A method for fabricating metal antennas based on visual recognition correction

Abstract

This invention relates to the field of antenna fabrication technology, and more particularly to a method for fabricating a metal antenna based on visual recognition correction. The steps include: S10: setting positioning marks on the front side of a substrate; printing a dot matrix on the back side of the substrate; applying adhesive to the front side of the substrate; S20: drying the substrate and then laminating it with metal foil; S30: simultaneously photographing the front and back sides of the substrate to determine the origin and coordinate system of the back image; S40: obtaining the designed positions of each point in the dot matrix and the measured positions of each point in the back image; establishing a correction model based on the deviation between the designed and measured positions, and verifying the error of the correction model; S50: using the coordinates of the original cutting position of the antenna as input, calculating the coordinates of the corrected position of the antenna after substrate deformation using the correction model, and cutting according to the corrected position to obtain the final finished metal antenna. This invention enables precise cutting of the metal antenna based on the deformation of the substrate.

Description

Technical Field

[0001] This invention relates to the field of antenna fabrication technology, and in particular to a method for fabricating a metal antenna based on visual recognition correction. Background Technology

[0002] During the fabrication of metal antennas, substrate deformation can occur due to material properties and processing techniques. After the substrate is deformed, it is necessary to perform laser cutting and other operations to cut out the antenna shape. If the cutting position is cut according to the original design, it may cut to the edge, resulting in an incomplete antenna structure. Therefore, the cutting position must be changed accordingly.

[0003] In existing related technologies, the common approach is to control the temperature and humidity of the environment, place the substrate in a special environment to cause deformation, and then experimentally determine the average deformation rate corresponding to the environment. The cutting position is then modified according to this average deformation rate. However, this method requires extensive experimental testing, which is time-consuming and costly. Furthermore, the deformation of each substrate is random; even in the same environment, the deformation of the substrate can vary significantly. Therefore, the cutting position modified using the average deformation rate is not applicable to all substrates, resulting in a low antenna yield.

[0004] Therefore, there is a need for an antenna fabrication method that can precisely cut the substrate based on its deformation characteristics. Summary of the Invention

[0005] This invention provides a method for fabricating a metal antenna based on visual recognition correction, which can effectively solve the problems in the background art.

[0006] This invention provides a method for fabricating a metal antenna based on visual recognition correction, comprising the following steps:

[0007] S10: Set positioning marks on the front side of the substrate; print a dot matrix on the back side of the substrate, with equal spacing between adjacent dots; apply adhesive to the front side of the substrate;

[0008] S20: The substrate is dried and then laminated with metal foil;

[0009] S30: Take photos of the front and back of the substrate simultaneously, and determine the origin and coordinate system of the back image based on the positioning mark information in the front image of the substrate.

[0010] S40: Obtain the designed position of each point in the dot matrix and the measured position of each point in the dot matrix in the back image. Based on the positional deviation between the designed position and the measured position of each point, establish a correction model and verify the error of the correction model.

[0011] S50: The coordinates of the original cutting position of the antenna are used as input. The coordinates of the corrected position of the antenna after the substrate deformation are calculated by the correction model. Then, the antenna is cut according to the corrected position to obtain the final metal antenna product.

[0012] Furthermore, in step S10, the positioning marks on the front side of the substrate include two rows of dots arranged in parallel, with the same number of marking points in each row of dots.

[0013] In step S30, the centroids of the distribution of each row of markers are calculated and denoted as points C1 and C2, respectively. Then, the center points of points C1 and C2 are denoted as the origin. The fitted line of each row of markers and the average slope kavg of the two fitted lines are calculated. The line with a slope of kavg and passing through the origin is used as the x-axis to form a coordinate system.

[0014] Further, in step S30, the points in the first row are denoted as A1(a1x,a1y), A2(a2x,a2y), ..., An(anx,any); and the points in the second row are denoted as B1(b1x,b1y), B2(b2x,b2y), ..., Bn(bnx,bny).

