A method and device for wafer processing path planning based on paraxial machine vision

Abstract

The present invention discloses a method and device for wafer processing path planning based on off-axis machine vision, aiming to solve the distortion problem caused by the non-coaxiality of the off-axis CCD and the galvanometer field of view and the difficult problem of direction control for anisotropic processing of wafers. It uses a two-dimensional galvanometer in combination with an XYZ three-axis machine tool and an off-axis CCD camera assembly, and combines a complete visual guidance process (mechanical calibration → image correction → path planning) to achieve automatic planning of the processing path on the wafer surface, reducing manual intervention. The wafer processing path generated by this solution has the characteristics of parameter setting, high precision, low cost, flexibility and controllability, and the processing path and the positioning edge of the wafer have a specific direction, so as to achieve the effect of accurately processing a specific path on the wafer surface.

Description

Technical Field

[0001] The present invention relates to the technical field of wafer processing, and particularly relates to a method and device for wafer processing path planning based on off-axis machine vision. Background Art

[0002] The production process of wafers determines that the outer contour of wafers is often a standard circle. During the processing of wafers such as cutting and polishing, the planning of the processing path has a certain impact on the processing effect. Wafers are crystals with different characteristics in different directions. Therefore, wafer blank manufacturers will leave an identification line on the outer contour of the wafer. This characteristic identification line has vector characteristics and is used to indicate the crystal orientation information of the wafer. Different processing directions based on this vector will result in different effects.

[0003] In a galvanometer processing system using vision guidance, there are two main vision guidance schemes relative to the position of the galvanometer: coaxial and off-axis CCD. In the coaxial CCD scheme, since the CCD camera and the galvanometer share the same optical axis, the processing field of view of the galvanometer and the field of view of the CCD camera are coaxial, and image correction is relatively simple. However, the optical path structure is relatively complex, and due to the compact structure of the optical path, it is difficult to coincide the depth of field of the CCD camera with the processing focal plane of the galvanometer. The optical path cost is high, and the field of view of the CCD camera is often very small, making it difficult to plan the processing path of larger workpieces. In the off-axis CCD scheme, since the optical paths of the CCD camera and the galvanometer are not coaxial, it can be placed outside the galvanometer, and the field of view size and eyepiece size of the CCD camera can be arbitrarily changed to obtain a suitable field of view size and depth of field size. However, the center of the off-axis CCD field of view and the center of the galvanometer are not coaxial, so there is distortion between the two fields of view. When using the CCD camera field of view as the galvanometer field of view for processing, how to accurately convert the CCD field of view into the galvanometer field of view is particularly crucial. Summary of the Invention

[0004] In order to solve the problems existing in the above background art, the present invention provides a method and device for wafer processing path planning based on off-axis machine vision, which uses an off-axis CCD camera, a two-dimensional galvanometer, and an XYZ three-axis machine tool to automatically plan the processing path on the wafer surface. The wafer processing path generated by this scheme has the characteristic that its parameters can be set, and the processing path and the positioning edge of the wafer have a specific direction, so as to achieve a specific path processing effect on the wafer surface, and has the advantages of low cost and small implementation difficulty.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In the first aspect of the present invention, there is provided a method for wafer processing path planning based on off-axis machine vision, including the following steps:

[0007] S1. Calibrate and synchronously control the XYZ moving axes of the machine tool. The galvanometer and the CCD camera assembly located on the side move up and down along the Z axis, and the wafer placed on the tray moves along the XY axis directions.

[0008] S2. After preliminarily calibrating the galvanometer, control the galvanometer to strike a matrix dot pattern on the calibration plate according to the theoretical dot matrix coordinates. Use magnifying devices such as microscopes to determine the coordinates of each point on the calibration plate, compare the deviation between the actual coordinates and the theoretical coordinates of the dot pattern marking points struck by the galvanometer on the calibration plate, compensate for the dynamic focusing of the galvanometer, and correct the Z-axis error of the two-dimensional galvanometer at each point position.

[0009] S3. Replace the appropriate camera eyepiece to adjust the imaging position and depth of field of the CCD camera assembly, so that the imaging plane of the CCD camera assembly is on the surface of the wafer to be processed, and at the same time adjust the camera depth of field to be 1 - 3 mm.

