensional vision system. This process uses known kinematic parameters to inversely determine the actual position and orientation of the microscopic measurement system, achieving accurate quantification of the coordinate relationship between the two measurement systems.

[0186] like Figure 1 As shown, a multi-scale visual measurement coordinate system and method based on standard parts and improved ICP includes a coordinate system standard part design, a 2D-3D hybrid point cloud acquisition algorithm, and a high-dimensional ICP algorithm improved based on geometric and annotation features. First, the overall and local structures of the standard part are designed based on a 3D measurement system and a microscopic measurement system. Then, combined with the 2D-3D hybrid point cloud acquisition algorithm, the local feature point clouds of the standard part are acquired using the 3D measurement system and the microscopic depth measurement system, respectively. Subsequently, the standard part CAD model is discretized to obtain the standard point cloud. Then, the coordinate transformation matrix of the two vision systems is obtained using the high-dimensional ICP algorithm improved based on geometric and annotation features. Finally, the unification of the on-machine 3D visual coordinate system and the microscopic depth measurement coordinate system is achieved by combining inverse kinematics and displacement of each axis.

[0187] like Figure 2 As shown, the standard part consists of a cuboid and a square pyramid connected together. The cuboid portion serves as the base of the standard part, with a square base. The square pyramid portion connects to the top rectangular face of the cuboid, and the base of the pyramid has the same shape and area as the top face of the cuboid. Each inclined face of the square pyramid and its directly connected side face of the cuboid are considered as a group. Two adjacent groups of faces use the same color as the first color scheme, while the other two adjacent groups use a second color scheme that differs from the first, thus forming two clearly distinguishable face group structures.

[0188] like Figure 3 The diagram shows the field of view of the 3D vision measurement system used in this method, illustrating the imaging area of ​​the measurement system at a given working distance and the geometric relationship between its width W, height H, and working distance L. Based on the field of view relationships in this diagram, the external dimensions and spatial layout of the standard part, as well as the geometric visibility conditions that must be met in the subsequent coordinate system step, can be derived to ensure that the accuracy and visibility of the 3D measurement results are maximized.

[0189] like Figure 4 The diagram shown is a clear imaging region of the microscopic depth-of-field measurement system used in this method, illustrating the depth of field DF, the object surface angle α, and the imaging region width W. x Based on the imaging relationship shown in this figure, the dimensions of the pyramidal part of the standard component can be derived to ensure the maximum accuracy of the microscopic depth-of-field measurement results. For example... Figure 5As shown, the coordinate system of the 3D measurement system and the microscopic depth-of-field measurement system is unified through the joint processing of multi-source data. First, the left and right sides respectively display 2D images acquired by the on-machine 3D vision system. Two local point clouds of the standard part are obtained through binocular reconstruction and depth-of-field reconstruction. Furthermore, the color information of the 2D images is mapped onto the surface of the reconstructed point cloud for subsequent feature enhancement and matching constraints. The ideal point cloud at the top represents the theoretical model of the standard part used for calibration or reference, providing unified morphological features and the expected geometric framework. Three sets of point cloud data (binocular reconstructed point cloud, microscopic depth-of-field reconstructed point cloud, and ideal point cloud) are input into the "ICP Registration and Coordinate Unification" module shown in the center. This module matches, aligns, and transforms the coordinate system of point clouds from different sources, achieving a consistent representation of the data under a unified coordinate system.

[0190] like Figure 6 The image shows a schematic diagram of the motion platform model and machine tool coordinate system used in this method. The blue model represents the 3D vision measurement system, and the yellow model represents the microscopic depth-of-field measurement system. Figure 7 The image shown is a final coordinate system diagram of this method. The left image is an external image captured by the microscopic depth-of-field measurement system after the blade automatically moves to the location of the micro-hole to be measured, based on the coordinates of the micro-hole position, after acquiring point cloud data at the three-dimensional measurement position. The right image is the first measurement image acquired during the reconstruction process of the microscopic depth-of-field measurement system.

[0191] Furthermore, the design of the dedicated standard parts in this invention takes into account the physical constraints of two types of systems: the cuboid base provides a stable orthogonal geometric benchmark at a macroscopic scale, facilitating robust spatial pose extraction by the 3D vision system; the apex of the square pyramid has its bevel angle precisely designed according to the depth of field, magnification, and pixel resolution of the microscopic system, ensuring that clear edges with significant gradient changes can still be obtained within a very small field of view, providing reliable topographic data for high-precision local reconstruction. Simultaneously, the surface grouping color matching strategy overcomes the ambiguity in point cloud matching easily caused by similar planes in traditional standard parts, guiding feature correspondence through prior color information, significantly improving the uniqueness and stability of cross-modal registration.