[0015] Let C1(c1x,c1y) and C2(c2x,c2y) be the points;

[0016] Then we have c1x = (a1x + a2x + a3x + a4x) / 4;

[0017] c1y=(a1y+a2y+a3y+a4y) / 4;

[0018] c2x=(b1x+b2x+b3x+b4x) / 4;

[0019] c2y=(b1y+b2y+b3y+b4y) / 4.

[0020] Furthermore, in step S40, the specific algorithm for correcting the model is as follows:

[0021] Number the points in the dot matrix from left to right and then from top to bottom as 1, 2...N. Let the designed position of the i-th point in the dot matrix be (xi, yi) and the measured position be (xi', yi').

[0022] Construct an N×10 mutual influence matrix M, where the elements in the i-th row of the mutual influence matrix M are: 1, xi, yi, xi 2 , xi·yi, yi 2 xi 3 xi 2 ·yi、xi·yi 2 yi 3 ;

[0023] Construct observation vectors Dx and Dy;

[0024] Dx=[x1',x2',……,xN'] T ;

[0025] Dy=[y1',y2',……,yN'] T ;

[0026] The correction model is set with two correction vectors, Tx and Ty:

[0027] Tx=[α1,α2,……,α10] T =(M T ·M+Λx) -1 ·M T ·Dx;

[0028] Ty=[β1,β2,……,β10] T =(M T ·M+Λy) -1 ·M T ·Dy;

[0029] The initial values ​​of Λx and Λy are both diagonal matrices with a set value λ0 on the diagonal and 0 for the rest.

[0030] In step S50, the original cutting position of the j-th antenna is denoted as (xj, yj), and the corrected position is denoted as (T(xj), T(yj)).

[0031] T(xj)=[1, xj, yj, xj 2 、xj·yj、yj 2 、xj 3 、xj 2 ·yj、xj·yj 2 yj 3 ]·Tx;

[0032] T(yj)=[1、xj、yj、xj 2 、xj·yj、yj 2 、xj 3 、xj 2 ·yj、xj·yj 2 yj 3 ]·Ty.

[0033] Furthermore, in step S40, the error verification model is specifically as follows:

[0034] The coordinates of the designed positions of each point in the dot matrix are used to calculate the correction reference position through the correction model. The coordinates of the correction reference position of the i-th point are denoted as (T(xi),T(yi)).

[0035] Calculate the mean square error (MSE) of the correction reference position and the measured position at each point within the lattice:

[0036] ;

[0037] When the MES is less than the set threshold, the error verification of the correction model is passed; otherwise, the elements in Λx and Λy are modified and the correction model is rebuilt until the error verification is passed.

[0038] Furthermore, the elements in Λx and Λy are modified as follows:

[0039] Calculate the average absolute difference of the abscissa Δxavg, the average absolute difference of the ordinate Δyavg, and the average distance Δdavg between the correction reference position and the measured position of all points in the dot matrix.

[0040] For the element in row p and column p in Λx and Λy, update it to λxp and λyp;

[0041] λxp=λ0+k0·mxp·exp(-△avg / △xavg);

[0042] λyp=λ0+k0·myp·exp(-△avg / △yavg);

[0043] Where k0 is the set base multiplier;

[0044] mxp is the power of xi in the p-th column of the mutual influence matrix M;

[0045] myp is the power of yi in the p-th column of the mutual influence matrix M.

[0046] Furthermore, after each calculation of all λxp and λyp, the maximum and minimum values ​​of α1~α10 and β1~β10 are obtained. If the maximum or minimum value of α1~α10 exceeds the set range, all λxp are adjusted so that α1~α10 can all fall within the set range while maintaining the proportions between them. If the maximum or minimum value of β1~β10 exceeds the set range, all λxp are adjusted so that β1~β10 can all fall within the set range while maintaining the proportions between them.

[0047] Furthermore, in step S50, a cutting border is set on the substrate, and multiple equally spaced design border points are set on the cutting border. All design border points are calculated using a correction model to obtain corresponding correction border points, and then all correction border points are connected in sequence to form a cutting area.

[0048] After calculating the antenna's correction position, determine whether each antenna region is within the cutting area and does not overlap with other antenna regions. If not, translate the antenna's correction position until the requirements are met.

