Non-contact detection method for complex assembly workpiece assembly based on laser three-dimensional scanning

Abstract

The application discloses a complex assembly workpiece assembly non-contact detection method based on laser three-dimensional scanning and relates to the technical field of automation. The application aims at the problem of large amount of manual participation and low efficiency of the technical route of the existing three-dimensional digital model comparison measurement, realizes a high-precision large-size assembly workpiece rapid three-dimensional scanning system based on linear laser structured light scanning combined with industrial robot technology, and can realize rapid three-dimensional point cloud measurement. The application designs a full-automatic measurement method for large-size assembly workpieces based on three-dimensional point cloud data, realizes full-automatic point cloud data processing through the design of point cloud alignment, point cloud classification and geometric information parameterization method, is convenient for efficient and accurate assembly between complex assembly workpiece parts, and effectively avoids the problem of large amount of manual participation and low efficiency of the technical route of the existing three-dimensional digital model comparison measurement.

Description

Technical Field

[0001] This invention relates to the field of automation technology, and in particular to a non-contact inspection method for assembling complex workpieces based on laser three-dimensional scanning. Background Technology

[0002] In modern component assembly technology, the inspection and assembly of multiple components are generally completed through a relatively complex three-dimensional digital model comparison and measurement technique. However, the existing three-dimensional digital model comparison and measurement technique involves a large amount of manual intervention, which leads to a significant reduction in the efficiency of component inspection and assembly.

[0003] Therefore, given the problems of high manual intervention and low efficiency in the existing technical approach of 3D digital model comparison and measurement, designing a system component for the detection and analysis of complex assembled workpiece parts and implementing an efficient and accurate detection and analysis algorithm has become a problem that needs to be overcome in the automated production process of complex assembled workpieces. Summary of the Invention

[0004] The technical content of this invention is to provide a non-contact inspection method for complex assembly workpieces based on laser three-dimensional scanning, which realizes fully automatic point cloud data processing, facilitates efficient and accurate assembly between components of complex assembly workpieces, and avoids the problems of large amount of manual intervention and low efficiency in the existing three-dimensional digital model comparison and measurement technical route.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention provides a non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning. The specific method is as follows: S1. Constructing and analyzing a 3D scanning cloud computing benchmark for assemblies: Based on the position of the industrial robot's base and ground rail, the end effector of the industrial robot uses a wired structured light scanner to acquire 3D scanning frame point cloud data and performs real-time scanning frame point cloud computing.

[0006] S2. Establishment of the point cloud reference coordinate system: S2.1. The assembly is placed according to the reference in step S1, and the point cloud is projected onto the XOY, XOZ and YOZ planes.

[0007] S2.2. Take the center coordinates of each segment to construct the product centerline and calculate the angle of the centerline relative to the coordinate axis. Rotate the point cloud around the origin so that the centerline is parallel to a certain coordinate axis.

[0008] S2.3. The centerline of the rotated point cloud is parallel to one of the coordinate axes. Then, take the maximum and minimum average values ​​of the point cloud along the parallel axis and translate the point cloud along the parallel axis.

[0009] S2.4. Rotate and translate the three projection planes respectively, so that the central axis of the three plane projections of the point cloud is always parallel to one of the axes.

[0010] In particular, the roll angle of the point cloud is calculated during the rotation and translation of the YOZ plane.

[0011] S2.5. Repeat the above process until the rotation and translation of the point cloud no longer change.

[0012] S3. Point cloud self-alignment method: S3.1. Let the value of the point cloud's X-axis be located at... Inside, the value of the Y-axis y is located at Within, the Z-axis value z is located at Inside.

[0013] S3.2. Translate the point cloud, with the translation amounts for each axis as follows: , and The origin point will be moved to the center of the assembly.

[0014] S3.3. Take the minimum and maximum values ​​of Y along the X-axis in segments. Take the minimum and maximum values ​​of Z segment by segment along the X-axis. ,by , and A series of center point sets are constructed, namely, the set P of the cross-sectional point cloud of the assembly scan point cloud along the X-axis, where, for Any point within.

[0015] S3.4. Fit a spatial straight line to the set of center points P. Calculate the angle α between the projection onto the XOY plane and the X-axis, and the angle β between the projection onto the YOZ plane and the Y-axis, respectively.

