Automatic pattern overprinting and printing method for textile fabric

Abstract

The invention belongs to the technical field of digital printing machines, and particularly discloses and provides an automatic pattern overprinting method for textile fabrics, which performs real-time acquisition through a high-definition industrial scanning camera, performs multiple operations such as image preprocessing, feature recognition and position adjustment, effectively solves the problem of insufficient printing precision at present, and improves the printing accuracy of the textile fabrics. According to the method, it can be guaranteed that no matter how the specification and size of equipment change, how the size of a to-be-printed textile fabric change and how the printing parameter mode change, images can be collected in a self-adaptive mode, the quality of a final product is guaranteed, and by calling RI software and board card printing software, deformed image data are called to the RI software to be subjected to image RI processing, so that the quality of the finished product is guaranteed. PRT printing data is produced and then sent to board card printing software through a network port for real-time printing. High integration is completely achieved, synchronous parallel operation of the stations of the multifunctional virtual assembly line is achieved, and the printing precision and efficiency are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of digital printing machines, and in particular relates to a method for automatically overprinting patterns on textile fabrics. Background Art

[0002] With the development of computer technology and inkjet printing technology, digital printing technology has begun to be applied to the textile industry. This technology allows designers to transfer patterns directly from computers to printers and then print them on fabrics, which can quickly process complex printing.

[0003] At present, there are two types of textile fabric printing product demands. The first type requires digital printing on jacquard fabrics to provide people with fabrics with three-dimensional color patterns. The second type requires double-sided printing on high-end fabrics, and the patterns on both sides require strict alignment, such as high-end silk fabrics. Obviously, in response to the above two types of printing product demands, the current textile fabric printing processing still has the following deficiencies: 1. There are still some deficiencies in printing accuracy. The current printing technology will cause the position deviation of the printed pattern due to factors such as the stretching and deformation of the fabric, affecting the quality of the final product, especially when double-sided printing, the current image capture design still has certain limitations.

[0004] 2. There is still a certain deviation in printing adaptability. At present, the printing production line is usually composed of multiple independent equipment, and the coordination between each link is difficult.

[0005] 3. The image processing process is relatively complicated. Currently, there is no optimized design for image acquisition, which increases the complexity of subsequent image processing. At the same time, image processing in current printing technology usually requires complex algorithms and a large amount of computing resources, and is less efficient. Summary of the invention

[0006] In view of this, in order to solve the problems raised in the above background technology, a method for automatically overprinting and printing patterns of textile fabrics is now proposed.

[0007] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a method for automatic overprinting and printing of patterns on textile fabrics, the method comprising: step 1, AI feature template construction: used to import the design drawing to be printed, record the resolution of the design drawing and set the coordinate system, and establish an AI feature template, and identify the position information of each target pattern from the AI ​​feature template.

[0008] Step 2: Image acquisition and upload: The high-definition industrial scanning camera installed on the back of the corresponding carriage of the digital printing guide machine is used to collect real-time images of textile fabrics entering the printing work area.

[0009] Step 3, image analysis and acquisition adjustment: Analyze the imaging status of the image of the textile fabric, obtain the imaging score of the image, and determine whether the imaging score reaches the set imaging score. If not, confirm to adjust the light source setting index and return to step 2 until the set imaging score is reached. If reached, execute the next step.

[0010] Step 4: Image preprocessing: Transform the image into spatial coordinates to make it consistent with the position information and resolution of the design drawing.

[0011] Step 5: Image feature recognition: Use AI feature recognition algorithm on the transformed image to find the AI ​​features and coordinate information of the pattern in the image.

[0012] Step 6: Pattern position adjustment: Based on AI features and coordinate information, combined with the set coordinate system of the printed design drawing, the position of the target pattern outline in the design drawing is adjusted through the position correction algorithm to generate the printing pattern of the current physical textile fabric.

[0013] Step 7, pattern format processing: perform RIP processing on the print pattern of the current physical textile fabric to generate PRT format data.

[0014] Step 8: Pattern printing execution: Send the PRT format data to the board printing software through the network port for real-time printing.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention effectively solves the problem of insufficient current printing accuracy by utilizing a high-definition industrial scanning camera for real-time acquisition and performing multiple operations such as image preprocessing, feature recognition, and position adjustment. It can ensure that no matter how the specifications and dimensions of the equipment change, how the dimensions of the textile fabric to be printed change, and how the printing parameter mode changes, images can be adaptively acquired and the quality of the final product can be guaranteed, breaking the limitations of the current image capture design.

