Double-cutter cutting precision optimization method and device based on image alignment, equipment and medium

Abstract

The invention discloses a double-cutter cutting precision optimization method and device based on image alignment, equipment and a medium, and the method comprises the following steps: obtaining a reference point data set, building a coordinate system based on the reference point data set, shooting a first Mark point of a cutting surface of a substrate through an industrial camera, extracting the coordinates of the first Mark point of cutter feeding and cutter discharging, and obtaining a second Mark point of the cutting surface of the substrate; calculating a straight line according to the first Mark point coordinates, and determining an initial cutting direction angle; constructing a cutting path model, determining a theoretical cutting path according to an initial cutting direction angle under actual deformation of the material, photographing a second Mark point through an industrial camera according to the theoretical cutting path, extracting coordinates of the second Mark point, measuring an actual cutting path, and calculating deformation deviation according to the coordinates of the first Mark point and the coordinates of the second Mark point; a traditional static model is replaced, complex material deformation can be dynamically predicted, and the cutting precision is remarkably improved. According to the method, the Y-direction deviation is reduced to the maximum extent through collaborative optimization and a dynamic adjustment algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of dicing machine cutting, and specifically to a double - knife cutting accuracy optimization method, device, equipment and medium based on image alignment. Background Art

[0002] When a dicing machine cuts an LED substrate, before die bonding and encapsulation heating, the substrate size is regular, and there are special image points of in - and - out knife marks on both the horizontal and vertical sides. The ideal cutting effect is that the knife marks are cut at the center positions of the in - and - out knife marks. However, after heating and encapsulation, the deformation of the two - side marks is inconsistent due to heating and encapsulation. If the first - knife in - and - out knife mark center is straightened and then the last - knife in - and - out knife mark center is aligned, it will be found that the cutting knife marks of the last knife and the first knife are not parallel and there is an angle. If the first - knife straightening angle is used for cutting and cutting is carried out one knife at a time according to the spacing, the deviation of the last knife is relatively large in the end.

[0003] During the process of a dicing machine cutting an LED substrate, before cutting, the substrate size is regular, and there are special image points of in - and - out knife marks on both the horizontal and vertical sides. During cutting, the center of the 1 and 2 point marks is straightened parallel to the cutting blade, and the 3 and 4 points are aligned according to the set spacing and number of knives.

[0004] The existing cutting method is as follows: There are a total of 29 knives on the current cutting surface, and the spacing is 3.7 mm. First, image alignment is performed on the 1 and 2 points, and their center positions are straightened. The straight line is made parallel to the Y - axis main - axis blade through the platform rotation axis, so as to determine the coordinate positions of the 1 and 2 points. Then, according to the set spacing, the theoretical position of the 3rd point is the Y position of the 1st point + 3.7×(29 - This method can ensure uniform spacing and reduce the deviation of the last - knife cutting. However, when the deformation of the 3 and 4 points is inconsistent, the accuracy of cutting the 3 and 4 points knife still does not reach the best effect. The deformation of the material edge is mostly linear deformation on both sides.

[0005] Regarding the cutting situation of the last knife (the 29th knife) during linear deformation: If the deformation directions of the 3 and 4 points are the same as the theoretical value and the magnitudes are the same, after image alignment of the 3 and 4 points, the final deviation is compensated to each knife, and there will be no position deviation in the last knife; if the deformation directions of the 3 and 4 points are opposite to the theoretical value, resulting in an angle between the straight line of the 3 and 4 points and the theoretical cutting path, at this time, the average value of the deviations of the 3 and 4 points in the Y - direction is calculated, and the accuracy of the actual cutting path and the straight line of the 3 and 4 points in the end is the difference between the Y - directions of the two points divided by 2. Although this accuracy is higher than that according to the theoretical position, for materials with high - precision requirements, the last knife still cannot meet the accuracy requirements.

[0006] For these problems, we propose a double - knife cutting accuracy optimization method, device, equipment and medium based on image alignment. Summary of the Invention

[0007] To solve one of the above technical problems, a method, device, equipment and medium for optimizing the double - knife cutting accuracy based on image alignment are provided. Aiming at the problem that the traditional method cannot fully cope with the non - linear deformation at both ends of the material during the double - knife cutting process, by introducing high - precision mathematical modeling, dynamic iterative optimization algorithm and multi - dimensional space compensation strategy, the cutting accuracy is fundamentally improved, the waste loss rate is reduced, and the above problems are solved.