[0010] S4. Determine the offset vector value in the machine tool coordinate system between the vision center Oc of the CCD camera assembly and the processing vision center Og of the galvanometer;

[0011] S5. Determine the processing size ratio between the image in the field of view of the CCD camera assembly and the processing field of view of the galvanometer k ;

[0012] S6. Use the image detection method to obtain the center P(x0, y0) of the wafer and the two end points M(x1, y1) and N(x2, y2) of the feature edge in the field of view of the CCD camera assembly. According to the sign of, judge whether the feature vector is or ;

[0013] S7. The wafer processing path consists of several groups of parallel straight lines. Different processing effects are achieved by setting the distance between the parallel straight lines and the angle formed between the parallel straight lines and the feature vector. First, draw a straight line GH through the center P of the wafer, intersecting the outer contour of the wafer at points G and H. GH is the perpendicular bisector of MN, and obtain the pixel coordinates of points G and H. Discretize the line segment GH into a point set {Sn}, and the distance between Sn and Sn - 1 is d. Then, draw a straight line RQ through point Sn on GH and forming an angle with the feature vector, intersecting the outer contour of the wafer at points Rn and Qn. Then the straight line set {RnQn} is the automatically planned processing path. Finally, {RnQn} is multiplied by the conversion ratio k in step S5 to obtain the processing path set in the galvanometer coordinate system. Finally, after the processing head moves along the offset vector obtained in step S4, it can move according to the processing path k{RnQn}.

[0014] Adopt the above technical solution:

[0015] (1) In this application, a design of separating the off-axis CCD from the galvanometer is adopted, which supports flexible replacement of the eyepiece of the CCD camera component to adapt to different field of view and depth of field requirements. By combining with the LED ring light source, the imaging quality is optimized, the optical path limitation of the coaxial scheme is broken through, the system complexity and cost are reduced, and at the same time, the size adaptation range of the workpiece is expanded;

[0016] (2) Using the kinematic model of the XYZ three-axis machine tool, the orthogonality error is calibrated by combining with a laser interferometer or a coordinate measuring machine (CMM), and corrected through a compensation algorithm; at the same time, the galvanometer marking dot matrix is measured multiple times to generate a compensation file. Through dynamic calibration and error compensation, the motion accuracy of the multi-axis system is improved, and the influence of mechanical installation error on the machining path is reduced;

[0017] (3) The off-axis CCD image is corrected twice through the perspective transformation matrix H to eliminate the field of view tilt distortion; combined with the offset vector and the machining size scale factor k , an accurate mapping between the CCD pixel coordinate system and the galvanometer machining coordinate system is realized, the distortion problem caused by the non-coaxial field of view between the off-axis CCD and the galvanometer is solved, and a complete coordinate conversion chain (image correction → offset compensation → scale conversion) is proposed to ensure the geometric accuracy of the machining path;

[0018] (4) By using the image detection method, the center and outer contour of the wafer are identified, and the direction of the feature vector is judged. Based on the direction of the feature vector, a set of machining paths {RnQn} parallel to the included angle formed with the feature vector is generated, and combined with the machining size scale factor k to be converted into the path in the galvanometer coordinate system. The crystal orientation information (feature vector) is directly associated with the machining path direction, realizing the directional control of the anisotropic machining effect on the wafer surface.

[0019] This application provides a method based on off-axis CCD machine vision recognition, which uses a two-dimensional galvanometer with an XYZ three-axis machine tool, and combines with a complete vision guidance process (calibration → correction → path generation) to realize the automatic planning of the machining path on the wafer surface, reduce manual intervention. The wafer machining path generated by this scheme has the characteristic of parameter setting, and the machining path and the positioning edge of the wafer have a specific direction, so as to realize the precise machining effect of a specific path on the wafer surface.

[0020] Further, the specific process of replacing a suitable camera eyepiece to adjust the imaging position and depth of field of the CCD camera component is as follows: Use the galvanometer to project a pattern on the calibration plate, move the XY axis so that the pattern appears in the field of view of the off-axis CCD camera component, and replace the suitable camera eyepiece until the CCD camera component clearly observes the image of the galvanometer machining plane, that is, the depth of field of the camera covers the focus position of the galvanometer machining.

[0021] Further, in the step S3, the specific process of adjusting the imaging plane of the CCD camera assembly on the surface of the wafer to be processed (parallel to the XY plane) is as follows: According to four marking points located on the tray and respectively at the four vertex positions of a square in the machine tool coordinate system, observe the pixel coordinates of the four marking points in the image captured by the paraxial CCD camera assembly, and perform tilt correction on the CCD camera assembly and / or perform secondary correction on the pixel coordinates in the CCD image using an algorithm until the pixel coordinates of the four marking points in the CCD image are also the four vertices of a square.

[0022] Further, adjust the mechanical clamping tool of the CCD camera assembly for tilt correction.