[0016] (2) The present invention establishes an AI feature template and combines a high-definition industrial scanning camera and a printer to achieve the linkage and coordination of multiple devices in the front printing production line, avoiding the cumbersomeness of the current single-item control, improving the adaptability of the current fabric printing, and also ensuring the efficiency and accuracy of the current fabric printing.

[0017] (3) This aspect effectively shortens the overall processing time of the fabric printing process by establishing an AI feature template, reduces errors caused by human misjudgment, can adapt to printing scenarios with high repeatability and high detail requirements, and can quickly identify target patterns.

[0018] (4) The present invention analyzes the imaging status of the image of the textile fabric and adjusts the light source, thereby breaking the disadvantage of the current image processing process being relatively complicated, realizing the optimized design during image acquisition, reducing the complexity of subsequent image processing, and reducing the demand for processing algorithms and computing resources, thereby ensuring image processing efficiency.

[0019] (5) The present invention calls RIP software and board printing software to process the deformed image data with RIP software, produces PRT printing data, and then sends the board printing software through the network port for real-time printing. It fully achieves real-time image acquisition, real-time image recognition, real-time image correction, and real-time image RIP, integrates multiple steps such as image acquisition, recognition, correction, RIP processing and printing, and forms an efficient virtual assembly line. All processes can run synchronously and in parallel, greatly improving production efficiency and the accuracy of pattern overprinting. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0021] Figure 1 The present invention is a schematic flow chart of the steps for implementing the method.

[0022] Figure 2 This is a specific implementation flow chart of step 3 of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] See also Figure 1 As shown, the present invention provides a method for automatic overprinting and printing of patterns of textile fabrics, the method comprising: step 1, AI feature template construction: used to import the design drawing to be printed, record the resolution of the design drawing and set the coordinate system, and establish the AI ​​feature template, and identify the position information of each target pattern from the AI ​​feature template.

[0025] Step 2: Image acquisition and upload: The high-definition industrial scanning camera installed on the back of the corresponding carriage of the digital printing guide machine is used to collect real-time images of textile fabrics entering the printing work area.

[0026] See also Figure 2 As shown, step 3, image analysis and acquisition adjustment: analyze the imaging status of the image of the textile fabric, obtain the imaging score of the image, and determine whether the imaging score reaches the set imaging score. If not, confirm to adjust the light source setting index and return to step 2 until the set imaging score is reached. If reached, execute the next step.

[0027] Step 4: Image preprocessing: Transform the image into spatial coordinates to make it consistent with the position information and resolution of the design drawing.

[0028] Step 5: Image feature recognition: Use AI feature recognition algorithm on the transformed image to find the AI ​​features and coordinate information of the pattern in the image.

[0029] Step 6: Pattern position adjustment: Based on AI features and coordinate information, combined with the set coordinate system of the printed design drawing, the position of the target pattern outline in the design drawing is adjusted through the position correction algorithm to generate the printing pattern of the current physical textile fabric.

[0030] Step 7, pattern format processing: perform RIP processing on the print pattern of the current physical textile fabric to generate PRT format data.

[0031] Step 8: Pattern printing execution: Send the PRT format data to the board printing software through the network port for real-time printing.

[0032] Understandably, what needs to be added about step 1 is that the high-definition industrial linear array camera is fixedly installed on the back of the printing carriage and moves synchronously with the printing carriage. When the equipment is producing and printing, the carriage moves back and forth along the X beam, and the camera synchronously scans the image of the textile fabric to be printed along the X direction. Every time the carriage prints a PASS, the equipment guide belt Y moves forward a set distance, and the camera synchronously collects images of the Y forward height.

[0033] Understandably, the step 1 of establishing the AI ​​feature template includes: A1, using a high-definition industrial linear array camera to scan the selected textile fabric sample to obtain a return image of the textile fabric.

[0034] A2. On the return image, use image editing software to select the target pattern area and record it as the AI ​​feature recognition area.

[0035] It should be added that selecting an area using image editing software is a common method, and the specific selection details will not be repeated here.

[0036] It should also be added that the area where the target pattern is selected can also be selected using a special AI tool, and the size and shape of the feature recognition area should be adjusted according to the characteristics of the target pattern to ensure that the key features of the target pattern can be accurately included. If there are repeated flower patterns on the fabric, then key feature points such as the center point of the flower and the edge of the petals need to be defined.

[0037] A3. Standardize the position information of each AI feature recognition area on the return image, including the horizontal coordinate, vertical coordinate and rotation angle.

[0038] A4. Based on the marked return image and AI feature recognition area, the AI ​​model is preliminarily trained. After the preliminary training is completed, it is tested with actual samples and the model parameters are adjusted according to the test results. The AI ​​model that passes the test is used as the AI ​​feature template.