[0049] Furthermore, the translation of the correction position is specifically as follows:

[0050] If the antenna area exceeds the cutting area, select the two correction frame points closest to the antenna area and move the correction position along the vertical direction of the line connecting the two correction frame points.

[0051] If two antenna regions overlap, compare the overlap lengths of the two antenna regions in the horizontal and vertical directions, select the minimum overlap length and the direction in which the minimum overlap length is located, and simultaneously translate the two antenna regions away from each other.

[0052] Furthermore, in step S40, before calculating the positional deviation, the number of points in the dot matrix of the back image is counted. If the counted number is different from the designed number, the missing points are identified, and the missing points are determined by interpolation using the neighboring points of the missing points.

[0053] The technical solution of this invention can achieve the following technical effects:

[0054] This method uses a dot matrix to accurately reflect the shape changes of various parts of the substrate, and then corrects the original cutting position based on these shape changes. This makes the formation of the correction position no longer dependent on experiments, but can be accurately corrected according to the actual deformation of each substrate. It upgrades the traditional static correction that relies on empirical data to dynamic adaptive correction, which can be applied to the fabrication of metal antennas in any production environment and material. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a method for fabricating a metal antenna based on visual recognition correction according to the present invention;

[0057] Figure 2 This is a schematic diagram of the front side of the substrate and the positioning marks in this invention;

[0058] Figure 3 This is a schematic diagram of the back side of the substrate and the dot matrix before deformation in this invention;

[0059] Figure 4 This is a schematic diagram of the back side of the substrate and the dot matrix after the substrate is deformed in this invention. Detailed Implementation

[0060] The basic principles and main features of the technical solution of the present invention will be described below with reference to the accompanying drawings of the embodiments of the present invention. The following description will use one or more embodiments for a more intuitive understanding. These embodiments are merely some, not all, of the embodiments of the present invention.

[0061] In the description of this invention, the terms indicating orientation or positional relationship (such as up, down, left, right, etc.) are based on the orientation shown in the drawings or some conventional positional relationships, and are only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the features referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting this invention.

[0062] A method for fabricating a metal antenna based on visual recognition correction, such as Figure 1 As shown, the steps include:

[0063] S10: Set positioning marks on the front side of the substrate, such as... Figure 2 As shown, the parts of the substrate cut to form the antenna shape are usually divided into multiple groups in columns. The positioning marks are set above and below each column to mark the cut area. The radius of the dot of the positioning mark can be relatively large, so that even if the substrate is deformed, the impact on the positioning mark as a whole will be relatively small, and the position of the origin and coordinates of each image in the subsequent detection will be relatively consistent.

[0064] Print a dot matrix on the back of the substrate, such as Figure 3 As shown, the dot matrix consists of P rows and Q columns of dots, for a total of N dots; the radius of each dot in the dot matrix can be relatively small, so that each dot can move more significantly with the deformation of the substrate, which facilitates the accurate calculation of the degree of substrate deformation.

[0065] Apply adhesive to the front of the substrate, ensuring that the adhesive completely covers the antenna area and avoids the positioning mark area, so that the metal foil does not obscure the positioning mark during lamination.

[0066] S20: The substrate is dried and then laminated with metal foil;

[0067] S30: Take pictures of the front and back of the substrate simultaneously. Determine the origin and coordinate system of the front image based on the positioning mark information in the front image of the substrate. Since the cameras on the front and back are fixed, the correspondence between the front and back images of all substrates is the same. At this time, the origin and coordinate system in the front image can be directly used as the origin and coordinate system of the back image.

[0068] S40: Obtain the design position of each point in the dot matrix. The design position is mainly determined based on the distance between each point during the dot matrix design. That is, there will be a virtual coordinate system during the substrate design. Each point in the dot matrix will generate a virtual coordinate. After the actual origin and coordinate system are determined, each virtual coordinate can be converted into the corresponding design position (xi, yi), where i takes 1~N.