[0023] Further, the specific process of performing secondary correction on the pixel coordinates in the CCD image using an algorithm is as follows: First, obtain the transformation matrix H of the pixel coordinates of the four vertices in the CCD image under the tilted viewing angle before correction to the CCD image coordinates of the vertical downward field of view along the Z axis, and then use the transformation matrix H to perform secondary correction on the pixel coordinates of the four vertices in the CCD image to obtain an imaging picture parallel to the tray plane.

[0024] Further, the four marking points on the tray are respectively painted with different colors.

[0025] Further, in the step S6, >0, then the characteristic quantity is ;

[0026] <0, then the characteristic quantity is .

[0027] Further, in the step S4, first place the calibration plate under the field of view of the galvanometer, adjust the Z-axis distance so that the processing focal plane of the galvanometer falls on the calibration plate. At this time, the positions of the XY axes of the galvanometer processing center are Og(x1, y1). Then process a calibration pattern (such as a point) on the calibration plate. After completion, keep the Z-axis position unchanged and move the XY axes so that the calibration pattern is located at the center of the CCD camera's field of view. Record the positions of the XY axes at this time as Oc(x2, y2). At this time, a set of offset vectors =(x2 - x1, y2 - y1) is obtained. Repeat the above method to measure the offset vector value multiple times, and use its average value to represent the offset vector value. Its meaning is the offset in the machine tool coordinate system from the center point of the galvanometer processing field of view to the center point of the CCD camera assembly's field of view.

[0028] Further, in step S2, move the Z-axis up and down so that the distance between the center point of the galvanometer optical axis and the tray is the focal length d of the galvanometer. Record the position of the galvanometer Z-axis at this time. Use the galvanometer to punch an n*n matrix target (such as 9*9 or 25*25) on the calibration plate placed on the tray. Observe the dot matrix punched by the galvanometer with a microscope, compare the deviation between the actual coordinates and the theoretical coordinates of the dot matrix marking points, generate a compensation file, and correct the two-dimensional galvanometer.

[0029] In the second aspect of the present invention, a wafer processing path planning device based on off-axis machine vision is provided for implementing the above-mentioned wafer processing path planning method. The device includes an XYZ three-axis machine tool, a tray, a galvanometer, and a CCD camera assembly. The tray is installed on the XY orthogonal motion axis, the Z-axis motion axis is fixed in the middle of the gantry, the galvanometer is installed at the output end of the Z-axis motion axis, a CCD camera assembly is fixedly installed on the side of the galvanometer, the field of view of the CCD camera assembly is downward along the Z-axis, and an LED lighting lamp ring is installed on the outer side under the galvanometer.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) In this application, a design of separating the off-axis CCD and the galvanometer is adopted, which simplifies the optical path structure, supports flexible replacement of the eyepiece of the CCD camera assembly to adapt to different field of view and depth of field requirements, combines an LED ring light source to optimize the imaging quality, adapts to the imaging requirements of different wafer sizes and surface materials, breaks through the optical path limitations of the coaxial scheme, reduces the system complexity and cost, and at the same time expands the size adaptation range of the workpiece to be processed;

[0032] (2) Utilize the kinematic model of the XYZ three-axis machine tool, combine a laser interferometer or a coordinate measuring machine (CMM) to calibrate the orthogonality error, and correct it through a compensation algorithm; at the same time, perform multiple measurements on the dot matrix punched by the galvanometer to generate a compensation file. Through dynamic calibration and error compensation, improve the motion accuracy of the multi-axis system, reduce the influence of mechanical installation error accumulation on the processing path, and improve the coincidence degree between the processing focal plane of the galvanometer and the wafer surface;

[0033] (3) Perform secondary correction on the off-axis CCD image through the perspective transformation matrix H to eliminate the field of view tilt distortion; combine the offset vector and the processing size scale factor k to achieve an accurate mapping between the CCD pixel coordinate system and the galvanometer processing coordinate system, ensure the geometric mapping accuracy of the processing path, solve the distortion problem caused by the non-coaxiality of the off-axis CCD and the galvanometer field of view, and propose a complete coordinate conversion chain (image correction → offset compensation → scale conversion) to ensure the geometric accuracy and stability of the processing path;

[0034] (4) Identify the center and outer contour of the wafer through image detection methods, and determine the direction of the feature vector. Based on the direction of the feature vector, generate the angle formed with the feature vector Parallel processing path set {RnQn}, combined with the processing size ratio coefficient k Convert to the path in the galvanometer coordinate system. Directly associate the crystal orientation information (characteristic vector) with the processing path direction to achieve directional control of the anisotropic processing effect on the wafer surface and meet the requirements of the anisotropic characteristics of the wafer on the processing effect (such as directional optimization of cutting, polishing and other processes).

