UV transfer printing texture intelligent matching and closed loop correction method, device and equipment

Abstract

The invention relates to a UV transfer printing texture intelligent matching and closed-loop correction method, device and equipment, and belongs to the field of UV transfer printing image analysis and processing.The method comprises the steps that target texture data is obtained and stored as a standard template, and an actual texture image of a workpiece after transfer printing is collected; extracting actual texture contour feature points and standard template reference feature points, and calculating a position deviation vector; the X-Y plane displacement, the Z-axis pressure value and the rotation angle of the imprinting head are adjusted according to the deviation vector, and micron-scale attachment is achieved; impressing force and UV light intensity sensor signals are monitored in real time, and when parameter fluctuation exceeds a threshold value, PID is triggered to adjust and correct impressing parameters; corrected parameters are transmitted to the motion control unit through a CAN bus, and closed-loop control is formed. According to the method, through sub-pixel-level feature matching and multi-parameter correction, the problems that a traditional method is insufficient in matching precision and lacks closed-loop control are solved, high-precision transfer printing of texture similarity and fitting precision is achieved, and the transfer printing yield and consistency of complex workpieces are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of UV transfer image analysis and processing technology, and in particular to a method, apparatus and device for intelligent matching and closed-loop correction of UV transfer textures. Background Art

[0002] In precision manufacturing fields such as optical components and consumer electronics glass covers, UV transfer technology uses an imprint head to transfer textures with micro-nano structures onto the surface of workpieces, making it a key process for achieving high-precision surface decoration and functional integration. Traditional UV transfer processes rely on manual presets or offline calibration for texture matching and parameter adjustment, resulting in two major technical bottlenecks: First, insufficient texture feature matching accuracy. Existing visual inspection methods mostly use template matching based on grayscale values, which is easily affected by lighting fluctuations and workpiece positioning deviations. Furthermore, edge feature extraction only achieves pixel-level accuracy, leading to errors in calculating the positional deviation between the actual texture and the target template, and causing the imprint head posture adjustment to lag behind real-time conditions. Second, the lack of a closed-loop control mechanism. Fluctuations in key parameters such as imprint force and UV curing intensity during the imprinting process lack real-time monitoring and dynamic correction. When disturbances such as equipment vibration and material deformation occur, parameters such as the imprint head pressure and UV lamp power cannot be adjusted in time, ultimately causing defects such as blurred edges and uneven depth in the transferred texture, severely impacting product yield. Furthermore, traditional methods rely solely on visual positioning information for spatial orientation adjustment of the imprint head, failing to integrate force sensor feedback data. This makes it difficult to address the complex bonding requirements of curved workpieces, hindering the achievement of micron-level precision control. As emerging products such as 5G terminal glass and AR optical lenses increasingly demand higher precision in texture transfer (requiring texture similarity ≥95% and bonding error ≤5μm), a UV transfer method with intelligent feature matching and real-time closed-loop correction capabilities is urgently needed to address issues such as lag in parameter adjustment, insufficient multi-source data fusion, and limited control precision in existing technologies. Summary of the Invention

[0003] The main objective of this invention is to provide a method, apparatus, and device for intelligent matching and closed-loop correction of UV transfer textures, which solves the problems of insufficient matching accuracy and lack of closed-loop control in traditional methods, achieves high-precision transfer with high texture similarity and bonding accuracy, and significantly improves the transfer yield and consistency of complex workpieces.

[0004] To achieve the above objectives, the present invention provides a UV transfer texture intelligent matching and closed-loop correction method, comprising the following steps: Acquire target texture data and store it as a standard template, and then use a high-resolution line scan camera to capture the actual texture image of the workpiece surface after transfer. The contour feature points of the actual texture are extracted using an edge detection algorithm, and the reference feature points of the corresponding region are extracted from the standard template. The positional deviation vector between the two is calculated based on the feature point matching algorithm. The spatial attitude parameters of the imprint head are adjusted by driving the servo motor according to the position deviation vector. The spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micrometer level. During the imprinting process, the feedback signals from the imprinting force sensor and the UV curing light intensity sensor are monitored in real time by the embedded controller. When the parameter fluctuation carried by the feedback signal exceeds the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. The corrected printing parameters are transmitted to the motion control unit of the UV transfer equipment via the CAN bus to achieve closed-loop control of the printing process.

[0005] Further, the steps of extracting contour feature points of the actual texture using an edge detection algorithm, simultaneously extracting reference feature points of the corresponding region from the standard template, and calculating the positional deviation vector between the two based on a feature point matching algorithm include: The actual texture image is subjected to gradient calculation and contour extraction using an edge detection algorithm. Continuous contour lines are obtained through non-maximum suppression and dual threshold filtering. Contour feature points with sub-pixel precision are extracted on the contour lines at preset intervals or curvature change rates. Based on the coordinate mapping relationship of the target texture data, a region of interest (ROI) corresponding to the actual texture acquisition area is defined in the standard template, and the same edge detection parameters are used to extract reference feature points in the ROI region. The actual contour feature points and the reference feature points are normalized in terms of coordinates. A matching cost matrix is ​​constructed based on the geometric positional relationship of the feature points. Pairs of the same feature points are selected through bidirectional matching. The coordinate difference of the pairs of the same feature points is fitted by the least squares method to obtain a position deviation vector containing the XY plane offset and rotation angle deviation.

[0006] Further, the step of driving the servo motor to adjust the spatial attitude parameters of the imprint head according to the position deviation vector includes: The XY plane offset in the position deviation vector is converted into pulse control signals for the X-axis and Y-axis servo motors, driving the imprint head to perform translational motion in the horizontal plane. The rotation angle deviation is converted into an angle control signal for the rotary axis servo motor to adjust the rotation posture of the imprint head around the Z-axis; The contact pressure of the imprint head is adjusted by a pressure servo valve based on the Z-axis pressure value in the position deviation vector. The contact force data is collected in real time by a six-dimensional force sensor installed on the imprint head, and the adjustment process is corrected by combining the feedback information of the vision positioning system until the contact accuracy between the imprint head and the workpiece surface meets the preset error requirements.