[0069] To obtain the measured positions of each point in the dot matrix in the back image, the back image needs to be identified. For example... Figure 4 As shown, due to the deformation of the substrate, the position of each point in the dot matrix will change with the deformation of the substrate. Therefore, it is necessary to identify each point in the dot matrix in the back image, and then determine its coordinates based on the current actual position of each point, which is the measured position (xi', yi'), where i takes 1~N.

[0070] Based on the positional deviation between the designed and measured positions of each point, the deformation of the substrate at each point can be determined, and a correction model can be established. Of course, the correction model generally cannot solve for the optimal solution in one go. Therefore, it is necessary to verify the error of the correction model and correct it according to the error until the correction model passes the error verification.

[0071] S50: The coordinates of the original cutting position of the antenna are used as input. The coordinates of the corrected position of the antenna after the substrate deformation are calculated by the correction model. Then, the antenna is cut according to the corrected position to obtain the final metal antenna product.

[0072] This method uses a dot matrix to accurately reflect the shape changes of various parts of the substrate, and then corrects the original cutting position based on these shape changes. This eliminates the reliance on experiments or human experience in determining the correction position, allowing for precise correction based on the actual deformation of each substrate. It upgrades traditional static correction, which relies on experience-based data, to dynamic adaptive correction, making it applicable to the fabrication of metal antennas in any production environment and with any material. Furthermore, the correction process is intelligently and automatically implemented using a vision-based algorithm, eliminating the need for cumbersome preliminary experiments and significantly reducing errors caused by manual inspection, thereby significantly improving antenna performance consistency and yield.

[0073] In step S10, the positioning marks on the front side of the substrate include two rows of dots arranged in parallel. The two rows of dots are located above and below the cutting area, respectively. Each row of dots has the same number of marking points, and the marking points of the two rows of dots are vertically aligned.

[0074] In step S30, the centroids of the distribution of each row of markers are calculated and denoted as point C1 and point C2 respectively. The centroid can be regarded as the center of the polygon formed by connecting multiple markers in each row. The deformation of the substrate is mainly caused by the temperature change of the environment. Therefore, larger deformations will accumulate at the edge of the substrate. Therefore, using the simulated centroid as a reference can make the reference positions between different substrates coincide as much as possible.

[0075] Then, the center point of points C1 and C2 is recorded as the origin. The center point obtained at this time can be regarded as the centroid of the substrate. After the above two centroid calculations, the origins formed by different substrates can basically coincide with each other in space.

[0076] Calculate the fitted straight line for each row of marked points and the average slope kavg of the two fitted straight lines. Use the straight line with slope kavg that passes through the origin as the x-axis to form a coordinate system. Since the orientation of the substrate is fixed during the previous processing, all substrates can be set to a common x-axis orientation in this step, without needing to distinguish each substrate individually.

[0077] In step S30, the specific calculation steps for points C1 and C2 are as follows: Let the points in the first row be A1(a1x,a1y), A2(a2x,a2y), ..., An(anx,any); let the points in the second row be B1(b1x,b1y), B2(b2x,b2y), ..., Bn(bnx,bny);

[0078] Let the coordinates of point C1 be (c1x, c1y) and the coordinates of point C2 be (c2x, c2y);

[0079] Then we have c1x = (a1x + a2x + a3x + a4x) / 4;

[0080] c1y=(a1y+a2y+a3y+a4y) / 4;

[0081] c2x=(b1x+b2x+b3x+b4x) / 4;

[0082] c2y=(b1y+b2y+b3y+b4y) / 4;

[0083] The coordinates of the origin are ((c1x+c2x) / 2, (c1y+c2y) / 2).

[0084] Preferably, in step S40, the specific algorithm for correcting the model is as follows:

[0085] Number the points in the dot matrix from left to right and then from top to bottom as 1, 2...N. Let the designed position of the i-th point in the dot matrix be (xi, yi) and the measured position be (xi', yi').

[0086] Since the deformation of the substrate is not necessarily linear, multi-order interference needs to be considered when simulating material deformation. Therefore, an N×10 mutual influence matrix M is constructed, and the elements in the i-th row of the mutual influence matrix M are: 1, xi, yi, xi 2 , xi·yi, yi 2 xi 3 xi 2 ·yi、xi·yi 2 yi 3 , where xi and yi are the design positions (xi, yi) of the i-th point, and each element simulates the results of the interaction between xi and yi to different degrees. The interaction matrix M is as follows:

[0087] .