[0016] Where y = k1x + b1 is the line formed by projecting the spatial line fitted by the set of center points P onto the XOY plane, k1 is the slope of line y, and b1 is the constant of the line equation. z = k2x + b2 is the line formed by projecting the spatial line fitted by the set of center points P onto the YOZ plane, k2 is the slope of line z, and b2 is the constant of the line equation.

[0017] S3.5. Rotate around the Z axis by an angle of -α, and around the Y axis by an angle of -β.

[0018] S3.6. Construct a straight line in the YOZ plane by taking the points corresponding to the maximum and minimum Z values, calculate the angle θ between the straight line and the Y-axis, and rotate the point cloud around the X-axis by -θ.

[0019] S4. Point cloud classification method: S4.1. Based on the relationship between multiple parts of the assembly, first extract the point cloud of the main component, and then divide the remaining no more than four parts into four quadrants independently, and extract the point cloud of the corresponding part for each quadrant.

[0020] S5. Point Cloud Classification and Extraction: S5.1. Using the X-axis direction of the main component as a reference, find its maximum and minimum values ​​on the X-axis, divide the point cloud into segments along the X-axis, construct the equation of the center point of the main component at the origin of the YOZ coordinate, calculate the points of the point cloud on the boundary line of the main component, if the number of points on the boundary line of the main component reaches a certain number, it is determined that the point cloud of the main component has been found, take out all the points located on the boundary line of the main component and use the least squares method to fit and obtain the corresponding equation.

[0021] S5.2. Determine the points located in the fitted equation, and set the distance threshold between points located on the boundary line of the main component according to the scanner accuracy parameters. Take all points located on the boundary line of the main component as the point cloud of the main component, and complete the point cloud classification of the main component.

[0022] S5.3. After completing the point cloud classification of the main components, analyze the gap rules of the mutual cooperation based on the specifications and parameters of the remaining components and their mutual cooperation methods, and complete the point cloud classification of the remaining components.

[0023] Preferably, when constructing and analyzing the 3D scanning cloud computing benchmark of the assembly: the origin of the industrial robot base coordinates coincides with the ground rail, the opposite direction of the Earth's gravity is the Z-axis direction, the direction of the robot's movement on the ground rail is the Y-axis direction, and according to the right-hand rule, the X-axis direction can be determined to be on the right side of the ground rail and the robot, forming the robot position K=(x, y, z) and the ground rail position T=(0, y, 0).

[0024] Preferably, when constructing and analyzing the 3D scanning cloud computing benchmark of the assembly: the coordinate origin is located at the head of the assembly at Kmm, and the X-axis points from the head of the assembly to its tail, where Kmm is the error range reserved when the assembly is installed and measured.

[0025] Preferably, when constructing and analyzing the 3D scanning cloud computing benchmark of the assembly: the attitude angle generated by the industrial robot's rotation relative to the X-axis is α, the attitude angle generated by its rotation relative to the Y-axis is β, and the attitude angle generated by its rotation relative to the Z-axis is γ, resulting in an attitude of M = (α, β, γ). The rotation matrix R can be calculated from M. Let the data scanned by the line structured light camera be Q = (x, 0, z), then the current scan frame point cloud computing is: P = R * Q + K + T t ...(Formula 1), through the calculation and analysis of the scanned frame point cloud P in Formula 1, the scanned point cloud and the robot are unified into a coordinate system, where R is the rotation matrix calculated based on the robot's posture, Q is the cross-sectional data currently scanned by the scanner, and T... tThis is the translation matrix for the orbital motion.

[0026] The preferred method for calculating the roll angle of the point cloud during rotation and translation in the YOZ plane is as follows: Take the projection of the YOZ plane and calculate the extreme value Z of the projection. max and Z min Calculate Z max and Z min Calculate the angle α of the line z = k3y + b3 in the YOZ plane formed by the points. y Point cloud rotated around the X-axis by -α y Where k3 is Z max and Z min Points form the slope of line z, y is the independent variable of line z, and b3 is the constant of the equation of line z.

[0027] Preferably, in the point cloud classification method: the main component is segmented into point clouds to obtain the point cloud parameters formed by the corresponding segments, and the coaxiality of the main component in the coordinate system is determined.

[0028] Preferably, in the point cloud classification method: the remaining parts that are not the main parts are classified according to the inherent maximum width parameter W of each part. max Take the Z coordinate as located at Points within the range are used as point clouds for quadrants I and III, and quadrants I and III are distinguished according to the sign of the Y coordinate.