[0007] Furthermore, the step of obtaining continuous contour lines through non-maximum suppression and dual threshold filtering, and extracting contour feature points with sub-pixel precision on the contour lines at preset intervals or curvature change rates, includes: The Sobel operator is used to calculate the gradient magnitude and direction of the actual texture image to generate a gradient matrix. Non-maximum suppression is performed on the gradient matrix by comparing the gradient magnitude of the current pixel with that of its neighboring pixels in the gradient direction, and retaining only local maxima to refine the edges; Two thresholds, high and low, are set for dual-threshold filtering. Points with gradient magnitude greater than the high threshold are designated as strong edge points, and points between the high and low thresholds are designated as weak edge points. Weak edge points are connected to strong edge points through 8-neighborhood to form a continuous contour line. Subpixel interpolation fitting is performed on the contour line. Based on the fitted curve, contour feature points are extracted according to the preset equal interval sampling interval, or the curvature value of each point on the curve is calculated. Points with curvature greater than the preset threshold are selected as contour feature points. The coordinate accuracy of all contour feature points is improved to the subpixel level through bilinear interpolation algorithm.

[0008] Further, based on the coordinate mapping relationship of the target texture data, the step of defining a Region of Interest (ROI) in the standard template corresponding to the actual texture acquisition area, and extracting reference feature points within the ROI using the same edge detection parameters, includes: Based on the field of view of the high-resolution line scan camera and the workpiece positioning coordinate system, a coordinate mapping matrix of the actual texture acquisition area in the standard template coordinate system is established. The coordinates of the upper left and lower right corners of the actual acquisition area are mapped to the standard template through affine transformation, and a rectangular ROI area containing complete target texture features is delineated. The configuration parameters of the edge detection algorithm are retrieved, including the Sobel operator template, the non-maximum suppression window size, and the high and low thresholds for dual threshold filtering. The same gradient calculation, non-maximum suppression, and dual threshold connection operations are performed within the ROI region to generate the outline of the standard template. The standard template outline is subjected to the same subpixel interpolation fitting process as the actual texture image. Reference feature points are extracted according to preset interval or curvature selection conditions, and the accuracy of the reference feature point coordinates is improved to the subpixel level through bilinear interpolation algorithm.

[0009] Further, the actual contour feature points and reference feature points are normalized in terms of coordinates. A matching cost matrix is ​​constructed based on the geometric positional relationship of the feature points. Pairs of corresponding feature points are selected through bidirectional matching. The coordinate differences of the pairs of corresponding feature points are fitted using the least squares method to obtain a position deviation vector containing the XY plane offset and rotation angle deviation. The steps include: Calculate the centroid coordinates of the actual contour feature point set Centroid coordinates of the reference feature point set The two sets of feature points are translated to a local coordinate system with their respective centroids as the origin, resulting in centroid-free coordinates. and ; The centroid-decentered coordinates are then scaled to calculate the scale factor of the actual feature point set. Scale factor of the baseline feature point set Scale the coordinates to the inside of the unit circle to obtain normalized coordinates. and ; Construct an n×m matching cost matrix C, where, This represents the squared Euclidean distance between the actual feature point i and the reference feature point j. ; The matching cost matrix is ​​subjected to bidirectional matching and filtering. First, the baseline point with the minimum cost in each row is selected as the candidate match. Then, the candidate match in each column is verified to be the bidirectional minimum cost. The bidirectional consistent point pairs are retained as feature point pairs with the same name. Transform the pairs of feature points with the same name back to the original coordinate system to obtain the data formula. The transformation parameters were fitted using the least squares method. , and The XY plane offset is obtained. ) and rotation angle deviation The position deviation vector.

[0010] Furthermore, by real-time monitoring of the feedback signals from the imprint force sensor and the UV curing light intensity sensor via an embedded controller, when the parameter fluctuation carried by the feedback signal is detected to exceed a preset threshold, the PID adjustment module is automatically triggered to correct the imprint parameters, including: Real-time acquisition of pressure signals from the imprint force sensor and light intensity signals from the UV curing light intensity sensor; When the pressure deviation exceeds the preset pressure threshold or the light intensity deviation exceeds the preset light intensity threshold, the sliding window algorithm is activated to analyze the parameter fluctuation trend of the most recent 20 cycles. If the fluctuation amplitude of three consecutive window cycles exceeds the corresponding threshold, it is determined to be a parameter abnormality. The PID control module processes parameters based on their abnormality type. When the pressure is abnormal, it adjusts the feed speed of the servo motor and the pressure compensation coefficient according to the proportional-integral-derivative algorithm. When the light intensity is abnormal, it synchronously adjusts the drive current and curing time of the UV lamp group to control the printing pressure fluctuation within ±2% of the rated pressure and the light intensity fluctuation within ±5% of the rated light intensity.

[0011] Furthermore, the step of transmitting the corrected imprinting parameters to the motion control unit of the UV transfer equipment via the CAN bus includes: The imprinting parameters output by the PID control module, including displacement, pressure, angle and curing parameters, are encapsulated into a standard data frame according to the CANopen protocol. After adding the device address identifier and CRC check code, the frame is sent to the device bus network at a baud rate of 500kbps through the CAN bus interface of the embedded controller. The motion control unit of the UV transfer printing equipment receives data frames through a bus transceiver. First, it parses the device address field to filter the corresponding instructions, then performs CRC verification on the parameters. After the verification is successful, it updates the internal parameter register and generates servo motor control signals and UV lamp drive signals based on the valid parameters, driving the actuator to complete the real-time adjustment of the printing parameters.