[0192] Secondly, the 2D-3D hybrid point cloud generation mechanism uses image grayscale as an enhanced dimension of the point cloud, introducing texture semantic information while preserving the three-dimensional geometric structure. For the 3D system, a calibration projection matrix is ​​used to map the grayscale of the ROI region in the binocular image to a sparse point cloud; for the microscopic system, a neighborhood-weighted averaging method is used to improve the local grayscale signal-to-noise ratio, enabling even micrometer-scale structures to possess distinguishable photometric features. This strategy effectively bridges the gap between the two types of systems in terms of sampling density (millimeter-level vs. micrometer-level) and data dimensionality.

[0193] Building upon this foundation, a high-dimensional improved ICP algorithm based on the fusion of geometric and color features constructs a seven-dimensional feature vector comprising coordinates, normal vectors, and grayscale values. An adjustable-weight distance metric function is introduced to simultaneously constrain geometric consistency and appearance similarity during iterative registration. Compared to traditional ICP methods that rely solely on spatial distance, this algorithm is more robust to noise, local missing values, and scale mismatches. Even with minor machining deviations or clamping disturbances in standard parts, it can still achieve sub-pixel level registration accuracy.

[0194] Finally, by jointly solving the rotation matrix obtained through point cloud registration and the platform displacement, the mathematical transformation relationship obtained from point cloud registration is transformed into the physical pose of the microscopic measurement system in the machine tool's global coordinate system, thereby establishing a unified framework for repeatable, traceable, and high-precision cross-scale measurement. This technology system not only achieves seamless integration of "macroscopic positioning - microscopic measurement" but also provides key support for the automated in-machine inspection of complex components such as aero-engine blades, significantly improving the reliability and efficiency of the measurement link in the high-precision manufacturing closed loop.

[0195] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A multi-scale visual measurement coordinate system and method, characterized in that, Includes the following steps: Step S1: Design a CAD model of a standard component for unifying the coordinate system of a large field-of-view 3D vision measurement system and a microscopic depth-of-field vision measurement system. The standard component is integrally machined from a cuboid base and a square pyramid located on top of it, wherein the bottom surface of the square pyramid has the same outline as the top surface of the cuboid and completely covers it. Step S2: Use a 3D vision measurement system to perform a 3D scan on the standard part and collect point cloud data of part of the outer surface of the standard part. Use a microscopic depth of field measurement system to measure a part of the standard part and collect point cloud data of the part of the standard part. Step S3: Randomly sample the point cloud of the CAD model of the standard part to generate an ideal reference point cloud with geometric structure and color annotation. Construct a high-dimensional feature vector based on geometric and color features. Use a high-dimensional ICP algorithm improved based on geometric and annotation features to register the collected point cloud data of the outer surface and the point cloud data of the local part of the standard part with the ideal reference point cloud. Solve the rigid body transformation matrix of each relative to the ideal coordinate system. Calculate the coordinate transformation relationship between the three-dimensional vision measurement coordinate system and the microscopic depth measurement coordinate system based on the two rigid body transformation matrices. Combine the motion of the motion platform to calculate the pose of the microscopic depth measurement system in the three-dimensional vision measurement system coordinate system.

2. The multi-scale visual measurement coordinate system and method according to claim 1, characterized in that, Preferably, step S1 includes, Step S11: Determine the overall size of the standard part based on the working distance and measurement field of view of the three-dimensional vision measurement system, so as to capture the overall contour features of the standard part in the three-dimensional vision measurement system. Step S12: Determine the size and angle of the pyramidal part of the standard component based on the equipment parameters of the microscopic depth-of-field measurement system and the reconstruction algorithm; Step S13: Surface treatment of standard parts. The outer surface of the standard parts is grouped and colored according to the adjacent relationship between pyramids and cuboids, which is used to distinguish similar features later. Step S14: The surface roughness of the standard cuboid and pyramidal parts is set according to the smooth surface standard g, as shown in the following formula: ， in: Let be the angle of incidence of the light ray. The incident light wavelength, denoted as surface roughness, when g < 1, the object is considered to have a smooth surface.

3. The multi-scale visual measurement coordinate system and method according to claim 2, characterized in that, S11 includes, S111， The overall height of the standard part is determined by the working distance and field of view of the 3D measurement system. L is the nominal working distance of the measurement system, and W and H are the field of view of the measurement system at this working distance, respectively. The corresponding horizontal outer dimension Dx and the corresponding vertical outer dimension Dy of the standard part in the field of view satisfy the following necessary constraints: ， in, For safety margin coefficient, To ensure that the vertex of the square pyramid is located in the center of the field of view, let the projected pixel coordinates of this vertex on the image plane be... The principal point coordinates are The pixel distance between a vertex and the principal point satisfies: ， in The central region of the field of view S12 includes, S121, the angle of the inclined plane of the pyramid is calculated based on the camera pixel size p, optical magnification M, system depth of field DF parameter, and the ideal pixel width range n of the inclined plane in the image, as shown in the following formula: ， A defined slope angle ensures that the clear imaging area of ​​the slope within the system's depth of field corresponds to n pixels, thereby guaranteeing that the grayscale gradient of the pyramid's slope edge can be effectively sampled in subsequent reconstruction algorithms, meeting the target measurement accuracy requirements.