[0029] Similarly, the point clouds of quadrants II and VI are distinguished by the combination of Y and Z coordinates, and the point clouds of the other components that are not the main components in each quadrant are also distinguished.

[0030] Compared with existing technologies, the beneficial effects of this invention are: This invention employs a combination of a line structured light scanner, an industrial robot, and a ground track to achieve automatic detection of the object being measured. By fusing and calculating the data obtained from the ground track, the industrial robot, and the line structured light scanner, dense, high-precision 3D scanning point cloud data is finally obtained. Furthermore, by designing point cloud alignment, point cloud classification, and geometric information parameterization methods, fully automated point cloud data processing is achieved. This facilitates efficient and precise assembly of complex workpiece components, avoiding the problems of high manual intervention and low efficiency associated with existing 3D digital model comparison and measurement techniques. Attached Figure Description

[0031] Figure 1 This is a schematic diagram showing the location arrangement of the three-dimensional scanning device in this invention.

[0032] Figure 2 This is a flowchart of the rotation, translation, and alignment process in this invention.

[0033] Figure 3 This is a flowchart of the point cloud roll calculation process in this invention.

[0034] Figure 4 This is a schematic diagram of the alignment of the reference in this invention. Figure 1 .

[0035] Figure 5 This is a schematic diagram of the alignment of the reference in this invention. Figure 2 .

[0036] Figure 6 This is a schematic diagram of the remaining point cloud after extracting component A in this invention.

[0037] Figure 7 This is a schematic diagram showing the area limitation of component E in this invention.

[0038] Figure 8 This is a schematic diagram of the XOZ projection of components B, C, and D in this invention.

[0039] Figure 9 This is a flowchart of the point cloud classification method for component A in this invention.

[0040] Figure 10 This is a flowchart illustrating the classification of components B, C, and D in this invention.

[0041] Figure 11 This is a schematic diagram of the overall structure of the assembly in this invention. Figure 1 .

[0042] Figure 12 This is a schematic diagram of the overall structure of the assembly in this invention. Figure 2 .

[0043] Figure 13 This is a schematic diagram of the skew angle detection in the present invention. Figure 1 .

[0044] Figure 14 This is a schematic diagram of the skew angle detection in the present invention. Figure 2 .

[0045] Figure 15 This is a schematic diagram of the installation angle detection in this invention.

[0046] Figure 16 This is a schematic diagram of the inversion angle detection in this invention.

[0047] Figure 17 This is a flowchart of the deflection angle calculation for component A in this invention.

[0048] Figure 18 This is a flowchart illustrating the calculation of the mounting angle of component B in this invention.

[0049] Figure 19 This is a flowchart of the anti-angle calculation for component C in this invention. Detailed Implementation

[0050] 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.

[0051] Example 1: The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning designed in this invention uses a laser three-dimensional scanning system based on line structured light to collect workpiece information and process the measurement information. The data processing mainly includes point cloud coordinate establishment, point cloud alignment, point cloud classification, and feature parameter extraction.

[0052] Point cloud coordinate establishment refers to the construction of an initial point cloud coordinate system based on the reference positions of the industrial robot base, ground rail, and end-effector structured light scanner.

[0053] Point cloud alignment refers to the process of aligning the scanned point cloud with the defined reference coordinate system when a line structured light scanner scans and models an assembly. Since the orientation of the assembly is random, the scanning point cloud needs to be aligned with the defined reference coordinate system during the measurement process.

[0054] Point cloud classification is used to establish relationships between scattered point clouds and components.

[0055] Feature parameter extraction: The detection parameters of the assembly include skew angle, mounting angle, and reverse angle.

[0056] When collecting workpiece information, a combination of a line structured light scanner, an industrial robot, and a ground track is used to achieve automatic detection of the object being measured. The line structured light scanner enables rapid cross-sectional detection of the workpiece, the industrial robot uses its end effector to hold the line structured light scanner for multi-angle scanning and transmits pose data in real time, and the ground track carries the industrial robot to detect axial movement. By fusing and calculating the data obtained from the ground track, industrial robot, and line structured light scanner, a dense, high-precision 3D scan point cloud data is finally obtained.