[0012] This invention proposes a UV transfer texture intelligent matching and closed-loop correction device, comprising: The acquisition unit is used to acquire target texture data and store it as a standard template, and to acquire the actual texture image of the workpiece surface after transfer using a high-resolution line scan camera. The computing unit is used to extract the contour feature points of the actual texture using an edge detection algorithm, and at the same time extract the reference feature points of the corresponding region from the standard template, and calculate the positional deviation vector between the two based on the feature point matching algorithm. The adjustment unit is used to drive the servo motor to adjust the spatial attitude parameters of the imprint head according to the position deviation vector. The spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micrometer level. The correction unit is used to monitor the feedback signals of the imprinting force sensor and the UV curing light intensity sensor in real time through the embedded controller during the imprinting process. When the parameter fluctuation carried by the feedback signal is detected to exceed the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. The control unit transmits the corrected imprinting parameters to the motion control unit of the UV transfer equipment via the CAN bus, thereby achieving closed-loop control of the imprinting process.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described UV transfer texture intelligent matching and closed-loop correction method.

[0014] The UV transfer texture intelligent matching and closed-loop correction method, apparatus, and device provided by this invention have the following beneficial effects: (1) By using sub-pixel level edge feature extraction and bidirectional feature point matching algorithm, the positional deviation between the actual texture and the target template is accurately calculated, ensuring that the imprint head adjusts its posture based on the real-time detected micron-level deviation, and finally forms a high-precision transfer texture with high similarity to the target texture on the workpiece surface.

[0015] (2) The embedded controller integrates the feedback signals of the printing pressure and UV light intensity sensor in real time, and corrects the parameter fluctuations through the PID adjustment module. The printing pressure fluctuation is controlled within ±2% of the rated pressure range and the light intensity fluctuation is controlled within ±5% of the rated light intensity range, effectively suppressing the impact of equipment vibration, material thickness change and other disturbances on the transfer quality.

[0016] (3) Based on the position deviation vector, the XY plane displacement, Z-axis pressure value and rotation angle of the imprint head are synchronously adjusted. Combined with the fusion calibration of the six-dimensional force sensor and the vision positioning system, the micron-level bonding accuracy between the imprint head and the workpiece surface is achieved. This breaks through the limitation of relying on a single displacement adjustment in the traditional method and significantly improves the transfer consistency of curved and irregular workpieces. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the UV transfer texture intelligent matching and closed-loop correction method in one embodiment of the present invention. Figure 2 This is a structural block diagram of a UV transfer texture intelligent matching and closed-loop correction device in one embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] Reference Figure 1This is a flowchart illustrating a UV transfer texture intelligent matching and closed-loop correction method proposed in this invention. The method includes the following steps: S1, acquire target texture data and store it as a standard template, and use a high-resolution line scan camera to capture the actual texture image of the workpiece surface after transfer; In a specific embodiment of step S1, a digital model of the target texture is generated using a 3D scanner or CAD design software, its two-dimensional contour data is extracted and converted into a grayscale image format, a coordinate index is established based on the coordinate system of the workpiece positioning platform, and it is stored as a standard template file at a preset resolution (e.g., 1200 dpi) to form a template database containing texture geometric features and coordinate mapping relationships; the high-resolution line array camera is installed at the visual inspection station of the UV transfer equipment, and scans the surface of the transferred workpiece line by line with the LED line array light source. When the workpiece passes through the detection area with the motion platform at a constant speed (e.g., 50 mm / s), the camera acquires continuous image strips in a synchronous trigger mode, and synthesizes a complete actual texture image through an image stitching algorithm.

[0021] S2, use an edge detection algorithm to extract the contour feature points of the actual texture, and at the same time extract the reference feature points of the corresponding region from the standard template, and calculate the positional deviation vector between the two based on the feature point matching algorithm; In a specific embodiment of step S2, the position deviation vector is obtained as follows: The actual texture image is subjected to gradient calculation and contour extraction using an edge detection algorithm. Continuous contour lines are obtained through non-maximum suppression and dual threshold filtering. Contour feature points with sub-pixel precision are extracted on the contour lines at preset intervals or curvature change rates. Based on the coordinate mapping relationship of the target texture data, a region of interest (ROI) corresponding to the actual texture acquisition area is defined in the standard template, and the same edge detection parameters are used to extract reference feature points in the ROI region. The actual contour feature points and the reference feature points are normalized in terms of coordinates. A matching cost matrix is ​​constructed based on the geometric positional relationship of the feature points. Pairs of the same feature points are selected through bidirectional matching. The coordinate difference of the pairs of the same feature points is fitted by the least squares method to obtain a position deviation vector containing the XY plane offset and rotation angle deviation.

[0022] Of these three steps, the edge detection algorithm for gradient calculation and contour extraction of the actual texture image is existing technology and will not be elaborated upon in this specification. In the step of obtaining continuous contour lines through non-maximum suppression and double threshold filtering, and extracting sub-pixel-level precision contour feature points on the contour lines at preset intervals or curvature change rates: the Sobel operator is used to calculate the gradient magnitude and direction of the actual texture image to generate a gradient matrix; non-maximum suppression is performed on the gradient matrix, comparing the gradient magnitude of the current pixel with that of its neighboring pixels in the gradient direction, retaining only local maxima. Large values ​​are used to refine the edges; two thresholds, high and low, are set for dual-threshold screening. Points with gradient magnitudes greater than the high threshold are designated as strong edge points, while points between the high and low thresholds are designated as weak edge points. Weak edge points are connected to strong edge points through 8-neighborhoods to form a continuous contour line; sub-pixel interpolation is performed on the contour line. Based on the fitted curve, contour feature points are extracted according to a preset equal-interval sampling interval, or the curvature value of each point on the curve is calculated, and points with curvature greater than a preset threshold are selected as contour feature points. The coordinate accuracy of all contour feature points is improved to the sub-pixel level through a bilinear interpolation algorithm.