[0035] The present application provides a method for machine vision recognition based on paraxial CCD, aiming to solve the distortion problem caused by the non-coaxial field of view of the paraxial CCD and the galvanometer and the direction control problem of anisotropic wafer processing. The method uses a two-dimensional galvanometer with an XYZ three-axis machine tool and a paraxial CCD camera assembly, and combines the completed visual guidance process (mechanical calibration → image correction → path planning) to realize automatic planning of the wafer surface processing path and reduce manual intervention. The wafer processing path generated by the scheme has the characteristics of settable parameters, high precision, low cost, flexible and controllable, and the processing path and the positioning edge of the wafer have a specific direction, thereby achieving the effect of precise processing of a specific path on the wafer surface. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] Figure 1 It is a schematic diagram of a wafer in the present invention;

[0038] Figure 2 It is a structural schematic diagram of the wafer processing path planning device in the present invention;

[0039] Figure 3 The image of the pallet marking points in the field of view of the front and rear CCD camera components in the present invention is corrected;

[0040] Figure 4 is the offset relationship between the field of view center Oc of the CCD camera assembly and the processing field of view center Og of the galvanometer in the present invention;

[0041] Figure 5 The processed part is under the field of view of the CCD camera assembly in the present invention;

[0042] Figure 6 is a schematic diagram of a wafer characteristic vector in the present invention;

[0043] Figure 7 A schematic diagram of a mathematical model for wafer processing path planning in the present invention;

[0044] Among them, the specific drawings are marked as:

[0045] XYZ three-axis machine tool 1, pallet 2, wafer 3, galvanometer 4, LED lighting ring 5, CCD camera assembly 6. Specific implementation mode

[0046] The following will combine the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0047] Embodiment 1

[0048] This embodiment provides a wafer processing path planning method based on off-axis machine vision, aiming to solve the distortion problem caused by the non-coaxial vision of the off-axis CCD and the galvanometer and the direction control problem of anisotropic processing of the wafer. It uses a two-dimensional galvanometer in combination with an XYZ three-axis machine tool and an off-axis CCD camera assembly, and combines the complete vision guidance process (mechanical calibration → image correction → path planning) to realize the automatic planning of the processing path on the wafer surface, reduce manual intervention. The wafer processing path generated by this solution has the characteristics of parameterizable, high-precision, low-cost, flexible and controllable, and the processing path and the positioning edge of the wafer have a specific direction, so as to achieve the effect of precisely processing a specific path on the wafer surface. The specific steps are as follows:

[0049] S1. Calibrate and synchronously control the XYZ motion axes of the machine tool. The galvanometer and the off-axis CCD camera assembly move up and down along the Z axis, and the wafer placed on the pallet moves along the XY axis. Figure 1 It is a schematic diagram of the wafer, P is the center of the wafer circle, and the vector L is the characteristic vector of the wafer.

[0050] S2. After initially calibrating the galvanometer, control the galvanometer to punch a matrix dot pattern on the calibration plate according to the theoretical dot matrix coordinates. Use a magnifying device such as a microscope to determine the coordinates of each point on the calibration plate, compare the deviation between the actual coordinates and the theoretical coordinates of the dot matrix marking points punched by the galvanometer on the calibration plate, compensate for the dynamic focusing of the galvanometer, and correct the Z-axis error of the two-dimensional galvanometer at each point.

[0051] S3. Replace the appropriate camera eyepiece to adjust the imaging position and depth of field of the CCD camera assembly, so that the imaging plane of the CCD camera assembly is on the surface of the wafer to be processed, and at the same time adjust the camera depth of field to 1-3 mm.

[0052] S4. Determine the offset vector value of the machine tool coordinate system between the vision center Oc of the CCD camera assembly and the processing vision center Og of the galvanometer.

[0053] S5. Determine the processing size ratio between the image within the field of view of the CCD camera assembly and the processing field of the galvanometer k ;

[0054] S6. Use the image detection method to obtain the center P(x0, y0) of the wafer and the two endpoints M(x1, y1) and N(x2, y2) of the feature edge within the field of view of the CCD camera assembly. According to the sign of, judge whether the feature vector is or ;

[0055] S7. The wafer processing path consists of several groups of parallel straight lines. Different processing effects can be achieved by setting the distance between the parallel straight lines and the angle formed between the parallel straight lines and the feature vector. First, draw a straight line GH through the center P of the wafer, intersecting the outer contour of the wafer at points G and H. GH is the perpendicular bisector of MN, and the pixel coordinates of points G and H are obtained. Discretize the line segment GH into a point set {Sn}, and the distance between Sn and Sn-1 is d. Then, on GH, draw a straight line RQ through point Sn, forming an angle with the feature vector, intersecting the outer contour of the wafer at points Rn and Qn. Then, the set of straight lines {RnQn} is the automatically planned processing path. Finally, multiply {RnQn} by the conversion ratio k in step S5 to obtain the processing path set in the galvanometer coordinate system. Finally, after the processing head moves along the offset vector obtained in step S4, move according to the processing path k{RnQn}.