[0057] Line structured light refers to the measurement of depth information by projecting a linear laser beam onto the surface of an object under test, employing the principle of triangulation. The linear laser beam emitted by the laser must have distinct features compared to other objects in the scene to ensure the detector can easily capture the corresponding image. After collimation and focusing, the line laser beam is incident perpendicularly onto the surface of the object under test. The light from the object's surface undergoes diffuse reflection and is imaged onto the detector's photosensitive array by a special receiving lens. When the surface of the object is moved, the image point reflected on the detector's photosensitive array also shifts. Line laser scanning technology offers a very large acquisition range due to the focusing effect of the lens, resulting in fast scanning speed, good repeatability, high accuracy, and small error.

[0058] Example 2: The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning designed in this invention adopts the following implementation technology: I. Cloud Computing Benchmark for 3D Scanning of Assemblies: A schematic diagram of the system implementing 3D scanning is shown below. Figure 1 As shown, the origin of the industrial robot's base coordinates coincides with the ground rail, and the opposite direction of Earth's gravity is the z-direction; the direction in which the robot moves on the ground rail is the y-direction; according to the right-hand rule, the x-axis direction can be determined to be to the right of the ground rail and the robot. The industrial robot's attitude angle α is the angle generated by the rotation of the X-axis, β is the angle generated by the rotation of the Y-axis, and γ is the angle generated by the rotation of the Z-axis.

[0059] Let the current position of the robot be K = (x, y, z), and its pose be M = (α, β, γ). The rotation matrix R can be calculated from M. The data scanned by the line structured light camera is Q = (x, 0, z), and the current position of the ground track is T = (0, y, 0). Then, the current scan frame point cloud computing is: The current scan frame point cloud computing is: P = R * Q + K + T t ... (Formula 1), through the calculation and analysis of the scanned frame point cloud P in Formula 1, the scanned point cloud and the robot are unified under a coordinate system.

[0060] Where R is the rotation matrix calculated based on the robot's posture, Q is the cross-sectional data currently scanned by the scanner, and T... t Given the translation matrix of the ground track motion, the transformation relationship between the (x, z) coordinates of the line structured light scanner and the pose coordinates (x, y, z, α, β, γ) of the six degrees of freedom of the industrial robot is obtained through Formula 1.

[0061] II. Establishment of the point cloud reference coordinate system: Point cloud reference coordinates address the issue of randomness in the placement and orientation of assemblies on a horizontal workbench, thus eliminating inspection errors caused by placement and orientation. The assembly is placed as described in the previous chapter. The point cloud projections on the XOY, XOZ, and YOZ planes are obtained. The projected point cloud is segmented, and the center coordinates of each segment are calculated. The center coordinates of each segment are used to construct the product's centerline, and the angle of the centerline relative to the coordinate axes is calculated. The point cloud is rotated around the origin until the centerline is parallel to one of the coordinate axes. After rotation, the centerline of the point cloud is parallel to one axis. Then, the maximum and minimum average values ​​of the point cloud along the parallel axis are taken, and the point cloud is translated along the parallel axis. Rotation and translation are performed on the three projection planes respectively, ensuring that the centerline of the point cloud's projections on the three planes is always parallel to one axis. This process is repeated until the rotation and translation of the point cloud no longer change. The point cloud rotation and translation alignment is achieved as follows: Figure 2 The process is shown below.

[0062] During rotation and translation, the roll angle of the assembly affects the calculation of the rotation angle and translation distance in the YOZ plane. Therefore, the roll angle calculation of the point cloud is added during the rotation and translation process in the YOZ plane. The calculation method is as follows: Figure 3 As shown.

[0063] Point cloud data processing methods: (i) Point Cloud Self-Alignment Method ① Suppose the value of the X-axis of the point cloud is located at... Inside, the Y-axis value is located at Within, the Z-axis value is located at Inside; ② Translate the point cloud, with the translation amounts for each axis being as follows: , and The origin point will be moved to the center of the assembly. ③ Take the minimum and maximum values ​​of Y along the X-axis in segments. The minimum and maximum values ​​of Z , with x c , and Construct a series of center point sets, namely, the set P of the cross-sectional point cloud of the assembly scan point cloud along the X-axis, where x c for Any point within; ④ Fit a spatial straight line to the set of center points P. Calculate the angle α between the projection onto the XOY plane and the X-axis, and the angle β between the projection onto the YOZ plane and the Y-axis, respectively. ⑤ Rotate around the Z-axis by an angle of -α, and around the Y-axis by an angle of -β; ⑥ Construct a straight line in the YOZ plane by taking the points corresponding to the maximum and minimum Z values, calculate the angle θ between the straight line and the Y-axis, and rotate the point cloud around the X-axis by -θ.