[0008] To achieve the above objectives, the technical invention adopted in the present invention is as follows:

[0009] In the first aspect: A method for optimizing the double - knife cutting accuracy based on image alignment, comprising the following steps:

[0010] Obtain a reference point data set, establish a coordinate system based on the reference point data set, capture the first Mark point on the cutting surface of the substrate through an industrial camera, extract the coordinates of the first Mark points for tool - in and tool - out, calculate a straight line according to the coordinates of the first Mark points, and determine the initial cutting direction angle;

[0011] Construct a cutting path model, determine the theoretical cutting path under the actual deformation of the material according to the initial cutting direction angle, take pictures of the second Mark points through an industrial camera according to the theoretical cutting path and extract the coordinates of the second Mark points, measure the actual cutting path, and calculate the deformation deviation according to the coordinates of the first Mark points and the second Mark points;

[0012] Based on the calculated deformation deviation data, construct a non - linear deformation model, divide the actual cutting path into several sub - intervals according to the non - linear deformation model, calculate the tool - in and tool - out deviations based on the coordinates of the first Mark points and the second Mark points as a reference, and allocate the Y - direction compensation value for each cutting position;

[0013] According to the blade rotation angle and the scaling factor, construct an affine transformation matrix, map the coordinates of the theoretical cutting path points to the actual cutting path, and update the tool - in and tool - out positions;

[0014] Based on the updated tool - in and tool - out positions, when cutting every two knives, calculate the included angle between the current knife and the previous knife, adjust the blade rotation angle, and adjust the position of the platform in real time according to the coordinates after affine transformation to ensure that the blade path coincides completely with the cutting path.

[0015] Preferably: The industrial camera captures the first Mark point on the cutting surface of the substrate, specifically including:

[0016] The multi - lens array camera synchronously collects data, and calculates the three - dimensional coordinates of the first Mark point in response to the stereo vision algorithm. The first Mark point includes M1(x1, y1, z1) and M2(x2, y2, z2);

[0017] Image preprocessing: According to the dynamic range imaging technology and convolutional neural network, improve the recognition rate of the first Mark point, enhance the image features, and filter out interference.

[0018] Preferably: Construct a non-linear deformation model based on the calculated deformation deviation data, specifically including:

[0019] Obtain historical cutting data, and input the historical cutting data and the position of the second Mark point obtained in real time into the non-linear deformation model, where the second Mark point includes M3(x3, y3, z3) and M4(x4, y4, z4);

[0020] Combine the image features and material properties according to the multi-modal model of the Transformer architecture, and output the predicted value of the global non-linear deformation during the cutting process.

[0021] Preferably: The predicted value F(x, y) of the global non-linear deformation, the specific formula is:

[0022] Loss function:

[0023]

[0024] Among them, N represents the number of samples;

[0025] F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0026] F 实际 (x i ,y i ) is the actual deformation value at the coordinate (x i ,y i );

[0027] Predicted output, output the deformation compensation value of each tool path point:

[0028] δy i =F 预测 (x i ,y i )-Y 理论,i ;

[0029] Among them, F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0030] Y 理论,i is at the coordinate (xi , y i ) theoretical deformation value at

[0031] Preferably: calculating the feed-in and feed-out deviation based on the coordinates of the first Mark point and the second Mark point, and allocating the Y-direction compensation value for each cutting position, specifically including:

[0032] Calculating the feed-in and feed-out deviation in combination with a collaborative optimization function, and performing real-time deviation iterative compensation;

[0033] Updating the blade cutting path based on the deformation value predicted by the non-linear deformation model;

[0034] Finding the optimal rotation angle and position compensation value according to the Newton iteration method to optimize the deviation value of each cut;

[0035] Adjusting the coordinates of the feed-in and feed-out points according to the optimized compensation value and synchronizing them to the numerical control machine tool control system.

[0036] Preferably: the collaborative optimization function, the specific formula is:

[0037]

[0038] where, δy 1,i and δy 2,i are the Y-direction deviations of the feed-in and feed-out, Δθ i is the blade rotation adjustment amount, and λ is the weight coefficient.

[0039] Preferably: the method further includes digital twin verification and online feedback, specifically including:

[0040] Constructing a virtual model of material deformation and cutting path, simulating the cutting process and evaluating the accuracy, and optimizing the cutting path;

[0041] Performing virtual cutting based on the optimized path, calculating the maximum deviation of each cut and recording the accuracy distribution;

[0042] Comparing the simulation results with the actual cutting results in real time, updating the model, and optimizing the cutting path of the next cut.