[0023] In the step of defining the ROI region corresponding to the actual texture acquisition area in the standard template according to the coordinate mapping relationship of the target texture data, and extracting reference feature points in the ROI region using the same edge detection parameters: Based on the field of view of the high-resolution line scan camera and the workpiece positioning coordinate system, a coordinate mapping matrix of the actual texture acquisition area in the standard template coordinate system is established. The coordinates of the upper left and lower right corners of the actual acquisition area are mapped to the standard template through affine transformation, defining a rectangular ROI region containing complete target texture features. The configuration parameters of the edge detection algorithm are retrieved, including the Sobel operator template, non-maximum suppression window size, and high and low thresholds for dual-threshold filtering. The same gradient calculation, non-maximum suppression, and dual-threshold connection operations are performed in the ROI region to generate the contour line of the standard template. The standard template contour line is subjected to the same sub-pixel interpolation fitting process as the actual texture image. Reference feature points are extracted according to preset interval or curvature filtering conditions, and the accuracy of the reference feature point coordinates is improved to the sub-pixel level using a bilinear interpolation algorithm.

[0024] Finally, the coordinates of the actual contour feature points and the reference feature points are normalized. A matching cost matrix is ​​constructed based on the geometric positional relationship of the feature points. Pairs of corresponding feature points are selected through bidirectional matching. The coordinate differences between the pairs of corresponding feature points are fitted using the least squares method to obtain the position deviation vector containing the XY plane offset and rotation angle deviation. The process involves calculating the centroid coordinates of the actual contour feature point set. Centroid coordinates of the reference feature point set The two sets of feature points are translated to a local coordinate system with their respective centroids as the origin, resulting in centroid-free coordinates. and The centroid-decentered coordinates are then scaled to calculate the scale factor of the actual feature point set. Scale factor of the baseline feature point set Scale the coordinates to the inside of the unit circle to obtain normalized coordinates. and Construct an n×m matching cost matrix C, where, This represents the squared Euclidean distance between the actual feature point i and the reference feature point j. The matching cost matrix is ​​subjected to bidirectional matching and filtering. First, the reference point with the minimum cost in each row is selected as a candidate match. Then, the candidate match in each column is verified to be the bidirectional minimum cost. Point pairs that are consistent in both directions are retained as feature point pairs with the same name. The feature point pairs with the same name are transformed back to the original coordinate system to obtain the data formula. The transformation parameters were fitted using the least squares method. , and The XY plane offset is obtained. ) and rotation angle deviation The position deviation vector.

[0025] In this embodiment of step S2, First, edge detection processing is performed on the actual texture image acquired by the high-resolution line scan camera: the Sobel operator is used to calculate the gradient magnitude (reflecting edge strength) and gradient direction (reflecting edge direction) of each pixel in the image, generating a matrix containing gradient information. To avoid excessively thick edges affecting subsequent matching accuracy, non-maximum suppression is performed on the gradient matrix—in the gradient direction of each pixel (e.g., 8 directions such as 45°, 90°, etc.), only the pixel with the largest gradient magnitude in that direction is retained, and the rest are set to zero, thereby compressing the edge width to the single-pixel level.

[0026] Subsequently, contour continuity was further optimized using a dual-threshold screening method: two thresholds were set (e.g., the high threshold was 1.5 times the average gradient of the image, and the low threshold was 0.8 times). Pixels with gradient magnitudes greater than the high threshold were marked as "strong edge points" (reliable edges), while those between the high and low thresholds were marked as "weak edge points" (possible edges). Through 8-neighborhood connectivity detection, weak edge points were connected to adjacent strong edge points to form continuous closed contour lines, avoiding contour breaks caused by noise.

[0027] To meet the micrometer-level matching accuracy requirements, the contour lines need to be refined to a sub-pixel level: a bilinear interpolation algorithm is used to interpolate and fit the pixels on the contour lines, improving the original pixel-based coordinates (e.g., (100, 150)) to sub-pixel accuracy (e.g., (100.3, 150.7)). Finally, based on the feature complexity of the actual texture, either "equal-interval sampling" or "curvature filtering" is selected to extract feature points: if the texture is a regular geometric shape (e.g., a grid), it is sampled uniformly at a preset interval (e.g., every 5 μm); if it is a complex curved surface texture (e.g., a micro / nano optical structure), the curvature of each point on the contour line (reflecting the degree of curvature) is calculated, and only points with curvature greater than a threshold (e.g., curvature ≥ 0.1 / μm) are retained as feature points, ensuring that the extracted feature points can accurately represent the key structures of the texture.

[0028] To ensure the comparability of feature points between the actual texture and the target template, a Region of Interest (ROI) that strictly corresponds to the actual acquisition area needs to be defined in the standard template. This is achieved through coordinate mapping: Based on the field of view of the high-resolution line scan camera (e.g., 20mm width) and the coordinate system of the workpiece positioning platform (with the lower left corner of the workpiece as the origin), a mapping relationship is established between the actual texture acquisition area (e.g., a 30mm×30mm area on the workpiece surface) and the coordinate system of the standard template. Through affine transformation (including translation, rotation, and scaling parameters), the coordinates of the upper left corner (e.g., (5mm, 5mm)) and lower right corner (e.g., (35mm, 35mm)) of the actual acquisition area are mapped to the standard template, defining a rectangular ROI containing complete target texture features (e.g., the corresponding area in the template is (100 pixels, 100 pixels) to (400 pixels, 400 pixels)).

[0029] Within the defined ROI, edge detection parameters identical to the actual texture (including Sobel operator template size (3×3), non-maximum suppression window (3×3), and high and low thresholds for dual-threshold filtering) are used to repeatedly perform gradient calculation, non-maximum suppression, and dual-threshold connection operations to generate the contour line of the standard template. Subsequently, this contour line is fitted with sub-pixel interpolation identical to the actual texture, and reference feature points are extracted at the same intervals or curvature conditions (e.g., sampling every 5μm or curvature ≥0.1 / μm). Finally, bilinear interpolation is used to improve the coordinate accuracy of the reference feature points to the sub-pixel level, ensuring that the two sets of feature points are matched at the same precision dimension.