[0088] Then, two N×1 observation vectors, Dx and Dy, are constructed;

[0089] Dx=[x1',x2',……,xN'] T ;

[0090] Dy=[y1',y2',……,yN'] T ;

[0091] Where xi' and yi' are the measured positions (xi', yi') of the i-th point.

[0092] Using the mutual influence matrix M and the observation vectors Dx and Dy, a preliminary correction model T can be fitted and calculated. The correction model T includes two N×1 correction vectors: Tx, used to calculate the position along the x-axis, and Ty, used to calculate the position along the y-axis.

[0093] Tx=[α1,α2,……,α10] T =(M T ·M+Λx) -1 ·M T ·Dx;

[0094] Ty=[β1,β2,……,β10] T =(M T ·M+Λy) -1 ·M T ·Dy;

[0095] Where Λx and Λy are both initial values ​​of diagonal matrices with a set value λ0 on the diagonal and all other elements being 0. Both Λx and Λy are 10×10 matrices. λ0 is the regularization coefficient, used to balance the fitting accuracy and generalization ability of the model. λ0∈

[10] −6 [1], which is usually taken as 0.1 in this method.

[0096] After obtaining the correction model T, the correction position of each point can be calculated in step S50; if the original cut position of the j-th antenna is denoted as (xj, yj), and the corrected position of that point is denoted as (T(xj), T(yj)), then:

[0097] T(xj)=[1, xj, yj, xj 2 、xj·yj、yj 2 、xj 3 、xj 2 ·yj、xj·yj 2 yj 3 ]·Tx;

[0098] T(yj)=[1、xj、yj、xj 2 、xj·yj、yj 2 、xj 3 、xj 2 ·yj、xj·yj 2 yj 3 ]·Ty;

[0099] It should be noted that the values ​​of xj and yj are not necessarily in x1~xN and y1~yN, that is, the original cutting position coordinates of the antenna may not coincide with the designed position of the point in the array. However, since the correction model T mainly simulates the correspondence between the points before and after the substrate deformation, it can still be calculated through the correction model T even if the cutting position does not coincide with the point in the array.

[0100] Of course, the corrected model T obtained through a single fitting may not be the optimal solution. Therefore, when calculating a corrected model T in step S40, it is necessary to perform error verification on the corrected model T. The specific error verification model is as follows:

[0101] The coordinates of the designed positions of each point in the dot matrix are used to calculate the correction reference position through the correction model. The coordinates of the correction reference position of the i-th point are denoted as (T(xi),T(yi)).

[0102] Calculate the mean square error (MSE) of the correction reference position and the measured position at each point within the lattice:

[0103] ;

[0104] ||(T(xi),T(yi))-(xi',yi')|| represents the distance between point (T(xi),T(yi)) and point (xi',yi');

[0105] MSE represents the difference between the correction reference position calculated by the correction model T for each point within the dot matrix and the actual measured position of each point within the dot matrix. When MSE is less than the set threshold, it indicates that the correction reference position and the actual measured position of each point within the dot matrix are close, the fitted correction model T is more in line with the actual situation, and the error verification of the correction model T is passed. Otherwise, the elements in Λx and Λy are modified and the correction model is reconstructed until the error verification is passed.

[0106] When error verification fails and elements in Λx and Λy need to be modified, the amount of modification for each element in Λx and Λy can be pre-set and then modified. However, this modification method is not very targeted and cannot quickly approximate the optimal values ​​of Λx and Λy. Therefore, this method provides a modification method that can more accurately modify the corresponding elements in Λx and Λy, specifically:

[0107] Calculate the average value of the absolute difference between the correction reference position and the measured position of all points in the matrix on the x-axis, △xavg; calculate the average value of the absolute difference between the correction reference position and the measured position of all points in the matrix on the y-axis, △yavg; calculate the average distance between the correction reference position and the measured position of all points in the matrix, △davg.