4. The multi-scale visual measurement coordinate system and method according to claim 1, characterized in that, Step S2 includes, Step S21: Clamp the standard part onto the motion platform and move the motion platform within the depth and field of view of the 3D vision measurement system. Step S22: Adjust the exposure value of the 3D vision measurement system according to the imaging quality, and use the single-frame measurement of the 3D measuring instrument to obtain the topographic point cloud containing the standard part, wherein the acquired point cloud includes at least one side of the cuboid base and at least one inclined surface of the square pyramid structure. Step S23: The moving platform positions the top of the pyramidal part of the standard component at the center of the field of view of the microscopic depth-of-field measurement system, ensuring a clear image of the top. Step S24: Use a microscopic depth of field measurement system to acquire an image sequence until the clearly imaged area disappears from the field of view of the microscopic depth of field measurement system, and use a microscopic depth of field algorithm to acquire the point cloud of the pyramidal part of the standard part. Step S25: First, use a large depth-of-field lens to acquire an image of the top of the standard part, extract the grayscale values ​​of the pixels, and then switch to a small depth-of-field lens, ensuring that the optical axes of the two lenses are the same. Step S26: Use a microscopic depth-of-field measurement system to obtain the point cloud of the pyramidal part of the standard part.

5. A multi-scale visual measurement coordinate system and method according to claim 4, characterized in that, Step S23 includes, S231, first use a three-dimensional vision measurement system to scan and reconstruct the standard part, then combine the statistical outlier removal method to perform point cloud filtering to obtain a noise-free single-area point cloud of the standard part. The statistical outlier removal method calculates the average distance between each point in the point cloud and its k nearest neighbors, thereby removing points that exceed the threshold. S232, then based on the correspondence between the point cloud and the pixels of the 2D image acquired by the binocular camera, the ROI region of the standard part in the image is obtained, and the grayscale value of the pixels in it is extracted as one-dimensional feature information for subsequent point cloud registration. The relationship between the point cloud and the image is obtained based on the projection matrix obtained by the calibrated 3D measurement system. ， Where u and v are pixel coordinates, and X, Y, and Z are 3D coordinates. This is the projection matrix, from which the pixel set is determined: 。 6. A multi-scale visual measurement coordinate system and method according to claim 1, characterized in that, Step S3 includes, Step S31: Discretize the CAD model of the standard part to obtain a standard point cloud. Step S32: Using a high-dimensional ICP algorithm improved based on geometric and annotation features, and taking the CAD point cloud as a reference, calculate the transformation matrix between the microscopic depth-of-field measurement point cloud and the 3D measurement point cloud. For any point P in the point cloud, construct its high-dimensional feature vector: ， Where XYZ are the three-dimensional coordinates of each point in the point cloud, and n x ,n y ,n z For each point, the normal vector is given, and I(P) represents the grayscale feature of that point. A high-dimensional distance metric function is established based on the high-dimensional feature vector. For different points P and Q in the point cloud: ， Among them, w g w n w I These are weights for geometric distance, normal consistency, and grayscale difference, respectively. A high-dimensional ICP error function is constructed based on the matching set. ， The goal is to find the rigid body transformation that minimizes the error: ， An incremental iterative model is constructed by performing a first-order Taylor expansion on the error term E. After removing the mean from the set of matching point pairs, the optimal transformation is obtained through singular value decomposition. This process is repeated iteratively until the convergence condition is met. ； Step S33: Using the rotation matrix R×, combined with the x, y, and z axis displacements of the motion platform moving from the three-dimensional measurement position to the microscopic measurement position, a translation matrix is ​​obtained, ultimately yielding the transformation relationship between the coordinate systems of the microscopic measurement system and the three-dimensional vision system.

7. A multi-scale visual measurement coordinate system and method according to claim 1, characterized in that, When using a 3D vision measurement system to perform 3D scanning to collect point cloud data of part of the outer surface of the standard part, a 3D point cloud containing at least one side of a cuboid and at least one inclined surface of a square pyramid is obtained, and the corresponding 2D image is acquired simultaneously. The grayscale information of the image is mapped to the point cloud to form a textured 3D point cloud. Through the calibration projection matrix of the 3D vision measurement system, the correspondence between the 3D coordinates of the point cloud and the pixel coordinates of the 2D image is established, and the pixel grayscale values ​​in the area of ​​the standard part are extracted as additional feature dimensions of the point cloud.

8. A system for performing the method as described in any one of claims 1 to 7, characterized in that, It includes: A 3D vision measurement system is configured to acquire macroscopic point clouds of standard parts and map image textures; A microscopic depth-of-field visual measurement system is configured to acquire point clouds of the vertex portion of a standard part's cone. A motion platform, which can movably mount the standard components; The processor calculates the pose of the microscopic depth-of-field measurement system in the coordinate system of the three-dimensional vision measurement system.

9. A computer storage medium, characterized in that, The storage medium includes computer instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes: Memory, processor, and computer programs stored in memory and executable on the processor, wherein, When the processor executes the program, it implements the method as described in any one of claims 1-7.

Images