[0043] Second aspect: A double-blade cutting accuracy optimization device based on image alignment, including:

[0044] A dataset acquisition module, configured to acquire a dataset of reference points, establish a coordinate system based on the dataset of reference points, photograph the first Mark point on the cutting surface of the substrate through an industrial camera, extract the coordinates of the first Mark points of the feed-in and feed-out, and calculate a straight line according to the coordinates of the first Mark points to determine the initial cutting direction angle;

[0045] The deformation deviation calculation module is used to construct a cutting path model, determine the theoretical cutting path according to the initial cutting direction angle under the actual deformation of the material, take a photo of the second Mark point through an industrial camera according to the theoretical cutting path and extract the coordinates of the second Mark point, measure the actual cutting path, and calculate the deformation deviation according to the coordinates of the first Mark point and the coordinates of the second Mark point;

[0046] The compensation module is used to construct a non-linear deformation model based on the calculated deformation deviation data, divide the actual cutting path into several sub-intervals according to the non-linear deformation model, calculate the in-and-out tool deviation based on the coordinates of the first Mark point and the coordinates of the second Mark point, and allocate the compensation value in the Y direction for each cutting position;

[0047] The cutting path update module is used to construct an affine transformation matrix according to the blade rotation angle and the scaling factor, map the coordinates of the theoretical cutting path points to the actual cutting path, and update the in-feed and out-feed positions;

[0048] The adjustment module is used to calculate the angle between the current tool and the previous tool when cutting every two tools based on the updated in-feed and out-feed positions, adjust the blade rotation angle, and adjust the position of the platform in real time according to the coordinates after affine transformation to ensure that the blade path coincides exactly with the cutting path.

[0049] In a third aspect: A computer device, comprising:

[0050] A processor;

[0051] A memory for storing executable instructions;

[0052] Wherein, the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the double-blade cutting accuracy optimization method based on image alignment.

[0053] In a fourth aspect: A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor is caused to implement the double-blade cutting accuracy optimization method based on image alignment.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] 1. The present invention replaces the traditional static model, can dynamically predict the deformation of complex materials, and significantly improves the cutting accuracy.

[0056] 2. The present invention minimizes the deviation in the Y direction through collaborative optimization and dynamic adjustment algorithms. The combination of affine transformation and local correction: the combination of overall and local optimization solves complex non-linear deformation.

[0057] 3. The present invention improves the cutting path planning efficiency and reliability through virtual simulation and actual result feedback.

[0058] 4. The present invention combines machine learning, image processing and numerical control technology, with the accuracy increased by about 1.5 times and the material waste rate reduced by 30%. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flowchart of the double-knife cutting accuracy optimization method based on image alignment of the present invention;

[0060] Figure 2 is a module diagram of the double-knife cutting accuracy optimization system based on image alignment of the present invention.

[0061] Figure 3 is a schematic structural diagram of a computer device of the present invention;

[0062] In the figure, 10 is a computer device; 1002 is a processor; 1004 is a memory; 1006 is a transmission device. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations.

[0064] Embodiment 1:

[0065] Referring to Figure 1 as shown, a double-knife cutting accuracy optimization method based on image alignment includes the following steps:

[0066] Step 10: Obtain a reference point data set, establish a coordinate system based on the reference point data set, capture the first Mark point on the cutting surface of the substrate through an industrial camera, extract the coordinates of the first Mark point for tool entry and tool exit, and calculate a straight line according to the coordinates of the first Mark point to determine the initial cutting direction angle;

[0067] The industrial camera captures the first Mark point on the cutting surface of the substrate, specifically including:

[0068] The multi-lens array camera synchronously collects data, and calculates the three-dimensional coordinates of the first Mark point in response to a stereo vision algorithm. The first Mark point includes M1(x1, y1, z1) and M2(x2, y2, z2);

[0069] Image preprocessing: According to the dynamic range imaging technology and convolutional neural network, improve the recognition rate of the first Mark point and enhance the image features, and filter out interference;

[0070] Specifically, camera calibration and enhancement: Use a multi-lens array camera to synchronously collect data, and combine with a stereo vision algorithm to calculate the three-dimensional coordinates of the first Mark point;

[0071] M1(x1, y1, z1) and M2(x2, y2, z2).

[0072] Apply high-dynamic range imaging (HDR) technology to improve the recognition rate of the first Mark point under complex lighting conditions. Use a convolutional neural network (CNN) to enhance the features of the image and filter out interference.

[0073] Material surface state modeling: Use a surface laser scanning system to obtain the material height distribution (i.e., the deformation amount H(x, y)) of the cutting area and fuse it with the image data.

[0074] Construct a Mark point offset vector:

[0075] ΔM = [Δx, Δy, Δz] = [x2 - x1, y2 - y1, z2 - z1];

[0076] Determine the initial cutting direction θ0:

[0077]

[0078] Where: (x1, y1) and (x2, y2) are the initial positions of the Mark points; θ0 is the initial angle between the blade and the material reference line.