[0108] For the element in row p and column p in Λx and Λy, update it to λxp and λyp;

[0109] λxp=λ0+k0·mxp·exp(-△avg / △xavg);

[0110] λyp=λ0+k0·myp·exp(-△avg / △yavg);

[0111] Wherein, k0 is the set base multiplier, which has a certain proportional relationship with λ0, and is usually 0.05λ0;

[0112] exp() represents an expression with the natural number e as the base and the content in parentheses as the power. For example, exp(3) means e 3 ;

[0113] mxp is the power of xi in the p-th column of the mutual influence matrix M. For example, in the 7th column of the mutual influence matrix M, xi is... 2 If yi, then mx7 is a power of xi, that is, mx7 = 2;

[0114] myp is the power of yi in the p-th column of the mutual influence matrix M. For example, in the 7th column of the mutual influence matrix M, it is xi. 2 If yi is a power of yi, then my7 is a power of yi, i.e., my7 = 1.

[0115] The meaning of this formula is as follows: α1~α10 and β1~β10 represent the result weights under the mutual influence of xi and yi. When the verification of the correction model T fails, the main problem lies in the numerical distribution of the weights α1~α10 and β1~β10. Therefore, it is necessary to adjust the values ​​of α1~α10 and β1~β10 based on the difference between the correction reference position and the measured position. According to Tx=[α1,α2,……,α10] T =(M T ·M+Λx) -1 ·M T ·Dx, and Ty=[β1,β2,…,β10] T =(M T ·M+Λy) -1 ·M T ·Dy, where the elements in M, Dx and Dy are all fixed values, so the adjustment of the values ​​of α1~α10 and β1~β10 in the correction model T can only be done by adjusting the elements on each diagonal of Λx and Λy. It can be seen that the larger the values ​​of λx1~λx10 and λy1~λy10, the larger the corresponding values ​​of α1~α10 and β1~β10.

[0116] When xavg / △avg approaches 1, it indicates that the error between the correction reference position and the measured position is mainly concentrated in the x-direction, meaning that the numerical ratios between α1 and α10 differ significantly from the optimal value. In this case, the value of exp(-△avg / △xavg) will be larger, increasing the basic change in α1 to α10 and enabling α1 to change rapidly. Conversely, if the difference between △xavg and △avg is larger, i.e., △avg / △xavg is greater than 1, it indicates that the error between the correction reference position and the measured position is significantly different. The less the error between positions is in the x-direction, meaning the numerical ratios between α1~α10 are closer to the optimal value, the smaller the value of exp(-△avg / △xavg) will be. This reduces the basic changes in α1~α10, maintaining the original numerical ratios between α1~α10 as much as possible. mxp is used to amplify the basic changes in α1~α10. When the power of xi is high, the corresponding influence in the x-direction will be greater. Therefore, it is necessary to appropriately amplify the corresponding changes in α to account for this difference in influence. The calculation meaning of λyp is similar to that of λxp, and will not be elaborated here.

[0117] To prevent λxp and λyp values ​​from being too large or too small, which could cause α1~α10 and β1~β10 values ​​to fall into unreasonable ranges, especially since α1 and β1 are constants, after each calculation of all λxp and λyp, the maximum and minimum values ​​of α1~α10 and β1~β10 are obtained. If the maximum or minimum value of α1~α10 exceeds the set range, all λxp values ​​are adjusted so that α1~α10 can fall within the set range while maintaining the relative proportions between them. Similarly, if the maximum or minimum value of β1~β10 exceeds the set range, all λxp values ​​are adjusted so that β1~β10 can fall within the set range while maintaining the relative proportions between them.

[0118] Preferably, in step S50, a cutting border is also set on the substrate. The cutting border can be the edge of the substrate or any shape set within the substrate. The cutting border is mainly set according to the antenna manufacturing requirements. For example, under a specific process, some parts of the substrate cannot be cut as antenna parts because of missing components or inability to attach metal. Setting a cutting border is to avoid these parts. Furthermore, a correction model T is needed to determine the change of the cutting border after the substrate is deformed, so as to avoid cutting these parts during cutting. When using the correction model T for calculation, multiple equally spaced design border points can be set on the cutting border first, and the corresponding correction border points can be calculated through the correction model. Then, all the correction border points are connected in sequence to form the corrected cutting area.