[0030] To eliminate matching errors caused by positional offsets or scale differences between actual feature points and reference feature points, coordinate normalization of the two sets of feature points is required: First, calculate the centroid of the actual feature point set (the average coordinates of all points) and the centroid of the reference feature point set, and translate the two sets of points to a local coordinate system with their respective centroids as the origin (i.e., "centroid decentroiding") to eliminate the influence of overall positional offset; then calculate the scale factor of the two sets of points (the average distance of all points to the centroid), and scale the coordinates to within the unit circle (i.e., "scale normalization") to eliminate scale differences caused by texture scaling.

[0031] After normalization, a matching cost matrix is ​​constructed based on the geometric positions of the feature points: each element of the matrix represents the squared Euclidean distance between an actual feature point and a reference feature point (the smaller the distance, the higher the matching probability). To avoid mismatches caused by one-way matching (e.g., actual point A matches reference point B, but the optimal match for reference point B is actual point C), a two-way matching screening is adopted: first, for each row of the matrix (actual points), the reference point with the minimum cost is selected as a candidate; then, for each column of the matrix (reference points), it is verified whether the candidate is the minimum cost point in that column, and only bidirectionally consistent point pairs are retained as "same-name feature point pairs" (i.e., one-to-one correspondence between actual points and reference points).

[0032] Finally, the corresponding feature point pairs are transformed back to the original coordinate system, and a two-dimensional rigid transformation model (including XY plane translation and rotation angle) is fitted using the least squares method to calculate the positional deviation vector between the actual texture and the target template. This vector contains the XY plane offset (e.g., Δx = 2.3 μm, Δy = -1.8 μm) and the rotation angle deviation (e.g., θ = 0.08°), providing a precise correction basis for subsequent imprint head posture adjustment.

[0033] S3, according to the position deviation vector, drive the servo motor to adjust the spatial attitude parameters of the imprint head, the spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micron level; In a specific embodiment of step S3, the XY plane offset in the position deviation vector is converted into pulse control signals for the X-axis and Y-axis servo motors to drive the imprint head to translate in the horizontal plane; the rotation angle deviation is converted into an angle control signal for the rotation axis servo motor to adjust the rotational posture of the imprint head around the Z-axis; the contact pressure of the imprint head is adjusted by a pressure servo valve according to the Z-axis pressure value in the position deviation vector; the contact force data is collected in real time by a six-dimensional force sensor installed on the imprint head, and the adjustment process is corrected by combining the feedback information of visual positioning until the contact accuracy between the imprint head and the workpiece surface meets the preset error requirements.

[0034] The XY plane offset (unit: μm) in the position deviation vector is mapped to pulse control signals (pulse equivalent ≤ 0.1μm / pulse) for the X-axis and Y-axis servo motors via the coordinate transformation algorithm of the motion control card. This drives the precision linear guide mechanism mounted on the impression head to perform translational motion in the horizontal plane. The translational speed is automatically adjusted according to the magnitude of the deviation (20mm / s when the deviation is ≤ 10μm, and reduced to 10mm / s when the deviation is > 10μm) to ensure positioning accuracy. For the rotation angle deviation (unit: mrad), the angle control signal is transmitted to the rotary axis servo motor via a high-precision rotary encoder (resolution ≤ 0.001°), driving the impression head to adjust its attitude around the Z-axis. During the adjustment process, a segmented PID control algorithm (proportional coefficient Kp=20, integral coefficient Ki=5, derivative coefficient Kd=10) is used to suppress overshoot and ensure that the rotational positioning error is ≤ 0.05°. For the Z-axis pressure value (unit: N), the pressure of the pneumatic spring assembly built into the imprint head is adjusted by the pressure servo valve (accuracy ±0.5%FS). The pressure servo valve corrects the output current in real time according to the deviation signal (0-10V control voltage corresponds to 0-500N pressure range) so that the contact pressure of the imprint head matches the surface morphology of the workpiece.

[0035] During the adjustment process, a six-dimensional force sensor (range ±50N for X / Y axes, ±500N for Z axis, resolution 0.1N) installed at the bottom of the impression head collects contact force data in real time at a frequency of 200Hz. After removing high-frequency noise using a Kalman filter algorithm, the data is fused with the position deviation data from visual positioning feedback. When the Z-axis contact force fluctuation exceeds ±5N or the X / Y axis tangential force exceeds 10N, the incremental PID adjustment module is automatically triggered to perform secondary correction on the displacement increment and pressure compensation coefficient of the servo motor. During the correction process, an error iteration algorithm is used (maximum number of iterations: 10). After each adjustment, the surface texture image of the workpiece is re-acquired visually to verify the edge alignment of the bonding area (error ≤ 3μm is acceptable) until the bonding accuracy between the imprint head and the workpiece surface meets the preset requirements (X / Y axis displacement error ≤ 2μm, rotation angle error ≤ 0.1°, contact pressure uniformity ≤ ±3%). This forms a multimodal closed-loop control mechanism based on "visual detection - force control feedback - motion correction", which effectively solves the posture coupling error problem in traditional single vision positioning and adjustment, and ensures high-precision bonding and transfer of complex curved workpieces.

[0036] S4. During the imprinting process, the embedded controller monitors the feedback signals of the imprinting force sensor and the UV curing light intensity sensor in real time. When the parameter fluctuation carried by the feedback signal is detected to exceed the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. In a specific embodiment of step S4, the pressure signal from the imprinting force sensor and the light intensity signal from the UV curing light intensity sensor are collected in real time. When the pressure deviation exceeds a preset pressure threshold or the light intensity deviation exceeds a preset light intensity threshold, a sliding window algorithm is activated to analyze the parameter fluctuation trend over the last 20 cycles. If the fluctuation amplitude exceeds the corresponding threshold for three consecutive window cycles, it is determined to be a parameter anomaly. The PID adjustment module processes the parameter anomaly according to the type of anomaly. When the pressure is abnormal, the feed speed of the servo motor and the pressure compensation coefficient are adjusted according to the proportional-integral-derivative algorithm. When the light intensity is abnormal, the drive current and curing time of the UV lamp group are adjusted synchronously to control the imprinting pressure fluctuation within ±2% of the rated pressure range and the light intensity fluctuation within ±5% of the rated light intensity range.