[0079] Through multi-channel data fusion, obtain high-precision information on the position of the first Mark point and the material surface state. Step 20, construct a cutting path model, determine the theoretical cutting path according to the initial cutting direction angle under the actual deformation of the material, take pictures of the second Mark point through an industrial camera according to the theoretical cutting path and extract the coordinates of the second Mark point, measure the actual cutting path, and calculate the deformation deviation according to the coordinates of the first Mark point and the second Mark point;

[0080] Calculate the deformation deviation: deviation value δ i = Y 实际,i - Y 理论,i ;

[0081] Establish a deformation distribution function: Based on the multi-point deviations measured and calculated, fit a non-linear deformation model:

[0082] f(y) = αy 2 + βy + γ;

[0083] Where, f(y) is the deformation function, and α, β, γ are fitting parameters.

[0084] Step 30: Based on the calculated deformation deviation data, construct a non-linear deformation model. According to the non-linear deformation model, divide the actual cutting path into several sub-intervals, calculate the approach and departure deviations based on the coordinates of the first Mark point and the second Mark point, and assign the Y-direction compensation value for each cutting position.

[0085] The construction of the non-linear deformation model based on the calculated deformation deviation data specifically includes:

[0086] Obtain historical cutting data, and input the historical cutting data and the position of the second Mark point obtained in real time into the non-linear deformation model, where the second Mark point includes M3(x3, y3, z3) and M4(x4, y4, z4);

[0087] Combine the image features and material properties according to the multi-modal model of the Transformer architecture, and output the predicted value of the global non-linear deformation during the cutting process;

[0088] The specific formula for the predicted value F(x, y) of the global non-linear deformation is:

[0089] Loss function:

[0090]

[0091] where N represents the number of samples;

[0092] F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0093] F 实际 (x i ,y i ) is the actual deformation value at the coordinate (x i ,y i );

[0094] Predict the output and output the deformation compensation value of each tool path point:

[0095] δy i =F 预测 (x i ,y i )-Y 理论,i ;

[0096] where F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0097] Y 理论,i is the theoretical deformation value at the coordinate (x i , y i ).

[0098] Calculating the feed-in and feed-out deviation based on the first Mark point coordinate and the second Mark point coordinate, and allocating the Y-direction compensation value for each cutting position specifically includes:

[0099] Calculating the feed-in and feed-out deviation in combination with the collaborative optimization function and performing real-time deviation iterative compensation;

[0100] Updating the blade cutting path based on the deformation value predicted by the non-linear deformation model;

[0101] Finding the optimal rotation angle and position compensation value according to the Newton iteration method to optimize the deviation value of each cut;

[0102] Adjusting the coordinates of the feed-in and feed-out points according to the optimized compensation value and synchronizing them to the numerical control machine tool control system;

[0103] The specific formula of the collaborative optimization function is:

[0104] ;

[0106] where δy 1,i and δy 2,i are the Y-direction deviations of the feed-in and feed-out, Δθ is the blade rotation adjustment amount, and λ is the weight coefficient.

[0107] Input the historical cutting data (including material type, size, deviation distribution) and the Mark point position and surface height distribution obtained in real time into the deep learning model.

[0108] Model training: Using a multi-modal model based on the Transformer architecture, combining image features and material properties to output the predicted value F(x, y) of the global non-linear deformation during cutting;

[0109] Using the deep learning model to predict the global deformation of the material to provide a more accurate basis for cutting path compensation.

[0110] Updating the blade cutting path using the deformation value predicted by the model:

[0111] Y 新,i = Y 理论,i + δy i .

[0112] Step 40: Construct an affine transformation matrix according to the blade rotation angle and the scaling factor, map the coordinates of the theoretical cutting path points to the actual cutting path, and update the feed-in and feed-out positions;

[0113] By rotating the angle θ 偏 and the scaling factor λ, an affine transformation matrix is constructed as follows:

[0114]

[0115] where Δx and Δy are the displacement compensation values for tool feed and tool retraction.

[0116] Map the coordinates (x, y) of the theoretical cutting path points to the actual cutting path:

[0117]

[0118] Update the tool feed and tool retraction positions:

[0119] For all tool numbers, the adjusted tool feed position is (x 进刀 , y′ 进刀 ), and the tool retraction position is (x 出刀, y′ 出刀 ).

[0120] Local fine-tuning compensation:

[0121] Perform local quadratic fitting correction on the cutting path for each tool:

[0122] Y 微调,i = Y 补偿,i + α i · (x i - x 中点 );

[0123] where α i is the local adjustment coefficient.

[0124] Step 50: Based on the updated tool feed and tool retraction positions, when cutting every two tools, calculate the angle between the current tool and the previous tool, adjust the blade rotation angle, and according to the coordinates after affine transformation, adjust the position of the platform in real time to ensure that the blade path coincides exactly with the cutting path;

[0125] When cutting every two tools, calculate the angle between the current tool and the previous tool, and adjust the rotation angle T:

[0126] T = T + Δθ 优化 ;

[0127] According to the coordinates after affine transformation, adjust the positions of the X, Y, and T axes of the platform in real time to ensure that the blade path coincides exactly with the cutting path.