[0064] By repeating the above steps, the rotation angle and translation distance are both less than the threshold. The origin of the point cloud coordinates is located at the center of the missile, and both the top and side views are symmetrical about the X-axis. See details. Figure 4 , Figure 5 .

[0065] Point cloud classification methods: The purpose of point cloud classification is to establish relationships between scattered point clouds and components, that is, to clarify which component each point originates from. The assembly mainly consists of five parts: component A, component B, component C, component D, and component E. Component A is the main body of the product, while components B, C, D, and E are distributed symmetrically across the four quadrants of the coordinate system. When classifying components, component A is extracted first. The remaining components B, C, D, and E are then divided into four quadrants. After extracting component E, specific component extraction is performed within each quadrant.

[0066] Part A has a conical front end and a cylindrical rear end. During classification, the projection of the assembly's segments on the X-axis onto the Y0Z plane is used. If the point cloud cross-sectional data forms a closed ring, the circle containing the segmented point cloud is fitted to obtain the center and radius of the cross-sectional circle of part A. The specific operation for point cloud classification of part A is divided into the point cloud of the front segment and the point cloud of the rear segment. Since the point cloud data of the front segment is relatively simple, it can be directly extracted to obtain the straight line formed by the center of the cross-sectional circle of the front segment point cloud and its maximum radius R. max The point cloud in the later section contains point cloud data of other components; the x-axis of the current section is taken. i Substituting the lines formed by the centers of the circles, we obtain the Y and Z coordinates (y = 0). ci , z ci The maximum radius R obtained from the previous segment max Using the radius as the coordinates, the points within the circle are calculated as the point cloud of the rear segment of component A. The equation of the projectile's central axis is obtained by calculating the coordinates of all circle centers. The straight line z=k, projected onto the XOZ plane from the central axis, is then taken. z x+b z The line y=k projected onto the XOY plane x x+b x This can be used to determine the coaxiality of component A.

[0067] The remaining point cloud after completing the classification of component A is as follows: Figure 6 As shown, the maximum width of components B, C, and D can be known from the design of the assembly. There is a certain gap between components B, C, and D and component E, which can distinguish components B, C, and D in each quadrant.

[0068] Let W be the maximum width parameter of components B, C, and D. max Take the Z coordinate as located at Points within the specified range are used as point clouds for quadrants I and III, and quadrants I and III are distinguished according to the sign of the Y-coordinate. Similarly, the point clouds for quadrants II and VI are distinguished by a combination of Y and Z coordinates. After distinguishing components B, C, and D in each quadrant, the remaining point clouds are component E and other unknown point clouds.

[0069] Construct a line z = ky + b at a 45° angle to the horizontal axis within the quadrant containing the remaining point cloud. Let b be an undetermined coefficient, and its initial value be... , where is the distance from component E to the center of the assembly. Calculate the points located on the straight line and determine if points exist within a certain distance outwards. If more point clouds exist, move the straight line outwards at a 45° angle until the outermost edge of component E is found. Calculate the slope of the straight line formed by the point clouds at the outermost edge, construct a straight line perpendicular to the outer edge line, calculate the width of the point cloud on the outer edge line, compare it with the standard parameters of component E, and after limiting the width of the point cloud of component E, classify the point cloud of component E. The quadrant region formed by component E is limited as follows: Figure 7 As shown.

[0070] The remaining point cloud, after separating the quadrants, contains components B, C, and D. Each component is rotated to the desired position. Figure 8 The XOZ projection of the point cloud is taken from the pose shown (quadrant I rotated 90° counterclockwise, quadrant III rotated 90° clockwise, quadrant VI rotated 180°).

[0071] Components C and D are mounted on top of component B, with a gap between them. The point cloud is layered along the Z-axis, and the continuity of the point cloud is determined to classify the point cloud of the three components.

[0072] V. Point Cloud Classification and Extraction: The product comprises five parts: Part A, Part B, Part C, Part D, and Part E. Part A is the main body of the product. Parts B, C, and D are distributed symmetrically across the four quadrants. Part E, after point cloud rotation and translation, may appear in any quadrant. The point cloud classification method for Part A is as follows: Figure 9 As shown. Find the maximum and minimum values ​​along the X-axis, segment the point cloud along the X-axis, construct the equation of a circle centered at the origin of the YOZ coordinate system, and calculate the points located on the arcs. If a certain number of points are located on the arcs, the point cloud of component A is considered found. Extract all points located on the arcs and fit the circle equation using the least squares method. Determine the points located on the fitted circle, setting a distance threshold based on the scanner's accuracy parameters. Use all points located on the arcs as the point cloud of component A, completing the point cloud classification for component A, as detailed below. Figure 9 As shown.