[0037] The embedded controller acquires pressure signals from an imprint force sensor (selected: range 0-500N, accuracy ±0.3%FS) and light intensity signals from a UV curing light intensity sensor (selected: measurement range 0-2000mW / cm², resolution 1mW / cm²) in real time at a sampling frequency of 100Hz. The raw signals are processed using a 5×5 window median filter to suppress high-frequency electromagnetic interference and noise fluctuations caused by equipment vibration. After signal preprocessing, the absolute deviation of the average pressure value within the current period (10ms) from the average pressure value of the previous 5 periods (50ms), and the absolute deviation of the average light intensity value from the average light intensity value of the previous 5 periods are calculated. When the pressure deviation exceeds ±3% of the rated pressure or the light intensity deviation exceeds ±8% of the rated light intensity, a sliding window algorithm is activated to perform trend analysis on the parameter sequence of the most recent 20 periods (200ms). The sliding window width is set to 5 cycles (50ms). It moves forward by 1 cycle each time. If the pressure fluctuation amplitude of 3 consecutive sliding windows exceeds ±3% of the rated pressure or the light intensity fluctuation amplitude exceeds ±8% of the rated light intensity, it is judged as an abnormal parameter to avoid false triggering caused by a single sudden signal.

[0038] When an abnormal pressure is detected, the PID control module automatically activates the pressure control mode based on the deviation. Through a combination of proportional coefficient (Kp=0.8), integral coefficient (Ki=0.05), and derivative coefficient (Kd=1.2), it outputs a servo motor feed speed correction (adjustment range ±20% of rated speed) and a pressure compensation coefficient (0-1.5). The former adjusts the dynamic contact pressure by changing the pressing rate of the impression head, while the latter corrects the pneumatic output pressure in real time through the pressure servo valve, forming a feedforward-feedback composite control. When the light intensity is abnormal, the control module synchronously generates a UV lamp drive current adjustment signal (0-24V corresponding to 0-100% of rated current) and a curing time compensation value (±50ms), achieving rapid stabilization of light intensity through PWM pulse width modulation technology. During the adjustment process, the printing pressure and light intensity are checked in a closed loop at a period of 20ms until the pressure fluctuation converges to ±2% of the rated pressure range and the light intensity fluctuation converges to ±5% of the rated light intensity range. This ensures the stability of key process parameters during UV transfer printing and effectively solves defects such as uneven texture depth and edge glue overflow caused by parameter drift in traditional open-loop control.

[0039] S5, the corrected printing parameters are transmitted to the motion control unit of the UV transfer equipment via the CAN bus to realize closed-loop control of the printing process.

[0040] In a specific embodiment of step S5, the imprinting parameters output by the PID adjustment module, including displacement, pressure, angle, and curing parameters, are encapsulated into a standard data frame according to the CANopen protocol. After adding a device address identifier and a CRC checksum, the frame is sent to the device bus network at a baud rate of 500kbps through the CAN bus interface of the embedded controller. The motion control unit of the UV transfer printing equipment receives the data frame through the bus transceiver. First, it parses the device address field to filter the corresponding instructions, then performs CRC check on the parameters. After the check passes, it updates the internal parameter register and generates servo motor control signals and UV lamp drive signals based on the valid parameters, driving the actuator to complete the real-time adjustment of the imprinting parameters.

[0041] The process of transmitting the corrected imprinting parameters in a closed loop via the CAN bus is as follows: The imprinting parameters output by the PID control module (including key parameters such as XY-axis displacement, Z-axis pressure, rotation angle correction, and UV curing time) are formatted by the embedded controller and encapsulated into standard data frames according to the DS301 communication specification of the CANopen protocol. Each data frame contains a 1-byte device address (motion control unit address set to 0x0A), a 1-byte parameter type identifier (displacement parameter 0x01, pressure parameter 0x02, angle parameter 0x03, curing parameter 0x04), an 8-byte data payload (using IEEE 754 single-precision floating-point format to store specific values), and a 2-byte CRC-16 checksum (check polynomial 0x1021). After encapsulation, the data frame is transmitted to the device bus network at a baud rate of 500kbps via the embedded controller's CAN bus interface (using an NXPTJA1042 transceiver), ensuring single-frame data transmission within 10ms to meet the real-time requirements of the transfer process.

[0042] After receiving data frames via the onboard CAN bus transceiver (of the same model), the motion control unit of the UV transfer printing equipment first parses the first byte of the device address field, processing only instructions with a target address of 0x0A to avoid bus conflicts. Then, it extracts a 2-byte CRC checksum and performs a CRC calculation on the first 10 bytes of the data frame (address + type + data). If the checksum matches the received value, the data is considered valid; otherwise, it requests retransmission by sending an NMT (Network Management) message containing an error code (0x01) until correct data is received. After successful verification, the motion control unit parses the 8-byte data payload into the corresponding physical quantity (e.g., 0x01 type is parsed as XY displacement in μm) according to the parameter type identifier and updates the internal parameter register (using a double buffering mechanism to avoid read / write conflicts). Ultimately, the motion control unit generates servo motor control signals (such as displacement parameters converted to pulse count and pressure parameters converted to servo valve current values) and UV lamp drive signals (such as curing time parameters converted to PWM duty cycle) based on the updated parameters. These signals are then used by the DO (digital output) module to drive the linear motor, rotary motor, and UV lamp actuators for real-time adjustments, forming a complete closed-loop control chain of "parameter correction - bus transmission - execution feedback." This transmission mechanism utilizes the multi-master and high anti-interference characteristics of the CAN bus (the bus uses shielded twisted-pair cable with a 120Ω terminating resistor) to ensure reliable transmission of imprinting parameters even under complex conditions such as equipment vibration and electromagnetic interference. The measured data packet loss rate is less than 0.01%, effectively guaranteeing the stability and consistency of the UV transfer process.