[0039] It should be added that model training and test judgment, which are the basic considerations and conventional means for building existing models, will not be explained again.

[0040] In a specific embodiment, the test pass can be detected by the consistency between the printed pattern and the target pattern in the design drawing, wherein the consistency detection includes multiple features such as position, image shape, etc., wherein the multiple features such as position, image shape, etc. can be constituted into a feature set, and calculated by Jaccard similarity, and the calculation result is used as the consistency detection result, and a passing threshold is set. For example, when the consistency detection result is greater than 0.98, it can be regarded as a test pass.

[0041] The embodiment of the present invention effectively shortens the overall processing time of the fabric printing process by establishing an AI feature template, reduces errors caused by human misjudgment, can adapt to printing scenarios with high repeatability and high detail requirements, and can quickly identify target patterns.

[0042] It can be understood that in step 2, a plurality of LED linear light sources are installed at the bottom of the high-definition industrial scanning camera, and the LED linear light sources move synchronously with the high-definition industrial scanning camera.

[0043] It should be added that an LED linear light source is also installed at the bottom of the high-definition industrial line array camera and moves synchronously with the camera to fill in the jacquard on the textile fabric or the printed pattern on the back to obtain a clear pattern that is easy for AI feature recognition. When the imaging status is analyzed, the LED linear light source can be adjusted if the set imaging score is not reached.

[0044] The embodiment of the present invention establishes an AI feature template and combines a high-definition industrial scanning camera and a printer to achieve the linkage and coordination of multiple devices in the front printing production line, avoids the cumbersomeness of the current monotonic control, improves the adaptability of the current fabric printing, and also ensures the efficiency and accuracy of the current fabric printing.

[0045] Understandably, the imaging state analysis of the image of the textile fabric in step 3 includes: B1, segmenting the image of the textile fabric through a sliding window to obtain segmented image blocks, and confirming various imaging evaluation indicators of the textile fabric.

[0046] B2. Compare each imaging evaluation index of the textile fabric with each set reference imaging evaluation index one by one. If a certain imaging evaluation index is greater than the corresponding reference imaging evaluation index, mark the imaging evaluation index as a matching index.

[0047] B3. Count the number of matching indices and the number of imaging evaluation indices, and record the ratio of the two as the imaging matching ratio.

[0048] B4. If a certain imaging evaluation index is less than the corresponding reference imaging evaluation index, the imaging evaluation index is subtracted from the corresponding reference imaging evaluation index, and the absolute value of the difference is taken as the deviation value of the imaging evaluation index. The imaging evaluation index is marked as the deviation index, and the matching correction factor is set. The imaging matching ratio is corrected to obtain the corrected imaging matching ratio.

[0049] In a specific embodiment, the reference imaging evaluation index includes a reference imaging quality conformance and a reference imaging quality uniformity. For example, the reference imaging quality conformance can be specifically taken as 0.5, and the reference imaging quality uniformity can be specifically taken as 0.2.

[0050] B5. The product of the corrected imaging coincidence ratio and the preset total imaging score is used as the imaging score of the image.

[0051] Furthermore, in step B1, various imaging evaluation indicators of the textile fabric are confirmed, including: B11, detecting each segmented image block by an edge detection algorithm to obtain the image clarity of each segmented image block.

[0052] It should be added that detecting each segmented image block by edge detection algorithm is one of the common tasks in image processing. Edge detection algorithm can help identify the boundaries and contours in the image. In a specific embodiment, the edge detection algorithm includes but is not limited to Canny edge detection and Sobel detection. For example, Canny edge detection can be selected as the edge detection algorithm described in the present invention.

[0053] B12. Generate a color histogram of each segmented image block, extract the minimum pixel intensity and the maximum pixel intensity corresponding to the non-zero frequency in the color histogram, and extract the pixel intensity and the total number of pixels of each channel.

[0054] It should be added that the process of generating a color histogram of each segmented image block can be achieved by using OpenCV and NumPy libraries, etc.

[0055] B13. Calculate the standard deviation of the pixel intensity of each channel to obtain the standard deviation of the pixel intensity corresponding to each segmented image block.

[0056] B14. Compare the maximum pixel intensity and the minimum pixel intensity to obtain the dynamic range corresponding to each segmented image block, set the weights of the pixel intensity standard deviation and the dynamic range, and obtain the contrast of each segmented image block by weighted summation.