[0056] Adopt the above technical solutions:

[0057] (1) In this application, the design of separating the off-axis CCD and the galvanometer for installation supports flexible replacement of the eyepiece of the CCD camera assembly to adapt to different field of view and depth of field requirements, combines the LED ring light source to optimize the imaging quality, breaks through the optical path limitation of the coaxial scheme, reduces the system complexity and cost, and at the same time expands the size adaptation range of the processed parts;

[0058] (2) Utilize the kinematic model of the XYZ three-axis machine tool, combine the laser interferometer or the coordinate measuring machine (CMM) to calibrate the orthogonality error, and correct it through the compensation algorithm; at the same time, perform multiple measurements on the galvanometer marking dot matrix to generate a compensation file; through dynamic calibration and error compensation, improve the motion accuracy of the multi-axis system and reduce the influence of mechanical installation errors on the processing path;

[0059] (3) Perform secondary correction on the off-axis CCD image through the perspective transformation matrix H to eliminate the field of view tilt distortion; combine the offset vector and the processing size ratio coefficient k, realize the precise mapping between CCD pixel coordinate system and galvanometer processing coordinate system, solve the distortion problem caused by the different axes of the paraxial CCD and galvanometer field of view, and propose a complete coordinate conversion chain (image correction → offset compensation → scale conversion) to ensure the geometric accuracy of the processing path;

[0060] (4) Identify the center and outer contour of the wafer through image detection methods, and determine the direction of the feature vector. Based on the direction of the feature vector, generate the angle formed with the feature vector Parallel processing path set {RnQn}, combined with the processing size ratio coefficient k Convert to the path in the galvanometer coordinate system. Directly associate the crystal orientation information (characteristic vector) with the processing path direction to achieve directional control of the anisotropic processing effect on the wafer surface.

[0061] The specific operation process of step S1 is as follows:

[0062] This step completes the calibration and software control of the machine tool XYZ motion axis. Figure 2 The structure shown is installed with XYZ motion axis, tray, CCD camera assembly and LED lighting ring. The galvanometer and the CCD camera assembly located on the side move up and down along the Z axis. The wafer placed on the tray moves along the XY axis, and the reference points of the XYZ axes are set. The reference points can be set arbitrarily on the motion path of each axis, and the kinematic model is established with the reference points of each axis as the reference points. The initial coordinate system of the machine tool is established using a laser interferometer or a three-dimensional coordinate measuring machine (CMM), the orthogonality error of the X / Y / Z axes is determined, and corrected by a compensation algorithm. The software control solution can use the EtherCAT bus and the corresponding XYZ three-axis motion model of Beckhoff to achieve three-axis synchronous control of the machine tool.

[0063] The specific operation process of step S2 is as follows:

[0064] Move the Z axis up and down so that the distance between the center point of the galvanometer optical axis and the pallet is the galvanometer focal length d, record the position of the galvanometer Z axis at this time, use the galvanometer to mark an n*n matrix target (for example, 9*9 or 25*25) on the calibration plate placed on the pallet (the calibration plate is a clean marking plate provided by the galvanometer manufacturer), observe the dot matrix marked by the galvanometer with a microscope, compare the deviation between the actual coordinates of the dot matrix marking points and the theoretical coordinates, generate a compensation file, and calibrate the two-dimensional galvanometer. After completing this step, the two-dimensional galvanometer can be calibrated.

[0065] The specific operation process of step S3 is as follows:

[0066] First, use a galvanometer to project a pattern on the calibration plate. Move the XY axes so that the pattern appears in the field of view of the off-axis CCD camera assembly. Replace the appropriate camera eyepiece until the CCD camera assembly can clearly observe the image of the galvanometer processing plane, that is, the depth of field of the camera covers the focus position of the galvanometer processing.