[0043] Reference Appendix Figure 2 The present invention provides a device structure block diagram for a UV transfer texture intelligent matching and closed-loop correction device, comprising: The acquisition unit is used to acquire target texture data and store it as a standard template, and to acquire the actual texture image of the workpiece surface after transfer using a high-resolution line scan camera. The computing unit is used to extract the contour feature points of the actual texture using an edge detection algorithm, and at the same time extract the reference feature points of the corresponding region from the standard template, and calculate the positional deviation vector between the two based on the feature point matching algorithm. The adjustment unit is used to drive the servo motor to adjust the spatial attitude parameters of the imprint head according to the position deviation vector. The spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micrometer level. The correction unit is used to monitor the feedback signals of the imprinting force sensor and the UV curing light intensity sensor in real time through the embedded controller during the imprinting process. When the parameter fluctuation carried by the feedback signal is detected to exceed the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. The control unit transmits the corrected imprinting parameters to the motion control unit of the UV transfer equipment via the CAN bus, thereby achieving closed-loop control of the imprinting process.

[0044] Reference Figure 3 This invention also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0045] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0046] In summary, the target texture data is acquired and stored as a standard template. The actual texture image of the workpiece after transfer is then captured. The feature points of the actual texture contour and the reference feature points of the standard template are extracted, and the positional deviation vector is calculated. Based on the deviation vector, the XY plane displacement of the imprint head, the Z-axis pressure value, and the rotation angle are adjusted to achieve micron-level bonding. The imprint force and UV light intensity sensor signals are monitored in real time; when parameter fluctuations exceed thresholds, PID control is triggered to correct the imprint parameters. The corrected parameters are transmitted to the motion control unit via the CAN bus, forming a closed-loop control. This invention solves the problems of insufficient matching accuracy and lack of closed-loop control in traditional methods through sub-pixel-level feature matching and multi-parameter correction, achieving high-precision transfer with high texture similarity and bonding accuracy, significantly improving the transfer yield and consistency of complex workpieces.

[0047] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0048] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0049] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for intelligent matching and closed-loop correction of UV transfer textures, characterized in that, The following steps are involved: Acquire target texture data and store it as a standard template, and then use a high-resolution line scan camera to capture the actual texture image of the workpiece surface after transfer. The contour feature points of the actual texture are extracted using an edge detection algorithm, and the reference feature points of the corresponding region are extracted from the standard template. The positional deviation vector between the two is calculated based on the feature point matching algorithm. The spatial attitude parameters of the imprint head are adjusted by driving the servo motor according to the position deviation vector. The spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micrometer level. During the imprinting process, the feedback signals from the imprinting force sensor and the UV curing light intensity sensor are monitored in real time by the embedded controller. When the parameter fluctuation carried by the feedback signal exceeds the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. The corrected printing parameters are transmitted to the motion control unit of the UV transfer equipment via the CAN bus to achieve closed-loop control of the printing process.

2. The UV transfer texture intelligent matching and closed-loop correction method according to claim 1, characterized in that, The steps of extracting contour feature points of the actual texture using an edge detection algorithm, simultaneously extracting reference feature points of the corresponding region from the standard template, and calculating the positional deviation vector between the two based on a feature point matching algorithm include: The actual texture image is subjected to gradient calculation and contour extraction using an edge detection algorithm. Continuous contour lines are obtained through non-maximum suppression and dual threshold filtering. Contour feature points with sub-pixel precision are extracted on the contour lines at preset intervals or curvature change rates. Based on the coordinate mapping relationship of the target texture data, a region of interest (ROI) corresponding to the actual texture acquisition area is defined in the standard template, and the same edge detection parameters are used to extract reference feature points in the ROI region. The actual contour feature points and the reference feature points are normalized in terms of coordinates. A matching cost matrix is ​​constructed based on the geometric positional relationship of the feature points. Pairs of the same feature points are selected through bidirectional matching. The coordinate difference of the pairs of the same feature points is fitted by the least squares method to obtain a position deviation vector containing the XY plane offset and rotation angle deviation.

3. The UV transfer texture intelligent matching and closed-loop correction method according to claim 2, characterized in that, The steps of adjusting the spatial attitude parameters of the impression head by driving the servo motor according to the position deviation vector include: The XY plane offset in the position deviation vector is converted into pulse control signals for the X-axis and Y-axis servo motors, driving the imprint head to perform translational motion in the horizontal plane. The rotation angle deviation is converted into an angle control signal for the rotary axis servo motor to adjust the rotation posture of the imprint head around the Z-axis; The contact pressure of the imprint head is adjusted by a pressure servo valve based on the Z-axis pressure value in the position deviation vector. The contact force data is collected in real time by a six-dimensional force sensor installed on the imprint head, and the adjustment process is corrected by combining the feedback information of the vision positioning system until the contact accuracy between the imprint head and the workpiece surface meets the preset error requirements.

4. The UV transfer texture intelligent matching and closed-loop correction method according to claim 2, characterized in that, The steps of obtaining continuous contour lines through non-maximum suppression and dual thresholding, and extracting sub-pixel-level precision contour feature points on the contour lines at preset intervals or curvature change rates, include: The Sobel operator is used to calculate the gradient magnitude and direction of the actual texture image to generate a gradient matrix. Non-maximum suppression is performed on the gradient matrix by comparing the gradient magnitude of the current pixel with that of its neighboring pixels in the gradient direction, and retaining only local maxima to refine the edges; Two thresholds, high and low, are set for dual-threshold filtering. Points with gradient magnitude greater than the high threshold are designated as strong edge points, and points between the high and low thresholds are designated as weak edge points. Weak edge points are connected to strong edge points through 8-neighborhood to form a continuous contour line. Subpixel interpolation fitting is performed on the contour line. Based on the fitted curve, contour feature points are extracted according to the preset equal interval sampling interval, or the curvature value of each point on the curve is calculated. Points with curvature greater than the preset threshold are selected as contour feature points. The coordinate accuracy of all contour feature points is improved to the subpixel level through bilinear interpolation algorithm.