[0119] The antenna region is defined, and it is usually larger than the actual shape of the antenna. The antenna region is a relatively simple rectangle, so that even if two antenna regions are close together, there is enough space for support during cutting to prevent the antenna from deforming. After calculating the correction position of the antenna, it is determined whether each antenna region is within the cutting area and does not overlap with other antenna regions. If not, the correction position of the antenna is translated until the requirements are met. The translation distance should be as short as possible to avoid the final position of the antenna deviating too far from the calculated correction position of the antenna.

[0120] Preferably, the translation of the correction position is specifically as follows:

[0121] If the antenna area exceeds the cutting area, select the two correction frame points closest to the antenna area and move the correction position along the vertical direction of the line connecting the two correction frame points.

[0122] If two antenna regions overlap, compare the overlap lengths of the two antenna regions in the horizontal and vertical directions, select the minimum overlap length l and the direction in which the minimum overlap length is located, and simultaneously translate the two antenna regions away from each other, with the translation length being l / 2.

[0123] To avoid calculation errors caused by missing points in the image due to reflections, occlusions, or other reasons, it is preferable to first count the number of points in the dot matrix of the back image in step S40 before calculating the positional deviation. If the counted number differs from the designed number, the missing points are identified, and the missing points are determined using an interpolation method based on their neighboring points. The interpolation method is an existing technology; for example, it can be implemented by calculating the average coordinates of points surrounding the missing point, or by taking the intersection of lines connecting two points above and below, or two points to the left and right.

[0124] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for fabricating a metal antenna based on visual recognition correction, characterized in that the steps include... include: S10: Set positioning marks on the front side of the substrate; Dot matrix is ​​printed on the back of the substrate, and the spacing between adjacent dots in the dot matrix is ​​equal; Apply adhesive to the front side of the substrate; S20: The substrate is dried and then laminated with metal foil; S30: Take photos of the front and back of the substrate simultaneously, and determine the origin and coordinate system of the back image based on the positioning mark information in the front image of the substrate. S40: Obtain the designed position of each point in the dot matrix and the measured position of each point in the dot matrix in the back image. Based on the positional deviation between the designed position and the measured position of each point, establish a correction model and verify the error of the correction model. S50: The coordinates of the original cutting position of the antenna are used as input. The coordinates of the corrected position of the antenna after the substrate deformation are calculated by the correction model. Then, the antenna is cut according to the corrected position to obtain the final metal antenna product. In step S40, the specific algorithm for correcting the model is as follows: Number the points in the dot matrix from left to right and then from top to bottom as 1, 2...N. Let the designed position of the i-th point in the dot matrix be (xi, yi) and the measured position be (xi', yi'). Construct an N×10 mutual influence matrix M, where the elements in the i-th row of the mutual influence matrix M are: 1, xi, yi, xi 2 , xi·yi, yi 2 xi 3 xi 2 ·yi、xi·yi 2 yi 3 ; Construct observation vectors Dx and Dy; Dx=[x1’,x2’,……,xN’] T ; Dy=[y1',y2',……,yN'] T ; The correction model is set with two correction vectors, Tx and Ty: Tx=[α1,α2,……,α10] T =(M T ·M+Λx) -1 ·M T ·Dx; Ty=[β1,β2,……,β10] T =(M T ·M+Λy) -1 ·M T ·Dy; The initial values ​​of Λx and Λy are both diagonal matrices with a set value λ0 on the diagonal and 0 for the rest. The error verification model is as follows: The coordinates of the designed positions of each point in the dot matrix are used to calculate the correction reference position through the correction model. The coordinates of the correction reference position of the i-th point are denoted as (T(xi),T(yi)). Calculate the mean square error (MSE) of the correction reference position and the measured position at each point within the lattice: ； When the MES is less than the set threshold, the error verification of the correction model is passed; otherwise, the elements in Λx and Λy are modified and the correction model is rebuilt until the error verification is passed. In step S50, the original cutting position of the j-th antenna is denoted as (xj, yj), and the corrected position is denoted as (T(xj), T(yj)). T(xj)=[1, xj, yj, xj 2 ,xj·yj,yj 2 ,xj 3 ,xj 2 ·yj, xj·yj 2 ,yj 3 ]·Tx; T(yj)=[1、xj、yj、xj 2 、xj·yy、yyy 2 、xj 3 、xj 2 ·yj、xj·yj 2 、y 3 ]·Ty: In step S50, a cutting border is set on the substrate, and multiple equally spaced design border points are set on the cutting border. All design border points are calculated through the correction model to form corresponding correction border points. Then, all correction border points are connected in sequence to form a cutting area. After calculating the antenna's correction position, determine whether each antenna region is within the cutting area and does not overlap with other antenna regions. If not, translate the antenna's correction position until the requirements are met. The specific translation method of the correction position is as follows: If the antenna area exceeds the cutting area, select the two correction frame points closest to the antenna area and move the correction position along the vertical direction of the line connecting the two correction frame points. If two antenna regions overlap, compare the overlap lengths of the two antenna regions in the horizontal and vertical directions, select the minimum overlap length and the direction in which the minimum overlap length is located, and simultaneously translate the two antenna regions away from each other.