[0057] It should be added that the pixel intensity standard deviation reflects the change of the pixel grayscale value within the image block. The larger the standard deviation, the more drastic the grayscale change within the image block and the higher the contrast. The dynamic range is defined as the ratio of the maximum pixel intensity to the minimum pixel intensity of the image block, which reflects the range of grayscale values ​​within the image block. The larger the dynamic range, the wider the grayscale change range within the image block and the higher the contrast. The pixel intensity standard deviation reflects the image contrast in breadth and is more comprehensive. Therefore, the weight of the pixel intensity standard deviation is set to be greater than the weight of the dynamic range. For example, the weight of the pixel intensity standard deviation can be 0.7, and the weight of the dynamic range can be 0.3.

[0058] It can be understood that the specific example statistical process of obtaining the contrast of each segmented image block by weighted summation is as follows: the weights of setting the pixel intensity standard deviation and dynamic range are recorded as w1 and w2 respectively, and the pixel intensity standard deviation and dynamic range corresponding to each segmented image block are recorded as σ j and DR j , j represents the number of the segmented image block, j = 1, 2, ... m.

[0059] Count the contrast of each segmented image block, denoted as (Co) j , (Co) j =w1*σ j +w2*log2(DR j ).

[0060] It needs to be explained that log2(DR j ) instead of using DR j The reason why it is directly involved in the calculation is that the dynamic range value may be very large. Taking the logarithm can compress it into a more reasonable range to avoid the influence of excessive values ​​on the results.

[0061] B15. By using the gray-level co-occurrence matrix to extract the energy value and entropy value of each segmented image block, the texture consistency of each segmented image block is calculated.

[0062] Among them, calculating the texture consistency of each segmented image block includes: R1, comparing the energy value and entropy value of each segmented image block with the energy value and entropy value of the image reference respectively, and marking the texture consistency of the image block whose energy value is lower than the energy value of the image reference and whose entropy value is higher than the entropy value of the image reference as 0.

[0063] R2. If the energy value of a segmented image block is higher than or equal to the energy value of the image reference and the entropy value is higher than the entropy value of the image reference, the energy value of the adjacent segmented image and the energy value of the image reference are recorded as ζ and ζ′ respectively. The texture matching degree of the image block is recorded as α, and ε is the set energy value matching compensation factor.

[0064] Understandably, higher energy values ​​indicate a more uniform distribution of gray levels in the image and better consistency. Entropy values ​​reflect the uncertainty of information. Lower entropy values ​​generally indicate more consistent and regular textures.

[0065] In a specific embodiment, the energy value ranges from [0, 1]. Specifically, 0.7 can be used as the specific value of ζ′.

[0066] R3. If the energy value of a segmented image block is lower than the energy value of the image reference and the entropy value is lower than or equal to the entropy value of the image reference, the entropy value of the adjacent segmented image and the entropy value of the image reference are recorded as and Will The texture matching degree of the image block is recorded as β, and ε′ is the set entropy value matching compensation factor.

[0067] In a specific embodiment, the entropy value of an image is generally between 0 and 8. Specifically, 4 can be used as The specific value of .

[0068] It should be added that in texture analysis, energy can be used to represent the uniformity of an image. If the pixel values ​​in an area do not change much, the energy value of the area will be relatively high, indicating that the texture of the area is relatively uniform. On the contrary, if the pixel values ​​in an area change greatly, the energy value of the area is low, indicating that the texture of the area is relatively complex, and entropy can be used to measure the complexity and diversity of textures. High entropy values ​​are usually associated with complex textures, while low entropy values ​​may indicate simple, repetitive texture patterns. Therefore, energy describes the uniformity of the image more, while entropy focuses more on describing the complexity and diversity of textures. When printing fabrics, more emphasis is placed on the consistency of printing. Therefore, the energy value matching compensation factor is set to be greater than the entropy value matching compensation factor. For example, the energy value matching compensation factor can be taken as 0.6, and the entropy value matching compensation factor can be taken as 0.4.

[0069] R4. If the energy value of a segmented image block is higher than or equal to the energy value of the image reference and the entropy value is lower than or equal to the entropy value of the image reference, As the texture matching degree δ of the image block, the texture matching degree of each segmented image block is obtained, wherein the value of the texture matching degree of each segmented image block belongs to one of 0, β or δ.

[0070] B16. Calculate the mean values ​​of image clarity, image contrast and texture consistency of each image block, set the weights of clarity, contrast and texture consistency, and obtain the imaging quality consistency through linear regression function statistics.

[0071] It should be added that, before the linear regression function is used to calculate the imaging quality conformity, the mean values ​​of the image clarity, image contrast and texture conformity corresponding to each image block are normalized. Specifically, the minimum-maximum normalization or Z-score normalization can be used. The specific example process of using the minimum-maximum normalization for image clarity is as follows: the mean value of the image clarity corresponding to each image block is recorded as The maximum and minimum image resolutions are selected from the image resolutions corresponding to each image block and are denoted as Sh. max and Sh min ,Will As the normalized result of image clarity.