[0073] The remaining point cloud after completing the classification of component A is as follows: Figure 8 As shown, the maximum width of components B, C, and D can be known from the product design. There is a certain gap between components B, C, and D and component E, which can distinguish components B, C, and D in each quadrant.

[0074] Components C and D are mounted on top of component B, with a gap between them. The point cloud is layered along the Z-axis, and the continuity of the point cloud is determined to classify the point clouds of the three components. The specific process is as follows: Figure 10 As shown. Take the maximum height Z of component B. B =HB +R,H B Let Z be the relative height of component B, and R be the maximum radius of component A. At height Z... B If multiple segments appear below, the current segment's point cloud is considered abnormal (theoretically, component B is closely attached to component A and is a continuous structure, therefore in Z...). B (Multiple segments will not appear below), otherwise it will be identified as component B.

[0075] The gap between components C and D, exceeding the height limit, causes the point cloud to be divided into two intervals along the X-axis: the front segment is component C, and the rear segment is component D. Therefore, the number of intervals along the X-axis obtained from the classification process of components B, C, and D should be B—BC—B—BD—B; otherwise, the point cloud data will be abnormal. At this point, the point cloud classification of all five components of the assembly is complete. Components A and E are single components, while components B, C, and D are divided into four quadrants.

[0076] Example 3: The present invention also relates to the analysis of assembly inspection parameters, the details of which are as follows: The large-size assembly being tested in this invention undergoes only external shape inspection after final assembly. An assembly coordinate system is established, with the origin located 10mm from the assembly's head (a pre-defined error range). The X-axis points from the assembly's head to its tail, the Y-axis points in the opposite direction of Earth's gravity, and the Z-axis direction is determined using the right-hand rule. Measurement point locations are as follows... Figure 11 , Figure 12 As shown, the measurement points are for ease of measurement. Measurement points for the head of the assembly should be at the front end, and for the rest at the rear end. According to the QJ 2079-91 measurement standard, the inspection parameters for the assembly include the inspection of the skew angle, mounting angle, and reversal angle.

[0077] The skew angle is a measurement parameter used to measure the degree of bending of the main body of an assembly; the skew angle η is... N The angle formed by the main body central axis Y=kx+b (or Z=kx+b) fitted by the head and tail measurement points of the assembly and the XOY plane (or XOZ plane) is expressed by the formula: ,in, This represents the height of the head measurement point from the horizontal reference platform. This represents the height of the tail measurement point from the horizontal reference platform. This is the distance between the head measurement point and the tail measurement point. For example... Figure 13 , Figure 14 The diagram shown is a schematic of the skew angle detection.

[0078] like Figure 15 As shown, the mounting angle refers to the angular deviation of an assembly component from the assembly body. Expressed as a formula , where HW1 and H W2 L represents the vertical height of two measuring points at the connection between the component and the main body of the assembly, from the horizontal reference plane. W12 This represents the distance between the two measurement points.

[0079] like Figure 16 As shown, the inversion angle test determines whether the assembly components are installed perpendicularly to the main body. Specifically, it measures the angle formed in the Y0Z plane by the straight lines connecting the assembly component to the main body and the end of the assembly component. This angle is Ψw, and can be expressed by the formula. , where H C1 and H C2 L represents the height of two measuring points of an assembly component from the horizontal reference platform. C12 This represents the distance between the two measurement points.

[0080] Example 4: This invention involves some basic point cloud processing methods. A brief overview of some of these basic methods is provided to facilitate understanding of the subsequent invention. The specific details are as follows: (1) Least squares method The least squares method is a mathematical optimization technique that finds the optimal matching function by minimizing the squared error. It can be used to conveniently obtain some unknown data. In curve fitting, let the function... , Let f(x) be a set of linearly independent functions, and f(x) be the coefficients to be determined. The criterion for least squares fitting a function is to minimize the sum of squares of the distances between the fitted functions f(x). Least squares fitting a curve is essentially the process of finding the coefficients to be determined using the least squares method, given the coordinates of scattered points and the form of the curve function.