[0067] Then, to determine whether the imaging plane of the CCD camera assembly and the surface of the wafer to be processed on the tray (located in the XY plane) are in parallel planes, four standard marking points are provided on the tray. Their actual coordinates in the machine tool coordinate system are A(x1, y1), B(x2, y2), C(x3, y3), D(x4, y4), which are located at the four vertices of a square respectively. To distinguish them, different colors are applied to the marking points to avoid confusion in the position of the CCD camera at the corresponding vertices. Move the CCD camera to capture these four vertices in the center of the field of view and detect these four vertices in the image. Their pixel coordinates are A1(u1, v1), B1(u2, v2), C1(u3, v3), D1(u4, v4). If A1, B1, C1, D1 are also the four vertices of a square, it proves that the imaging plane of the CCD camera assembly and the tray plane are parallel. However, due to mechanical installation errors, there is often a slight tilt in the field of view (as Figure 3 shown). The mechanical clamping tool of the CCD camera assembly can be adjusted for tilt correction. If the mechanical clamping mechanism cannot perform fine adjustment, the CCD image can be corrected twice using an algorithm. First, obtain the transformation matrix H of the pixel coordinates of the four vertices in the CCD image under the tilt angle before correction to the CCD image coordinates along the Z-axis perpendicular to the downward field of view. Then, use the transformation matrix H to correct the pixel coordinates of the four vertices in the CCD image twice to obtain an imaging picture parallel to the tray plane, that is, the secondary correction of the CCD field of view is achieved.

[0068] The specific steps for using the algorithm to perform secondary correction on the CCD image are as follows:

[0069] Use the Hough detection visual recognition to obtain the pixel coordinates of the four vertices in the CCD image before correction: A1(u1, v1), B1(u2, v2), C1(u3, v3), D1(u4, v4). The pixel coordinates of the corresponding vertices after correction are ( u i , v i ), i = 1, 2, 3, 4. Then, the conversion relationship between ( u i , v i ) and ( x i , y i ) satisfies the following perspective formula:

[0070] ;

[0071] Among them, the perspective transformation matrix H is:

[0072] ;

[0073] w is the coordinate ratio between the CCD pixel coordinate system and the actual machine tool coordinate system, which can be obtained by measuring the ratio of the actual radius of the calibration point on the pallet to the CCD pixel radius. For the convenience of calculation, w = 1 is taken here. For each point pair ( u i , v i ) → ( x i , y i ) two equations can be generated:

[0074] ;

[0075] A total of 4 point pairs yield 8 equations to solve for the 8 parameters of the transformation matrix H h 11 , h 12 , h 13 , h 21 , h 22 , h 23 , h 31 , h 32 . Rewrite the equations as the homogeneous linear equation system form of A · h = b . Through the least squares method or singular value decomposition, where

[0076] ;

[0077] ;

[0078] ;

[0079] Solve , then the solution of the perspective transformation matrix H is obtained.

[0080] Let the set of pixel points after CCD calibration be , from the set of pixel points of the original CCD image before calibration , the CCD camera image can be corrected by using the following formula to obtain an imaging picture that is almost parallel to the processing tray plane.

[0081] ;

[0082] where n is the pixel size of the CCD camera.

[0083] The specific operation process of step S4 is as follows:

[0084] Since the CCD camera assembly is installed paraxially relative to the galvanometer, there is an offset vector between the field center Oc of the CCD camera assembly and the processing field center Og of the galvanometer , specifically as shown in Figure 4 . To determine the offset vector, first place the calibration plate below the galvanometer field of view, adjust the Z-axis distance so that the processing focal plane of the galvanometer falls on the calibration plate. At this time, the XY-axis position of the galvanometer processing center is Og(x1, y1). Then, process a calibration pattern (such as a point) on the calibration plate. After completion, keep the Z-axis position unchanged and move the XY-axis so that the calibration pattern is located at the center of the CCD camera field of view. Record the XY-axis position at this time as Oc(x2, y2). At this time, a set of offset vectors =(x2 - x1, y2 - y1) is obtained. Repeat the above method multiple times to measure the value of the offset vector and use its mean value to represent the offset vector . Its meaning is the offset in the machine tool coordinate system from the center point of the galvanometer processing field of view to the center point of the CCD camera assembly field of view.

[0085] The specific operation process of step S5 is as follows:

[0086] Determine the processing size ratio between the image under the field of view of the CCD camera assembly and the processing field of the galvanometer k , that is, use the galvanometer to draw a straight line with a length of f on the calibration plate, move it according to the offset vector so that the straight line appears in the field of view of the CCD camera assembly, and use the image to detect the pixel length of this straight line as F. Then k = f / F, and the mean value can be obtained by measuring multiple times. The processing size ratio k means the conversion ratio when converting the processing path planned by the CCD vision to the galvanometer coordinate system.

[0087] The specific operation process of step S6 is as follows:

[0088] After the above steps are completed, a laser processing system with paraxial CCD vision guidance can be obtained. After that, the workpiece to be processed needs to be identified and the automatic path planning needs to be carried out.

