Silk-screen pattern dislocation identification and processing method for injection molding product

Abstract

The invention relates to the technical field of image recognition, in particular to a silk-screen pattern dislocation recognition and processing method for an injection molding product. Obtaining a silk-screen pattern image, carrying out feature extraction, and outputting a bounding box set containing coordinate information; performing comparative analysis on the bounding box set and a preset template to generate a dislocation analysis report; generating correction parameters according to the dislocation analysis report, feeding back the correction parameters to the silk-screen equipment, and adjusting the printing position of the silk-screen pattern; and if dislocation still exists after correction, the injection molding product is marked as an unqualified product, and early warning monitoring treatment is carried out, so that the unqualified product is prevented from flowing into a subsequent production link. Through full-process automatic treatment, the accuracy and efficiency of screen printing pattern dislocation recognition are improved, and the intelligent degree in the production process of injection molding products is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, in particular to a method for recognizing and processing the misalignment of a silk-screen pattern of an injection-molded product. Background Art

[0002] In the production process of injection molded products, the printing accuracy of the silk-screen pattern directly affects the appearance quality and market competitiveness of the product. However, due to the influence of factors such as equipment accuracy, material properties and process parameters, the silk-screen pattern is prone to misalignment, such as center point offset, dimensional deviation or rotation angle deviation. Traditional misalignment detection methods mostly rely on manual visual inspection or simple image comparison technology, which has problems such as low efficiency, poor accuracy and difficulty in quantification, and cannot meet the needs of large-scale production. In addition, the existing technology lacks the ability to finely classify misalignment types and dynamically optimize correction parameters, making it difficult to fundamentally solve the misalignment problem. With the development of intelligent manufacturing technology, misalignment recognition methods based on image processing and machine learning have gradually become a research hotspot, but in the case of complex background interference, pattern overlap and the simultaneous existence of multiple types of misalignment, existing algorithms still face challenges such as inaccurate feature extraction, unstable bounding box regression and unsatisfactory correction effect. Therefore, there is an urgent need for an efficient, accurate and traceable method for identifying and processing screen printing pattern misalignment, which can realize the automation of the entire process from image acquisition, feature extraction, misalignment analysis to correction feedback, and has the functions of intelligent marking and early warning monitoring of unqualified products, so as to improve the production quality and process level of injection molded products. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a method for identifying and processing the misalignment of a silk-screen pattern of an injection-molded product.

[0004] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention discloses a method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product, comprising the following steps:

[0006] Obtain the silk screen pattern image of the injection molded product, perform feature extraction on the silk screen pattern image, and output a set of silk screen pattern bounding boxes containing coordinate information;

[0007] Comparing and analyzing the silk-screen pattern boundary box set with a preset template to obtain a dislocation analysis report of the injection molded product;

[0008] Generate correction parameters based on the misalignment analysis report and feed them back to the screen printing equipment to adjust the printing position of the screen printing pattern;

[0009] If misalignment still exists after correction, the injection molded product will be marked as unqualified and early warning monitoring will be carried out.

[0010] More specifically, feature extraction is performed on the silk screen pattern image, and a set of silk screen pattern bounding boxes containing coordinate information is output, specifically:

[0011] The silk-screen pattern image is divided into multiple local regions. Each local region is used as a node in the graph structure. The edge connection relationship is constructed based on the color similarity and spatial distance between adjacent local regions to generate the initial graph topology.

[0012] In the initial graph topology, the Euclidean distance between each node and the rest of the nodes is calculated, and the Euclidean distance between each node and the rest of the nodes is weightedly aggregated to obtain the feature vector of each node;

[0013] Based on the feature vector of each node, determine whether each node belongs to the target area of the silk screen pattern and output the confidence score of each node;

[0014] Mark nodes with confidence levels higher than a preset threshold as feature regions, and obtain the centroid coordinates and coverage of each feature region;

[0015] For each feature region, extract the node features of its corresponding node, and obtain the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region based on the node features;

[0016] The coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are collected to output a set of silk screen pattern bounding boxes.

[0017] More specifically, based on the feature vector of each node, it is determined whether each node belongs to the target area of the silk screen pattern, and the confidence of each node is output, specifically:

[0018] A multi-layer perceptron classifier was constructed. The input layer dimension of the classifier was consistent with the dimension of the feature vector, and the output layer was a binary classification structure, representing the probability of a node belonging to the target area or non-target area of the silk screen pattern. The hidden layer of the classifier used a nonlinear activation function to enhance the distinguishing ability of the feature expression.

[0019] The feature vector of each node is input into the classifier, and the probability value of the node belonging to the target area is calculated through forward propagation as the confidence value of the node;

[0020] If the confidence of a node is lower than the preset threshold, the node is determined to be a non-target area node and marked as a background area;

[0021] If the confidence of the node is higher than the preset threshold, the node is determined to be a target area node, marked as a candidate feature area, and its spatial location information is recorded.