[0072] It can be understood that the normalization process of image contrast and texture consistency is the same as the normalization process of image clarity, and will not be shown here.

[0073] It should also be added that in fabric printing, setting the weights of clarity, contrast and texture matching needs to consider specific application requirements and goals. Because clarity directly determines the subsequent printing quality, the weight of clarity is set to the largest. For example, the weight of clarity can be taken as 0.5. Contrast determines the boundaries and details of the pattern, so the weight of contrast is set second and the weight of texture matching is the smallest. For example, the weight of contrast can be set to 0.3 and the weight of texture matching can be taken as 0.2.

[0074] B17. Calculate the standard deviation of image clarity, image contrast and texture consistency of each image segmentation block respectively, and calculate the imaging quality uniformity in the same way as the imaging quality conformity. Use the imaging quality conformity and imaging quality uniformity as the imaging evaluation indicators of textile fabrics.

[0075] Furthermore, in step B4, a matching correction factor is set, including: B41, the number of statistical deviation indicators, denoted as D0.

[0076] B42. Set a first reference deviation value and a second reference deviation value.

[0077] It should be supplemented that the specific values ​​of the first reference deviation value and the second reference deviation value can be set comprehensively based on historical experience, etc., that is, by collecting historical data, including the actual values ​​and reference values ​​of each imaging evaluation indicator, recording the deviation value of each imaging evaluation indicator, and calculating statistics such as the minimum value and median value of the deviation value. If the minimum value is 0, a value slightly greater than 0 is taken as the first reference deviation value, such as 0.1. If the minimum value is not 0, the minimum value is taken as the first reference deviation value, and the median value is taken as the second reference deviation value.

[0078] B43. Deviation indicators with deviation values ​​less than or equal to the first reference deviation value are marked as Class I deviation indicators.

[0079] B45. If the deviation value is greater than the first reference deviation value and less than or equal to the second reference deviation value, the deviation index shall be marked as a Class II deviation index.

[0080] B46. Deviation indicators with deviation values ​​greater than the set second reference deviation value are marked as Class III deviation indicators.

[0081] B47. Count the number of type I deviation indicators, type II deviation indicators, and type III deviation indicators, and compare them with the number of deviation indicators. The ratios are recorded as type I deviation ratio, type II deviation ratio, and type III deviation ratio, respectively, and are recorded as k a , k b and k c .

[0082] B48. Set the weights of the deviation ratios of Class I, Class II, and Class III, denoted as f1, f2, and f3 respectively, and set the coincidence correction factor ψ.

[0083] In a specific embodiment, the Class I deviation ratio represents the proportion of deviation indicators with smaller deviation values, the Class II deviation ratio represents the proportion of deviation indicators with deviation values ​​in the middle echelon, and the Class III deviation ratio represents the proportion of deviation indicators with larger deviation values. In terms of the degree of deviation, the degree of deviation of the Class III deviation ratio is higher than that of the Class II deviation ratio, and the degree of deviation of the Class II deviation ratio is higher than that of the Class I deviation ratio. Therefore, the weight of the Class III deviation ratio is set to be greater than the weight of the Class II deviation ratio and greater than the weight of the Class I deviation ratio. Exemplarily, the weight f3 of the Class III deviation ratio can be 0.6, the weight f2 of the Class II deviation ratio can be 0.3, and the weight f1 of the Class I deviation ratio can be 0.1.

[0084] It should be added that the corrected imaging coincidence ratio=imaging coincidence ratio×(1-coincidence correction factor).

[0085] It can be understood that regarding the confirmation of adjusting the light source setting indicators in step 3, it includes: T1, if the number of deviation indicators is 2 or the number of deviation indicators is 1 and it is the imaging quality compliance, overall adjustment is used as the adjustment method; if the number of deviation indicators is 1 and it is the imaging quality uniformity, local adjustment is used as the adjustment method.

[0086] T2. When the adjustment mode is overall adjustment, extract the current set light intensity and the maximum light intensity of the LED linear light source corresponding to the high-definition industrial scanning camera, which are recorded as I0 and I max , set the overall light source adjustment factor Will As an indicator for adjusting light source intensity and light source settings.

[0087] The overall light source adjustment factor is set, including: if the number of deviation indicators is 2, the deviation compensation factor is recorded as φ0; if the number of deviation indicators is 1, the deviation compensation factor is recorded as φ1, thereby obtaining the deviation compensation factor η, where η is φ0 or φ1, φ0>φ1.