[0128] Specifically, the method further includes digital twin verification and online feedback, which specifically include:

[0129] Construct a virtual model of the deformation and cutting path of the building material, simulate the cutting process, evaluate the accuracy, and optimize the cutting path;

[0130] Based on the optimized path, perform virtual cutting, calculate the maximum deviation of each cut, and record the accuracy distribution;

[0131] Compare the simulation results with the actual cutting results in real time, update the model, and optimize the cutting path for the next cut;

[0132] Specifically, the present invention solves the problem that traditional static models are difficult to handle complex deformations, and significantly improves the prediction accuracy.

[0133] The present invention dynamically adjusts the tool entry and exit paths, improving the overall accuracy and stability of cutting.

[0134] The present invention evaluates the cutting effect in advance through virtual simulation, reducing the cost and risk of on-site adjustment.

[0135] The accuracy of the present invention is improved by more than 1.5 times, and the waste rate is reduced by 30%. It is particularly suitable for precision cutting of large-sized and highly deformed materials.

[0136] Embodiment 2

[0137] To solve the above technical problems, based on Embodiment 1, another technical solution adopted in this application is: a double-tool cutting accuracy optimization method based on image alignment, including the following steps:

[0138] Cooperatively acquire the accurate three-dimensional coordinates of the Mark points through multiple sensors, and simultaneously model the deformation amount of the material surface to provide basic data for subsequent optimization.

[0139] Step 1: Obtain camera calibration and images: Use a multi-camera array to ensure coverage of the cutting area: The pixel points captured by each camera are mapped to the world coordinate system through the calibration matrix K, and the mapping formula is:

[0140] P 世界 = K · P pixel;

[0141] Where: K is the camera calibration matrix, including focal length, optical center offset, and perspective distortion correction parameters;

[0142] P 像素 =(u, v) represents the image pixel coordinates; P 世界 =(x, y, z) is the three-dimensional coordinate of the first Mark point. Determine the initial position of the first Mark point:

[0143] M1 = (x1, y1, z1), M2 = (x2, y2, z2);

[0144] Image enhancement: Use high dynamic range imaging (HDR) to improve the ability to identify details.

[0145] The convolutional neural network (CNN) extracts the Mark point feature map, filters out background interference, and outputs the center position of the Mark point.

[0146] Material surface deformation modeling:

[0147] Combined with laser scanning technology, the surface deformation of the material is obtained: H(x, y), which represents the height of the surface deformation of the material in the two-dimensional plane (x, y).

[0148] Initial blade cutting angle calculation: Calculate the offset vector between Mark points:

[0149] ΔM = (x2 - x1, y2 - y1, z2 - z1);

[0150] Initial cutting direction angle:

[0151]

[0152] where: (x1, y1) and (x2, y2) are the initial positions of the Mark points; θ0 is the initial angle between the blade and the material reference line.

[0153] Step 2: Data input, the input features include:

[0154] Use a deep learning model to predict the non-linear deformation of the material and provide a basis for dynamic cutting compensation.

[0155] Historical cutting data: cutting material type, thickness, size, deformation distribution;

[0156] Real-time collected data: Mark point position, surface deformation H(x, y).

[0157] Deep learning model architecture:

[0158] Use a multi-modal model based on Transformer:

[0159] F(x, y) = Model([historical features, H(x, y)]);

[0160] where: F(x, y): Predicted actual deformation of the material at position (x, y);

[0161] Model: Transformer model, which combines historical and real-time features for prediction.

[0162] Optimize the prediction error of the model:

[0163]

[0164] where N represents the number of samples;

[0165] F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0166] F 实际 (x i ,y i ) is the actual deformation value at the coordinate (x i ,y i );

[0167] Predicted output, output the deformation compensation value of each tool path point:

[0168] δy i = F 预测 (x i ,y i ) - Y 理论,i ;

[0169] where F 预测 (x i ,y i ) is the predicted deformation value at the coordinate (x i ,y i );

[0170] Y 理论,i is the theoretical deformation value at the coordinate (x i ,y i );

[0171] Output:

[0172] Output the deformation compensation value of each tool path point:

[0173] δy i = F(x i ,y i ) - Y 理论,i ;

[0174] Step 3: Collaborative optimization objective: Define the optimization objective for the deviation of the double - tool in - and - out positions:

[0175]

[0176] Among them:

[0177] δy 1,i , δy 2,i : The Y - deviation of tool 1 and tool 2 at position i;

[0178] Δθ: The adjustment amount of the blade rotation angle;

[0179] λ: The weight parameter, balancing accuracy and adjustment amplitude.