[0089] Such as Figure 5Shown is the workpiece under the field of view of the CCD camera assembly. By using mature image detection methods such as Hough detection, the center point P(x0, y0) of the wafer and the two end points M(x1, y1) and N(x2, y2) of the feature edge are obtained under the field of view of the CCD camera assembly. The feature edge of the wafer is a vector, and its direction satisfies the right-hand helix. For the processing paths at different angles relative to this vector edge, there are certain differences in the obtained wafer processing effects. Therefore, an algorithm is needed to identify whether the feature vector is or as shown in Figure 6 . To determine the feature vector, the following calculations are performed:

[0090] ;

[0091] where the direction of k is perpendicular to the paper surface and outward, that is, the positive direction of the Z-axis. Therefore, only by judging the sign of , the feature vector can be judged as or .

[0092] That is >0, then the feature quantity is ;

[0093] <0, then the feature quantity is .

[0094] The specific operation process of step S7 is as follows:

[0095] Assume that the direction of the feature vector has been determined in step S6 as . Next is to plan the processing path. The wafer processing path generally consists of several groups of parallel lines. By setting the distance between the parallel lines and the angle formed between the parallel lines and the feature vector, different processing effects can be achieved. First, through visual recognition, the center point P of the wafer and the outer contour set have been obtained. As shown in Figure 7 , draw a straight line GH through the center point P of the wafer, which intersects the outer contour of the wafer at points G and H. GH is the perpendicular bisector of MN, and the pixel coordinates of points G and H are obtained. The line segment GH is discretized into a point set {Sn}, and the distance between Sn and Sn-1 is d. Since the pixel coordinates of GH have been determined, the point set {Sn} can be obtained. Then, on GH, draw a straight line RQ through point Sn and forming an angle with the feature vector , which intersects the outer contour of the wafer at points Rn and Qn. Then the straight line set {RnQn} is the automatically planned processing path. Finally, {RnQn} is multiplied by the conversion ratio k in step S5 to obtain the processing path set in the galvanometer coordinate system; finally, the processing head moves along the offset vector obtained in step S4 After moving, move along the processing path k{RnQn}, and the method for planning a specific processing path of a wafer using off-axis vision according to the present invention can be realized.

[0096] Embodiment 2

[0097] This embodiment provides a wafer processing path planning device based on off-axis machine vision for implementing the above-mentioned wafer processing path planning method, as Figure 2 shown. The device includes an XYZ three-axis machine tool 1, a tray 2, a galvanometer scanner 4, and a CCD camera assembly 6. The tray 2 is installed on the XY orthogonal motion axes. The wafer 3 is placed on the tray 2. The Z-axis motion axis is fixed in the middle of the gantry. The galvanometer scanner 4 is installed at the output end of the Z-axis motion axis. A CCD camera assembly 6 is fixedly installed on the side of the galvanometer scanner 4. The field of view of the CCD camera assembly 6 is downward along the Z-axis. An LED light ring 5 is installed on the outer side surface below the galvanometer scanner 4.

[0098] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A wafer processing path planning method based on paraxial machine vision, characterized in that It includes the following steps: S1. Calibrate and synchronously control the XYZ moving axes of the machine tool. The galvanometer and the CCD camera assembly located on the side move up and down along the Z axis, and the wafer to be processed placed on the tray moves along the XY axis direction; S2. Compare the deviation between the actual coordinates and the theoretical coordinates of the dot matrix marking points hit by the galvanometer on the calibration plate, generate a compensation file, and correct the two-dimensional galvanometer; S3. Replace the appropriate camera eyepiece to adjust the imaging position and depth of field of the CCD camera assembly, so that the imaging plane of the CCD camera assembly is on the surface of the wafer to be processed, and at the same time adjust the camera depth of field to 1 - 3 mm; S4. Determine the offset vector in the machine tool coordinate system between the vision center Oc of the CCD camera assembly and the processing vision center Og of the galvanometer value; S5. Determine the processing dimension ratio between the image within the field of view of the CCD camera assembly and the processing field of view of the galvanometer scanner k ; S6. Obtain the center point P(x0, y0) of the wafer and the two end points M(x1, y1) and N(x2, y2) of the feature edge in the field of view of the CCD camera assembly by using an image detection method. According to the sign of to determine whether the feature vector is or S7. The wafer processing path consists of several sets of parallel straight lines. Different processing effects are achieved by setting the distance between the parallel straight lines and the angle formed between the parallel straight lines and the feature vector. First, draw a straight line GH through the center P of the wafer, intersecting the outer contour of the wafer at two points G and H. GH is the perpendicular bisector of MN, and the pixel coordinates of points G and H are obtained. The line segment GH is discretized into a point set {Sn}, and the distance between Sn and Sn-1 is d. Then, draw a straight line RQ through point Sn on GH, forming an angle with the feature vector, intersecting the outer contour of the wafer at points Rn and Qn. Then, the set of straight lines {RnQn} is the automatically planned processing path. Finally, {RnQn} is multiplied by the conversion ratio k in step S5 to obtain the set of processing paths in the galvanometer coordinate system. Finally, after the processing head moves along the offset vector obtained in step S4, it can move according to the processing path k{RnQn}.