[0088] The image quality conformity and image quality uniformity are recorded as FH and JY respectively, and the overall light source adjustment factor is set. FH′ and JY′ are the set reference imaging quality conformity and reference imaging quality uniformity respectively.

[0089] T3. When the adjustment mode is local adjustment, confirm the adjustment light source setting index of the local adjustment.

[0090] Among them, confirming the adjustment light source setting indicators for local adjustment includes: T31, extracting the setting irradiation area positions of each LED linear light source corresponding to the high-definition industrial scanning camera, and using each image block located at the setting irradiation area position corresponding to the LED linear light source as each associated image block of the corresponding LED linear light source.

[0091] T32. Calculate the imaging quality conformity of each LED linear light source corresponding to each associated image block, and record the LED linear light source of the associated image block whose imaging quality conformity is less than the reference imaging quality conformity as an adjusted light source.

[0092] T33. Set the overall light source adjustment factor Then, the adjusted light source intensity of the adjusted light source is confirmed in the same way as the adjusted light source setting index during overall adjustment, and used as the adjusted light source setting index for local adjustment, where e is a natural constant.

[0093] It should be added that the imaging quality conformity of each associated image block is calculated in a similar manner to that of FH. The imaging clarity and other indicators of each associated image block are imported into the calculation framework corresponding to FH to obtain the imaging quality conformity of each associated image block.

[0094] The embodiment of the present invention analyzes the imaging status of the image of the textile fabric and adjusts the light source, thereby breaking the disadvantage of the current image processing process being relatively cumbersome, achieving an optimized design during image acquisition, reducing the cumbersomeness of subsequent image processing, and reducing the demand for processing algorithms and computing resources, thereby ensuring image processing efficiency.

[0095] Understandably, what needs to be added about step 4 is that after the high-definition industrial linear array camera captures the return image, the image can be transformed into a spatial coordinate system through a pre-calibration algorithm to eliminate lens distortion and convert the image into a standard coordinate system so that the image is consistent with the coordinate system and resolution of the design drawing, wherein the pre-calibration parameter algorithm refers to the camera calibration parameter algorithm, wherein the specific conversion example process for the spatial coordinate transformation of the image is as follows: 1) Extract the internal parameters of the high-definition industrial linear array camera, such as focal length, principal point offset, lens distortion coefficient, etc., and at the same time extract the external parameters of the high-definition industrial linear array camera, such as the position and direction of the camera relative to the world coordinate system, and correct it through the distortion correction formula. After the distortion correction is completed, the image needs to be converted from the camera coordinate system to the coordinate system of the design drawing.

[0096] 2) Obtain the internal and external parameters of the camera through the calibration algorithm, and use the internal and external parameters of the camera to dedistort the image.

[0097] In a specific embodiment, the calibration algorithm may be specifically a cv2.calibrateCamera function in OpenCV, which uses the internal and external parameters of the camera to perform dedistortion processing on the image. In OpenCV, the cv2.undistort function may be used to complete this operation.

[0098] 3) Using the known rotation matrix and translation vector, the dedistorted image is transformed from the camera coordinate system to the coordinate system of the design drawing to complete the spatial coordinate transformation.

[0099] In a specific embodiment, the dedistorted image is converted from the camera coordinate system to the coordinate system of the design drawing by affine transformation or perspective transformation.

[0100] The embodiments of the present invention utilize a high-definition industrial scanning camera for real-time acquisition and perform multiple operations such as image preprocessing, feature recognition, and position adjustment, thereby effectively solving the problem of insufficient current printing accuracy. It can ensure that no matter how the specifications and dimensions of the equipment change, how the dimensions of the textile fabric to be printed change, and how the printing parameter mode changes, images can be adaptively acquired and the quality of the final product can be guaranteed, breaking the limitations of current image capture designs.

[0101] It is understandable that the feature recognition performed by the AI ​​feature recognition algorithm in step 5 is a relatively mature technology, and its specific recognition process will not be described here.

[0102] It is also understandable that what needs to be supplemented about step 6 is that, in a specific embodiment, the position correction algorithm may select the IDW interpolation algorithm formula to implement image correction calculation.

[0103] In another specific embodiment, in order to cope with the large amount of computing requirements in the high-speed production mode, a GPU can be used for image processing, which is much more efficient than an ordinary CPU, ensuring real-time processing even under high-speed production conditions of 80 meters / hour.