[0180] Real-time path adjustment:

[0181] Update the Y coordinate of the path point:

[0182] Y 新,i = Y 理论,i + δy i ;

[0183] Dynamic optimization algorithm: Use the gradient descent method for iterative adjustment:

[0184]

[0185] where: -η: learning rate; Gradient of the loss function with respect to the angle.

[0186] According to the deformation prediction result, dynamically adjust the blade path and rotation angle.

[0187] Step 4: Affine transformation model:

[0188] Adjust the blade direction through global affine transformation and perform local non-linear correction at the same time.

[0189] Construct a two-dimensional affine transformation matrix:

[0190]

[0191] where:

[0192] θ: rotation angle;

[0193] Δx, Δy: translation compensation amount..

[0194] Local fine-tuning:

[0195] Perform quadratic fitting correction on the path:

[0196] Y 微调,i = Y 补偿,i + α i ·(x i - x 中点 );

[0197] where:

[0198] α i : local adjustment coefficient;

[0199] x 中点 : midpoint position of the blade.

[0200] Step 5: Use digital twin technology to simulate and verify the optimization scheme, and continuously iterate and improve the cutting path.

[0201] Construct a virtual environment that includes the material deformation characteristics, blade path, and actual cutting parameters.

[0202] Feedback optimization:

[0203] Compare the simulation and actual cutting results:

[0204] δy i = Y 仿真,i - Y 实际,i ;

[0205] Update the compensation model to improve the cutting accuracy of the next cut.

[0206] The present invention replaces the traditional static model, can dynamically predict the deformation of complex materials, and significantly improves the cutting accuracy.

[0207] The present invention minimizes the deviation in the Y direction through collaborative optimization and dynamic adjustment algorithms. The combination of affine transformation and local correction: the combination of overall and local optimization solves complex non-linear deformation.

[0208] The present invention improves the efficiency and reliability of cutting path planning through virtual simulation and actual result feedback.

[0209] The present invention combines machine learning, image processing, and numerical control technology, with the accuracy increased by about 1.5 times and the material waste rate reduced by 30%.

[0210] Embodiment III

[0211] As Figure 2 shown, to solve the above technical problems, based on Embodiment I, another technical solution adopted by the present application is: a double-blade cutting accuracy optimization device based on image alignment, including the following steps:

[0212] An acquisition dataset module, used to acquire a reference point dataset, establish a coordinate system based on the reference point dataset, photograph the first Mark point on the cutting surface of the substrate through an industrial camera, extract the coordinates of the first Mark point for tool entry and tool exit, calculate a straight line according to the coordinates of the first Mark point, and determine the direction angle of the initial cut;

[0213] A calculation deformation deviation module, used to construct a cutting path model, determine the theoretical cutting path under the actual deformation of the material according to the direction angle of the initial cut, photograph the second Mark point through an industrial camera according to the theoretical cutting path and extract the coordinates of the second Mark point, measure the actual cutting path, and calculate the deformation deviation according to the coordinates of the first Mark point and the coordinates of the second Mark point;

[0214] A compensation module, which is used to construct a non - linear deformation model based on the calculated deformation deviation data, divide the actual cutting path into several sub - intervals according to the non - linear deformation model, calculate the feed - in and feed - out deviations based on the coordinates of the first Mark point and the second Mark point, and allocate the compensation value in the Y - direction for each cutting position.

[0215] An updated cutting path module, which is used to construct an affine transformation matrix according to the blade rotation angle and the scaling factor, map the coordinates of the theoretical cutting path points to the actual cutting path, and update the feed - in and feed - out positions.

[0216] An adjustment module, which is used to calculate the angle between the current cutting and the previous cutting when cutting every two cuts based on the updated feed - in and feed - out positions, adjust the blade rotation angle, and adjust the position of the platform in real - time according to the coordinates after affine transformation to ensure that the blade path coincides completely with the cutting path.

[0217] The foregoing Figure 1 All kinds of variation methods and specific examples of a double - blade cutting accuracy optimization method based on image alignment in the first embodiment are equally applicable to a double - blade cutting accuracy optimization system based on image alignment in this embodiment. Through the detailed description of a double - blade cutting accuracy optimization method based on image alignment above, those skilled in the art can clearly know the implementation method of a double - blade cutting accuracy optimization system based on image alignment in this embodiment. Therefore, for the sake of simplicity of the specification, it will not be described in detail here.

[0218] Embodiment 4

[0219] An embodiment of the present application provides a computer device, which includes a processor and a memory. At least one instruction or at least one program segment is stored in the memory, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement a double - blade cutting accuracy optimization method based on image alignment as provided in the above - mentioned method embodiment.