2. The method for wafer processing path planning based on paraxial machine vision according to claim 1, wherein In the step S3, the specific process of replacing the appropriate camera eyepiece to adjust the imaging position and depth of field of the CCD camera assembly is as follows: Use the galvanometer to hit a pattern on the calibration plate, move the XY axis to make the pattern appear in the field of view of the off-axis CCD camera assembly, and replace the appropriate camera eyepiece until the CCD camera assembly clearly observes the image of the galvanometer processing plane.

3. The method for wafer processing path planning based on paraxial machine vision according to claim 1 or 2, characterized in that, In the step S3, the specific process of adjusting the imaging plane of the CCD camera assembly to be on the surface of the wafer to be processed is as follows: According to the four marking points located on the tray and at the four vertex positions of the square in the machine tool coordinate system, observe the pixel coordinates of the four marking points in the image captured by the off-axis CCD camera assembly, and perform tilt correction on the CCD camera assembly and / or perform secondary correction on the pixel coordinates in the CCD image using an algorithm until the pixel coordinates of the four marking points in the CCD image are also the four vertices of the square.

4. The method for wafer processing path planning based on paraxial machine vision according to claim 3, wherein, Adjust the mechanical clamping tool of the CCD camera assembly for tilt correction.

5. The wafer processing path planning method based on paraxial machine vision according to claim 3, wherein The specific process of performing secondary correction on the pixel coordinates in the CCD image using an algorithm is as follows: First, obtain the transformation matrix H of the pixel coordinates of the four vertices in the CCD image under the tilted viewing angle before correction to the CCD image coordinates of the field of view vertically downward along the Z axis, and then use the transformation matrix H to perform secondary correction on the pixel coordinates of the four vertices in the CCD image to obtain an imaging picture parallel to the tray plane.

6. The method for wafer processing path planning based on paraxial machine vision according to claim 3, wherein The four marking points on the tray are respectively painted with different colors.

7. The method for wafer processing path planning based on paraxial machine vision according to claim 1, wherein In the said step S6, > 0, the feature amount is ; < 0, then the characteristic quantity is .

8. The method for wafer processing path planning based on paraxial machine vision according to claim 1, wherein In the step S4, first place the calibration plate below the field of view of the galvanometer, adjust the Z-axis distance so that the processing focal plane of the galvanometer falls on the calibration plate. At this time, the positions of the XY axes of the galvanometer processing center are Og(x1, y1). Then process the calibration pattern on the calibration plate. After completion, keep the Z-axis position unchanged and move the XY axes so that the calibration pattern is located at the center of the CCD camera's field of view. Record the positions of the XY axes at this time as Oc(x2, y2). At this time, a set of offset vectors is obtained =(x2 - x1, y2 - y1). Repeat the above method multiple times to measure the values of the offset vectors and use their mean value to represent the value of the offset vector .

9. The method for wafer processing path planning based on paraxial machine vision according to claim 1, wherein In the step S2, move the Z axis up and down so that the distance between the center point of the galvanometer optical axis and the tray is the focal length d of the galvanometer, record the position of the galvanometer Z axis at this time, use the galvanometer to hit an n*n matrix target on the calibration plate placed on the tray, observe the dot matrix hit by the galvanometer with a microscope, compare the deviation between the actual coordinates and the theoretical coordinates of the dot matrix marking points, generate a compensation file, and correct the two-dimensional galvanometer.

10. A wafer processing path planning device based on paraxial machine vision, characterized in that, A device for implementing the wafer processing path planning method as claimed in claim 1, the device includes an XYZ three-axis machine tool, a tray, a galvanometer and a CCD camera assembly. The tray is installed on the XY orthogonal moving axes, the Z-axis moving axis is fixed in the middle of the gantry, the galvanometer is installed at the output end of the Z-axis moving axis, a CCD camera assembly is fixedly installed on the side of the galvanometer, the field of view of the CCD camera assembly is downward along the Z axis, and an LED lighting lamp ring is installed on the outer side surface below the galvanometer.

Images