[0104] The embodiment of the present invention calls RIP software and board printing software to call RIP software to process the deformed image data, produce PRT printing data, and then send the board printing software through the network port for real-time printing. It fully achieves real-time image acquisition, real-time image recognition, real-time image correction, and real-time image RIP, integrates multiple steps such as image acquisition, recognition, correction, RIP processing and printing, and forms an efficient virtual assembly line. All processes can run synchronously and in parallel, greatly improving production efficiency and the accuracy of pattern overprinting.

[0105] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A method for automatically overprinting a pattern on a textile fabric, characterized in that: include: Step 1: AI feature template construction: used to import the design drawing to be printed, record the resolution of the design drawing and set the coordinate system, and establish the AI ​​feature template to identify the position information of each target pattern from the AI ​​feature template; Step 2: Image acquisition and upload: The high-definition industrial scanning camera installed on the back of the corresponding carriage of the digital printing guide machine is used to collect real-time images of the textile fabrics entering the printing work area; Step 3, image analysis and acquisition adjustment: perform imaging state analysis on the image of the textile fabric, obtain the imaging score of the image, and determine whether the imaging score reaches the set imaging score. If not, confirm to adjust the light source setting index and return to step 2 until the set imaging score is reached. If reached, proceed to the next step; Step 4: Image preprocessing: transform the image into spatial coordinates to make it consistent with the position information and resolution of the design drawing; Step 5: Image feature recognition: Use AI feature recognition algorithm on the transformed image to find the AI ​​features and coordinate information of the pattern in the image; Step 6: Pattern position adjustment: Based on AI features and coordinate information, combined with the set coordinate system of the printed design drawing, the position of the target pattern outline in the design drawing is adjusted through the position correction algorithm to generate the printing pattern of the current physical textile fabric; Step 7, pattern format processing: RIP processing is performed on the print pattern of the current physical textile fabric to generate PRT format data; Step 8: Pattern printing execution: Send the PRT format data to the board printing software through the network port for real-time printing.

2. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 1, characterized in that: The step of establishing an AI feature template includes: Use a high-definition industrial linear array camera to scan the selected textile fabric samples to obtain the textile fabric return map; On the return image, use the image editing software to select the area of ​​the target pattern and record it as the AI ​​feature recognition area; Mark the position information of each AI feature recognition area on the return map, including the horizontal coordinate, vertical coordinate and rotation angle; Based on the marked return image and AI feature recognition area, the AI ​​model is preliminarily trained. After the preliminary training is completed, it is tested with actual samples, and the model parameters are adjusted according to the test results. The AI ​​model that passes the test is used as the AI ​​feature template.

3. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 1, characterized in that: A plurality of LED linear light sources are installed at the bottom of the high-definition industrial scanning camera, and the LED linear light sources move synchronously with the high-definition industrial scanning camera.

4. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 1, characterized in that: The imaging state analysis of the image of the textile fabric comprises: The image of the textile fabric is segmented by a sliding window to obtain segmented image blocks, and various imaging evaluation indicators of the textile fabric are confirmed; Compare each imaging evaluation index of the textile fabric with each set reference imaging evaluation index one by one, and if an imaging evaluation index is greater than the corresponding reference imaging evaluation index, mark the imaging evaluation index as a matching index; The number of matching indices and the number of imaging evaluation indices were counted, and the ratio between the two was recorded as the imaging matching ratio; If a certain imaging evaluation index is less than the corresponding reference imaging evaluation index, the imaging evaluation index is subtracted from the corresponding reference imaging evaluation index, and the absolute value of the difference is taken as the deviation value of the imaging evaluation index, and the imaging evaluation index is marked as the deviation index, and the matching correction factor is set, and the imaging matching ratio is corrected to obtain the corrected imaging matching ratio; The product of the corrected imaging coincidence ratio and the preset total imaging score is taken as the imaging score of the image.

5. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 4, characterized in that: The imaging evaluation indicators for confirming the textile fabric include: Detect each segmented image block by using an edge detection algorithm to obtain the image clarity of each segmented image block; Generate a color histogram of each segmented image block, extract the minimum pixel intensity and the maximum pixel intensity corresponding to the non-zero frequency in the color histogram, and extract the pixel intensity and the total number of pixels of each channel; The standard deviation of the pixel intensity of each channel is calculated to obtain the standard deviation of the pixel intensity corresponding to each segmented image block; Comparing the maximum pixel intensity with the minimum pixel intensity to obtain a dynamic range corresponding to each segmented image block, setting weights of the pixel intensity standard deviation and the dynamic range, and obtaining the contrast of each segmented image block by weighted summation; The energy value and entropy value of each segmented image block are extracted by using the gray level co-occurrence matrix, and the texture consistency of each segmented image block is calculated; Calculate the mean values ​​of image clarity, image contrast and texture consistency of each image block, set the weights of clarity, contrast and texture consistency, and obtain the imaging quality consistency through linear regression function statistics; The standard deviation of image clarity, image contrast and texture consistency of each image segmentation block is calculated respectively. The imaging quality uniformity is statistically analyzed in the same way as the imaging quality conformity. The imaging quality conformity and imaging quality uniformity are used as imaging evaluation indicators of textile fabrics.

6. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 5, characterized in that: The step of calculating the texture consistency of each segmented image block includes: The energy value and entropy value of each segmented image block are compared with the energy value and entropy value of the image reference, and the texture consistency of the image block whose energy value is lower than the energy value of the image reference and whose entropy value is higher than the entropy value of the image reference is recorded as 0; If the energy value of a segmented image block is higher than or equal to the energy value of the image reference and the entropy value is higher than the entropy value of the image reference, the energy value of the adjacent segmented image and the energy value of the image reference are denoted as ζ and ζ′ respectively. As the texture matching degree of the image block, denoted as α, ε is the set energy value matching compensation factor; If the energy value of a segmented image block is lower than the energy value of the image reference and the entropy value is lower than or equal to the entropy value of the image reference, the entropy value of the adjacent segmented image and the entropy value of the image reference are recorded as and Will As the texture matching degree of the image block, it is recorded as β, and ε′ is the set entropy value matching compensation factor; If the energy value of a segmented image block is higher than or equal to the energy value of the image reference and the entropy value is lower than or equal to the entropy value of the image reference, As the texture consistency δ of the image block, the texture consistency of each segmented image block is obtained.

7. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 4, characterized in that: The setting of the matching correction factor includes: The number of statistical deviation indicators is denoted as D0; Setting a first reference deviation value and a second reference deviation value; The deviation index with a deviation value less than or equal to the first reference deviation value is marked as a Class I deviation index; If the deviation value is greater than the first reference deviation value and less than or equal to the second reference deviation value, the deviation index is marked as a Class II deviation index; The deviation index with a deviation value greater than the set second reference deviation value is marked as a Class III deviation index; The number of type I deviation index, type II deviation index, and type III deviation index are counted and compared with the number of deviation indexes respectively. The ratios are recorded as type I deviation ratio, type II deviation ratio, and type III deviation ratio, respectively, and recorded as k a , k b and k c ; Set the weights of the deviation ratios of Class I, Class II, and Class III, denoted as f1, f2, and f3 respectively, and set the coincidence correction factor ψ.

8. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 4, characterized in that: The step of confirming and adjusting the light source setting index includes: If the number of deviation indicators is 2 or the number of deviation indicators is 1 and it is the image quality conformity, the overall adjustment is used as the adjustment method; if the number of deviation indicators is 1 and it is the image quality uniformity, the local adjustment is used as the adjustment method; When the adjustment mode is overall adjustment, extract the current set light intensity and the maximum light intensity of the LED linear light source corresponding to the high-definition industrial scanning camera, which are recorded as I0 and I max , set the overall light source adjustment factor Will As an indicator for adjusting the intensity of light sources and for adjusting light source settings; When the adjustment method is local adjustment, confirm the adjustment light source setting index of the local adjustment.

9. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 8, characterized in that: The step of setting the overall light source adjustment factor includes: If the number of deviation indicators is 2, the deviation compensation factor is recorded as φ0, and if the number of deviation indicators is 1, the deviation compensation factor is recorded as φ1, thereby obtaining the deviation compensation factor η, where η is φ0 or φ1, φ0>φ1; The image quality conformity and image quality uniformity are recorded as FH and JY respectively, and the overall light source adjustment factor is set. FH′ and JY′ are the set reference imaging quality conformity and reference imaging quality uniformity respectively, and e is a natural constant.

10. The method for automatically overprinting a pattern on a textile fabric as claimed in claim 9, characterized in that: The adjustment light source setting index for confirming the local adjustment includes: Extract the set irradiation area position of each LED linear light source carried by the high-definition industrial scanning camera, and use each image block located at the set irradiation area position of the LED linear light source as each associated image block of the corresponding LED linear light source; Calculate the imaging quality conformity of each LED linear light source corresponding to each associated image block, and record the LED linear light source of the associated image block whose imaging quality conformity is less than the reference imaging quality conformity as an adjusted light source; Set the overall light source adjustment factor Then, the adjusted light source intensity of the adjusted light source is confirmed in the same way as the adjusted light source setting index during the overall adjustment, and is used as the adjusted light source setting index for the local adjustment.

Images