[0220] Figure 3 The figure shows a schematic hardware structure diagram of a device for implementing a double - blade cutting accuracy optimization method based on image alignment provided in an embodiment of the present application. The device can participate in constituting or include the device or system provided in the embodiment of the present application. As Figure 3 shown, the computer device 10 may include one or more processors 1002 (the processor may include, but is not limited to, a processing device such as a micro - processor MCU or a programmable logic device FPGA), a memory 1004 for storing data, and a transmission device 1006 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand,Figure 3 The structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computer device 10 may also include more or fewer components than those shown in Figure 3 , or have a different configuration from that shown in Figure 3 .

[0221] It should be noted that the above one or more processors and / or other data processing circuits can generally be referred to as "data processing circuits" in this article. The data processing circuit can be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit can be a single independent processing module, or be incorporated in whole or in part into any one of the other elements in the computer device 10 (or mobile device). As involved in the embodiments of the present application, the data processing circuit is a kind of processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0222] The memory 1004 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to a double-knife cutting precision optimization method based on image alignment in the embodiments of the present application. The processor runs the software programs and modules stored in the memory 1004 to perform various functional applications and data processing, that is, to implement the above-mentioned method. The memory 1004 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 1004 may further include a memory remotely set relative to the processor, and these remote memories can be connected to the computer device 10 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0223] The transmission device 1006 is used to receive or send data via a network. Specific examples of the above network may include a wireless network provided by the communication provider of the computer device 10. In one instance, the transmission device 1006 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission device 1006 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0224] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables the user to interact with the user interface of the computer device 10 (or mobile device).

[0225] Embodiment Five

[0226] The embodiment of the present application further provides a computer-readable storage medium, which can be disposed in the server to store at least one instruction or at least one segment of program related to an optimization method for double-knife cutting accuracy based on image alignment in the method embodiment. The at least one instruction or the at least one segment of program is loaded and executed by the processor to implement the optimization method for double-knife cutting accuracy based on image alignment provided in the above method embodiment.

[0227] Optionally, in this embodiment, the above storage medium may be located in at least one of multiple network servers in the computer network. Optionally, in this embodiment, the above storage medium may include but is not limited to: various media that can store program codes such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs.

[0228] Embodiment Six

[0229] The embodiment of the present invention further provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the optimization method for double-knife cutting accuracy based on image alignment provided in the above various optional implementation manners.

[0230] It should be noted that: the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of the present application have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims may be executed in a different order from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain implementation manners, multi-task processing and parallel processing are also possible or may be advantageous.

[0231] Each embodiment in the present application is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device, equipment, and storage medium, since they are basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0232] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware or by a program instructing relevant hardware. The program can be stored in a computer-readable storage medium, and the storage medium mentioned above can be a read-only memory, a disk, an optical disc, etc.

[0233] Inspired by the above ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of the present invention. The technical scope of the present invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

[0234] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various changes or substitutions, and these should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A double - knife cutting precision optimization method based on image alignment, characterized in that, It includes the following steps: Obtain a reference point data set, establish a coordinate system based on the reference point data set, photograph the first Mark point on the cutting surface of the substrate through an industrial camera, extract the coordinates of the first Mark points for tool entry and tool exit, calculate a straight line based on the coordinates of the first Mark points, and determine the initial cutting direction angle; Construct a cutting path model, determine the theoretical cutting path under the actual deformation of the material according to the initial cutting direction angle, photograph the second Mark point through an industrial camera according to the theoretical cutting path and extract the coordinates of the second Mark point, measure the actual cutting path, and calculate the deformation deviation according to the coordinates of the first Mark point and the coordinates of the second Mark point; Based on the calculated deformation deviation data, construct a non-linear deformation model, divide the actual cutting path into several sub-intervals according to the non-linear deformation model, calculate the tool entry and exit deviation based on the coordinates of the first Mark point and the coordinates of the second Mark point, and allocate the Y-direction compensation value for each cutting position; Construct an affine transformation matrix according to the blade rotation angle and the scaling factor, map the coordinates of the theoretical cutting path points to the actual cutting path, and update the tool entry and tool exit positions; Based on the updated tool entry and tool exit positions, when cutting every two cuts, calculate the included angle between the current cut and the previous cut, adjust the blade rotation angle, and adjust the position of the platform in real time according to the coordinates after affine transformation to ensure that the blade path coincides completely with the cutting path.