[0081] (2) Scatter point fitting of straight line If f(x) = kx + b, that is, to fit the equation of the line using the coordinates of scattered points located on the line, the undetermined coefficients to be determined are [k, b]. Let the coordinates of all scattered points be [x, b]. i y i The criterion of least squares is transformed into... To minimize this, we can take the partial derivatives of the objective function with respect to k and b respectively. .

[0082] Setting the partial derivatives to 0, solving the system of equations yields the undetermined coefficients [k, b]: .

[0083] By substituting the coordinates of all the scattered points, the values ​​of the undetermined coefficients can be obtained.

[0084] (3) Scatter point fitting circle Similar to fitting a straight line to a scatter plot, a target function is set. This can be achieved by taking the partial derivative and setting it to zero (note that this is done directly on the partial derivative). Finding the partial derivative does not yield an analytical solution; it is necessary to find the partial derivative of f(x) to obtain the undetermined coefficients of the circle equation, where (x0, y0) is the center of the fitted circle and R is the radius of the fitted circle.

[0085] (4) Calculation of detection parameters The calculation of some parameters is based on the design of a three-dimensional point cloud computing method to achieve equivalence according to existing measurement methods.

[0086] The original manual measurement method assessed the coaxiality of component A by measuring the height of four measurement points from multiple cross-sections on component A to the measuring platform. Since each cross-section of component A is a standard circle, measuring four measurement points with a 180° phase difference is equivalent to determining the positional deviation of the cross-section's center. Now, fully automated 3D scanning achieves... Figure 17 As shown.

[0087] After the measurement and calculation of component A, the installation angles of the remaining components B, C, and D all need to be calculated, such as... Figure 18 As shown, taking the installation angle calculation of component B in the first quadrant as an example, the installation angles of the other components are calculated using the same technique.

[0088] Reverse angle measurement mainly checks whether the component is installed perpendicular to component A. The assembly is divided into four quadrants, and the reverse angle is calculated for different components in each quadrant. The calculation of reverse angles for different components is generally the same. Taking the reverse angle measurement flowchart of component C as an example, see [link to flowchart]. Figure 19 .

[0089] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning, characterized in that: S1. Constructing and analyzing a 3D scanning cloud computing benchmark for assemblies: Based on the position of the industrial robot base and ground rail, the end effector of the industrial robot uses a wire-held structured light scanner to acquire 3D scanning frame point cloud data and perform real-time scanning frame point cloud computing. S2. Establishment of the point cloud reference coordinate system: S2.

1. The assembly is placed according to the reference in step S1, and the point cloud is projected onto the XOY, XOZ and YOZ planes; S2.

2. Take the center coordinates of each segment to construct the product centerline and calculate the angle of the centerline relative to the coordinate axis. Rotate the point cloud around the origin so that the centerline is parallel to a certain coordinate axis. S2.

3. The centerline of the rotated point cloud is parallel to one of the coordinate axes. Then, take the maximum and minimum average values ​​of the point cloud in the direction of the parallel axis and translate the point cloud along the parallel axis. S2.

4. Rotate and translate the three projection planes respectively, so that the central axis of the three plane projections of the point cloud is always parallel to one of the axes; Among them, the roll angle of the point cloud is calculated during the rotation and translation of the YOZ plane; S2.

5. Repeat the above process until the rotation and translation of the point cloud no longer change; S3. Point cloud self-alignment method: S3.

1. Let the value of the point cloud's X-axis be located at... Inside, the value of the Y-axis y is located at Within, the Z-axis value z is located at Inside; S3.

2. Translate the point cloud, with the translation amounts for each axis as follows: , and The origin point will be moved to the center of the assembly. S3.

3. Take the minimum and maximum values ​​of Y along the X-axis in segments. Take the minimum and maximum values ​​of Z segment by segment along the X-axis. ,by , and A series of center point sets are constructed, namely, the set P of the cross-sectional point cloud of the assembly scan point cloud along the X-axis, where, for Any point within; S3.

4. Fit a spatial straight line to the set of center points P. Calculate the angle α between the projection onto the XOY plane and the X-axis, and the angle β between the projection onto the YOZ plane and the Y-axis, respectively. Where y=k1x+b1 is the line formed by the projection of the spatial line fitted by the set of center points P onto the XOY plane, k1 is the slope of the line y, and b1 is the constant of the line equation. z = k2x + b2 is the line formed by the projection of the spatial line fitted by the set of center points P onto the YOZ plane, k2 is the slope of the line z, and b2 is the constant of the line equation. S3.