2. The method for fabricating a metal antenna based on visual recognition correction according to claim 1, characterized in that, In step S10, the positioning marks on the front side of the substrate include two rows of dots arranged in parallel, with the same number of marking points in each row. In step S30, the centroids of the distribution of each row of markers are calculated and denoted as points C1 and C2, respectively. Then, the center points of points C1 and C2 are denoted as the origin. The fitted line of each row of markers and the average slope kavg of the two fitted lines are calculated. The line with a slope of kavg and passing through the origin is used as the x-axis to form a coordinate system.

3. The method for fabricating a metal antenna based on visual recognition correction according to claim 2, characterized in that, In step S30, the points in the first row are denoted as A1(a1x,a1y), A2(a2x,a2y), ..., An(anx,any); and the points in the second row are denoted as B1(b1x,b1y), B2(b2x,b2y), ..., Bn(bnx,bny). Let C1(c1x,c1y) and C2(c2x,c2y) be the points; Then we have c1x = (a1x + a2x + a3x + a4x) / 4; c1y=(a1y+a2y+a3y+a4y) / 4; c2x=(b1x+b2x+b3x+b4x) / 4; c2y=(b1y+b2y+b3y+b4y) / 4.

4. The method for fabricating a metal antenna based on visual recognition correction according to claim 1, characterized in that, The elements in Λx and Λy are modified as follows: Calculate the average absolute difference of the abscissa Δxavg, the average absolute difference of the ordinate Δyavg, and the average distance Δdavg between the correction reference position and the measured position of all points in the dot matrix. For the element in row p and column p in Λx and Λy, update it to λxp and λyp; λxp=λ0+k0·mxp·exp(-△avg / △xavg); λyp=λ0+k0·myp·exp(-△avg / △yavg); Where k0 is the set base multiplier; mxp is the power of xi in the p-th column of the mutual influence matrix M; myp is the power of yi in the p-th column of the mutual influence matrix M.

5. The method for fabricating a metal antenna based on visual recognition correction according to claim 4, characterized in that, After each calculation of all λxp and λyp, obtain the maximum and minimum values ​​of α1~α10 and β1~β10. If the maximum or minimum value of α1~α10 exceeds the set range, adjust all λxp so that α1~α10 can all fall within the set range while maintaining the proportions between them. If the maximum or minimum value of β1~β10 exceeds the set range, adjust all λxp so that β1~β10 can all fall within the set range while maintaining the proportions between them.

6. The method for fabricating a metal antenna based on visual recognition correction according to claim 1, characterized in that, In step S40, before calculating the positional deviation, the number of points in the dot matrix of the back image is counted. If the counted number is different from the designed number, the missing points are identified, and the missing points are determined by interpolation using the neighboring points of the missing points.

Images