2. The method according to claim 1, wherein: The industrial camera photographs the first Mark point on the cutting surface of the substrate, which specifically includes: Multiple-lens array cameras synchronously collect data, and calculate the three-dimensional coordinates of the first Mark point in response to a stereo vision algorithm. The first Mark point includes M1(x1, y1, z1) and M2(x2, y2, z2); Image preprocessing: According to the dynamic range imaging technology and convolutional neural network, improve the recognition rate of the first Mark point and enhance the image features to filter out interference.

3. The method according to claim 1, characterized in that: The constructing of the non-linear deformation model based on the calculated deformation deviation data specifically includes: Obtain historical cutting data, and input the historical cutting data and the position of the second Mark point obtained in real time into the non-linear deformation model. The second Mark point includes M3(x3, y3, z3) and M4(x4, y4, z4); Combine the image features and material properties according to the multi-modal model based on the Transformer architecture, and output the predicted value of the global non-linear deformation during the cutting process.

4. The method according to claim 3, wherein: The predicted value F(x, y) of the global non-linear deformation, the specific formula is: Loss function: where N represents the number of samples; F 预测 (x i ,y i ) is the predicted deformation value at the coordinates (x i ,y i ); F 实际 (x i , y i ) is the actual deformation value at the coordinate (x i , y i ); Predictive output, output the deformation compensation value of each tool path point: δy i = F 预测 (x i , y i ) - Y 理论,i ; Among them, F 预测 (x i , y i ) is the predicted deformation value at the coordinate (x i , y i ); Y 理论,i is the theoretical deformation value at the coordinates (x i , y i ).

5. The method according to claim 1, wherein: The calculating of the tool entry and exit deviation based on the coordinates of the first Mark point and the coordinates of the second Mark point, and the allocation of the Y-direction compensation value for each cutting position specifically includes: Calculate the tool entry and exit deviation in combination with a collaborative optimization function for real-time deviation iterative compensation; Update the blade cutting path based on the deformation value predicted by the non-linear deformation model; Find the optimal rotation angle and position compensation value according to the Newton iteration method to optimize the deviation value of each cut; Adjust the coordinates of the tool entry and tool exit points according to the optimized compensation value and synchronize them to the numerical control machine tool control system.

6. The method according to claim 5, characterized in that: The collaborative optimization function, the specific formula is: Among them, δy 1,i and δy 2,i are the Y deviations for tool feed-in and tool feed-out, Δθ i is the blade rotation adjustment amount, and λ is the weight coefficient.

7. The method according to claim 1, characterized in that: The method further includes digital twin verification and online feedback, specifically including: Construct a virtual model of material deformation and cutting path, simulate the cutting process, evaluate the accuracy, and optimize the cutting path; Perform virtual cutting based on the optimized path, calculate the maximum deviation of each cut, and record the accuracy distribution; Compare the simulation results with the actual cutting results in real time, update the model, and optimize the cutting path for the next cut.

8. A double-knife cutting precision optimization device based on image alignment, which is applied to the double-knife cutting precision optimization method based on image alignment according to any one of claims 1-7, and is characterized in that, Including: A data set acquisition module for acquiring a reference point data set, establishing a coordinate system based on the reference point data set, photographing the first Mark point on the cutting surface of the substrate through an industrial camera, extracting the coordinates of the first Mark points for tool entry and exit, calculating a straight line based on the coordinates of the first Mark points, and determining the initial cutting direction angle; A deformation deviation calculation module for constructing a cutting path model, determining the theoretical cutting path under the actual deformation of the material according to the initial cutting direction angle, photographing the second Mark point through an industrial camera according to the theoretical cutting path and extracting the coordinates of the second Mark point, measuring the actual cutting path, and calculating the deformation deviation based on the coordinates of the first Mark point and the second Mark point; A compensation module for constructing a non-linear deformation model based on the calculated deformation deviation data, dividing the actual cutting path into several sub-intervals according to the non-linear deformation model, calculating the entry and exit deviation based on the coordinates of the first Mark point and the second Mark point as a reference, and allocating the Y-direction compensation value for each cutting position; An updated cutting path module for constructing an affine transformation matrix according to the blade rotation angle and the scaling factor, mapping the coordinates of the theoretical cutting path points to the actual cutting path, and updating the tool entry and exit positions; An adjustment module for calculating the angle between the current cut and the previous cut when cutting every two cuts based on the updated tool entry and exit positions, adjusting the blade rotation angle, and adjusting the position of the platform in real time according to the coordinates after affine transformation to ensure that the blade path coincides completely with the cutting path.

9. A computer device, characterized in that, Including: A processor; A memory for storing executable instructions; Wherein, the processor is used to read the executable instructions from the memory and execute the executable instructions to implement the method for optimizing the double-blade cutting accuracy based on image alignment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the processor is enabled to implement the method for optimizing the double-blade cutting accuracy based on image alignment as described in any one of claims 1 to 7.

Images