5. Rotate around the Z-axis by an angle of -α, and around the Y-axis by an angle of -β; S3.

6. Construct a straight line in the YOZ plane by taking the points corresponding to the maximum and minimum Z, calculate the angle θ between the straight line and the Y axis, and rotate the point cloud around the X axis by -θ. S4. Point cloud classification method: S4.

1. Based on the relationship between multiple parts of the assembly, first extract the point cloud of the main component, and then divide the remaining no more than four parts into four quadrants independently, and extract the point cloud of the corresponding part in each quadrant. S5. Point Cloud Classification and Extraction: S5.

1. Using the X-axis direction of the main component as a reference, find its maximum and minimum values ​​on the X-axis, divide the point cloud into segments along the X-axis, construct the equation of the center point of the main component at the origin of the YOZ coordinate, calculate the points of the point cloud on the boundary line of the main component, if the number of points on the boundary line of the main component reaches a certain number, it is determined that the point cloud of the main component has been found, take out all the points located on the boundary line of the main component and use the least squares method to fit and obtain the corresponding equation; S5.

2. Determine the points located in the fitted equation, and set the distance threshold between points located on the boundary line of the main component according to the scanner accuracy parameters. Take all points located on the boundary line of the main component as the point cloud of the main component, and complete the point cloud classification of the main component. S5.

3. After completing the point cloud classification of the main components, analyze the gap rules of the mutual cooperation based on the specifications and parameters of the remaining components and their mutual cooperation methods, and complete the point cloud classification of the remaining components.

2. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 1, characterized in that, When constructing and analyzing a 3D scanning cloud computing benchmark for an assembly: The origin of the industrial robot's base coordinates coincides with the ground rail. The direction opposite to Earth's gravity is the Z-axis, and the direction the robot moves on the ground rail is the Y-axis. According to the right-hand rule, the X-axis is located to the right of both the ground rail and the robot, resulting in the robot's position being K = (x, y, z) and the ground rail's position being T = (0, y, 0).

3. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 2, characterized in that, When constructing and analyzing a 3D scanning cloud computing benchmark for an assembly: The origin of the coordinate system is located at Kmm at the head of the assembly, and the X-axis points from the head of the assembly to its tail. Kmm is the error range reserved when the assembly is installed and measured.

4. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 2, characterized in that, When constructing and analyzing a 3D scanning cloud computing benchmark for an assembly: The industrial robot's attitude angle generated by rotating relative to the X-axis is α, its attitude angle generated relative to the Y-axis is β, and its attitude angle generated relative to the Z-axis is γ, resulting in an attitude M = (α, β, γ). The rotation matrix R can be calculated from M. Let the data scanned by the line structured light camera be Q = (x, 0, z), then the current scan frame point cloud computing is: P = R * Q + K + T t ... (Formula 1); By calculating and analyzing the point cloud P of the scanned frame in Formula 1, the scanned point cloud and the robot are unified into a coordinate system. Where R is the rotation matrix calculated based on the robot's posture, Q is the cross-sectional data currently scanned by the scanner, and T... t This is the translation matrix for the orbital motion.

5. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 1, characterized in that, The method for calculating the roll angle of a point cloud during rotation and translation in the YOZ plane is as follows: Project the image onto the YOZ plane and calculate the extreme value Z. max and Z min Calculate Z max and Z min Calculate the angle α of the line z = k³y + b³ in the YOZ plane formed by the points. y Point cloud rotated around the X-axis by -α y ; Where k3 is Z max and Z min Points form the slope of line z, y is the independent variable of line z, and b3 is the constant of the equation of line z.

6. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 1, characterized in that, In point cloud classification methods: The main component is segmented into point clouds to obtain the point cloud parameters formed by the corresponding segments, and the coaxiality of the main component in the coordinate system is determined.

7. The non-contact inspection method for complex assembled workpieces based on laser three-dimensional scanning according to claim 1, characterized in that, In point cloud classification methods: The remaining components, which are not the main components, are determined according to their inherent maximum width parameter W. max Take the Z coordinate as located at Points within the range are used as point clouds for quadrants I and III, and quadrants I and III are distinguished according to the sign of the Y coordinate; Similarly, the point clouds of quadrants II and VI are distinguished by the combination of Y and Z coordinates, and the point clouds of the other components that are not the main components in each quadrant are also distinguished.

Images