5. The UV transfer texture intelligent matching and closed-loop correction method according to claim 4, characterized in that, Based on the coordinate mapping relationship of the target texture data, the step of defining a Region of Interest (ROI) in the standard template corresponding to the actual texture acquisition area, and extracting reference feature points within the ROI using the same edge detection parameters, includes: Based on the field of view of the high-resolution line scan camera and the workpiece positioning coordinate system, a coordinate mapping matrix of the actual texture acquisition area in the standard template coordinate system is established. The coordinates of the upper left and lower right corners of the actual acquisition area are mapped to the standard template through affine transformation, and a rectangular ROI area containing complete target texture features is delineated. The configuration parameters of the edge detection algorithm are retrieved, including the Sobel operator template, the non-maximum suppression window size, and the high and low thresholds for dual threshold filtering. The same gradient calculation, non-maximum suppression, and dual threshold connection operations are performed within the ROI region to generate the outline of the standard template. The standard template outline is subjected to the same subpixel interpolation fitting process as the actual texture image. Reference feature points are extracted according to preset interval or curvature selection conditions, and the accuracy of the reference feature point coordinates is improved to the subpixel level through bilinear interpolation algorithm.

6. The UV transfer texture intelligent matching and closed-loop correction method according to claim 5, characterized in that, The steps include: normalizing the coordinates of actual contour feature points and reference feature points; constructing a matching cost matrix based on the geometric positional relationship of the feature points; filtering out pairs of corresponding feature points through bidirectional matching; and fitting the coordinate differences of the pairs of corresponding feature points using the least squares method to obtain a position deviation vector containing XY plane offset and rotation angle deviation. Calculate the centroid coordinates of the actual contour feature point set Centroid coordinates of the reference feature point set The two sets of feature points are translated to a local coordinate system with their respective centroids as the origin, resulting in centroid-free coordinates. and ; The centroid-decentered coordinates are then scaled to calculate the scale factor of the actual feature point set. Scale factor of the baseline feature point set Scale the coordinates to the inside of the unit circle to obtain normalized coordinates. and ; Construct an n×m matching cost matrix C, where, This represents the squared Euclidean distance between the actual feature point i and the reference feature point j. ; The matching cost matrix is ​​subjected to bidirectional matching and filtering. First, the baseline point with the minimum cost in each row is selected as the candidate match. Then, the candidate match in each column is verified to be the bidirectional minimum cost. The bidirectional consistent point pairs are retained as feature point pairs with the same name. Transform the pairs of feature points with the same name back to the original coordinate system to obtain the data formula. The transformation parameters were fitted using the least squares method. , and The XY plane offset is obtained. ) and rotation angle deviation The position deviation vector.

7. The UV transfer texture intelligent matching and closed-loop correction method according to claim 1, characterized in that, The embedded controller monitors the feedback signals from the imprint force sensor and the UV curing light intensity sensor in real time. When the parameter fluctuation carried by the feedback signal exceeds a preset threshold, the PID adjustment module is automatically triggered to correct the imprint parameters, including: Real-time acquisition of pressure signals from the imprint force sensor and light intensity signals from the UV curing light intensity sensor; When the pressure deviation exceeds the preset pressure threshold or the light intensity deviation exceeds the preset light intensity threshold, the sliding window algorithm is activated to analyze the parameter fluctuation trend of the most recent 20 cycles. If the fluctuation amplitude of three consecutive window cycles exceeds the corresponding threshold, it is determined to be a parameter abnormality. The PID control module processes parameters based on their abnormality type. When the pressure is abnormal, it adjusts the feed speed of the servo motor and the pressure compensation coefficient according to the proportional-integral-derivative algorithm. When the light intensity is abnormal, it synchronously adjusts the drive current and curing time of the UV lamp group to control the printing pressure fluctuation within ±2% of the rated pressure and the light intensity fluctuation within ±5% of the rated light intensity.

8. The UV transfer texture intelligent matching and closed-loop correction method according to claim 1, characterized in that, The steps for transmitting the corrected imprinting parameters to the motion control unit of the UV transfer equipment via the CAN bus include: The imprinting parameters output by the PID control module, including displacement, pressure, angle and curing parameters, are encapsulated into a standard data frame according to the CANopen protocol. After adding the device address identifier and CRC check code, the frame is sent to the device bus network at a baud rate of 500kbps through the CAN bus interface of the embedded controller. The motion control unit of the UV transfer printing equipment receives data frames through a bus transceiver. First, it parses the device address field to filter the corresponding instructions, then performs CRC verification on the parameters. After the verification is successful, it updates the internal parameter register and generates servo motor control signals and UV lamp drive signals based on the valid parameters, driving the actuator to complete the real-time adjustment of the printing parameters.

9. A UV transfer texture intelligent matching and closed-loop correction device, characterized in that, include: The acquisition unit is used to acquire target texture data and store it as a standard template, and to acquire the actual texture image of the workpiece surface after transfer using a high-resolution line scan camera. The computing unit is used to extract the contour feature points of the actual texture using an edge detection algorithm, and at the same time extract the reference feature points of the corresponding region from the standard template, and calculate the positional deviation vector between the two based on the feature point matching algorithm. The adjustment unit is used to drive the servo motor to adjust the spatial attitude parameters of the imprint head according to the position deviation vector. The spatial attitude parameters include XY plane displacement, Z-axis pressure value and rotation angle, so that the contact accuracy between the imprint head and the workpiece surface reaches the micrometer level. The correction unit is used to monitor the feedback signals of the imprinting force sensor and the UV curing light intensity sensor in real time through the embedded controller during the imprinting process. When the parameter fluctuation carried by the feedback signal is detected to exceed the preset threshold, the PID adjustment module is automatically triggered to correct the imprinting parameters. The control unit transmits the corrected imprinting parameters to the motion control unit of the UV transfer equipment via the CAN bus, thereby achieving closed-loop control of the imprinting process.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the UV transfer texture intelligent matching and closed-loop correction method according to any one of claims 1 to 8.

Images