[0022] More specifically, for each feature region, the node features of the corresponding node are extracted, and the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region are obtained according to the node features, specifically:

[0023] Extract the node features of the nodes corresponding to each feature region; use the weighted average method to integrate the node features of all adjacent nodes in each feature region to obtain the region-level feature vector;

[0024] A multi-layer perceptron regression network was constructed. The input layer dimension was consistent with the region-level feature vector dimension, and the output layer was a four-dimensional structure, representing the center point coordinates, width, and height of the minimum bounding rectangle. The hidden layer of the regression network used a nonlinear activation function.

[0025] The region-level feature vector is input into the regression network, and the initial center point, initial width, and initial height of the bounding box are calculated through forward propagation.

[0026] If the initial center point of the bounding box is within the valid area of the image, the coordinates of the initial center point are used as the center point coordinates; if the initial center point of the bounding box is at the edge or outside the image, the position of the initial center point is corrected according to the spatial position information of the feature area until the corrected center point is within the valid area of the image, and the coordinates of the corrected center point are used as the center point coordinates;

[0027] If the initial width or initial height is smaller than the actual coverage of the feature area, the bounding box size is enlarged according to the preset ratio; if the initial width or initial height is larger than the actual coverage of the feature area, the bounding box size is reduced according to the preset ratio until both the initial width and initial height are within the actual coverage of the feature area;

[0028] Finally, the coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are obtained, including the center point coordinates, width and height.

[0029] More specifically, the silk screen pattern boundary box set is compared and analyzed with a preset template to obtain a misalignment analysis report of the injection molded product, specifically:

[0030] Spatially align the preset template with the bounding box set of the silk screen pattern, and use affine transformation to convert the template coordinates to the same coordinate system as the bounding box set;

[0031] Traverse the set of silk screen pattern bounding boxes and match each bounding box with the corresponding area in the template; when the center point of the bounding box is within the preset range of the preset template area, the two are considered to be matched successfully and a mapping relationship is established;

[0032] For each successfully matched bounding box, calculate the size deviation parameters between it and the corresponding area of the template, including center point coordinate deviation, width deviation, and height deviation. If the rotation angle of the bounding box is inconsistent with the template area, additional angle deviation is calculated.

[0033] Compare the calculated size deviation parameters with the preset deviation thresholds; if all size deviation parameters are less than the preset deviation thresholds, the bounding box is determined to be a normal box; if any size deviation parameter exceeds the preset deviation threshold, the bounding box is marked as a misaligned box, and the specific misalignment type, value, and position are recorded;

[0034] Summarize the size deviation parameters of all bounding boxes and generate a misalignment analysis report, including the normal box, misaligned box, and the misalignment type, value, and position of the misaligned box.

[0035] More specifically, correction parameters are generated based on the misalignment analysis report and fed back to the screen printing equipment to adjust the printing position of the screen printing pattern, specifically:

[0036] According to the misalignment analysis report, the misalignment types of the misalignment frame are divided into three categories: center point offset, size deviation, and rotation angle deviation. The specific value and position information of each type of misalignment are extracted as the basic data for the correction parameter calculation.

[0037] For the misaligned frame with center point offset, the coordinate difference between it and the center point of the area corresponding to the preset template is calculated and converted into the displacement adjustment of the screen printing equipment; for the misaligned frame with size deviation, the width difference and height difference between it and the area corresponding to the preset template are calculated and converted into the scaling ratio of the screen printing equipment; for the misaligned frame with rotation angle deviation, the angle difference between it and the area corresponding to the preset template is calculated and converted into the rotation adjustment of the screen printing equipment; the generated correction parameters;

[0038] Apply the generated correction parameters to the real-time simulation model of the silk screen pattern, regenerate the virtual bounding box set, and compare it with the preset template;

[0039] If the deviation parameters between the virtual bounding box and the template are all less than the preset threshold, the correction parameters are considered valid; if there is still a deviation, the correction parameters are readjusted until the accuracy requirements are met;

[0040] The verified correction parameters are fed back to the screen printing equipment through the communication interface, driving the equipment to perform displacement, scaling and rotation adjustments. If the equipment feeds back an adjustment completion signal, the screen printing pattern image of the injection molded product is re-acquired and a second misalignment detection is performed to ensure the correction effect.

[0041] More specifically, if misalignment still exists after correction, the injection molded product is marked as unqualified and early warning monitoring is carried out, specifically:

[0042] Re-acquire the silk screen pattern image of the corrected injection molded product, perform feature extraction and bounding box set generation, and compare and analyze it with the preset template;

[0043] If there are still misaligned boxes in the corrected bounding box set, the injection molded product is determined to be unqualified. At the same time, the specific misalignment value, location information and correction history of the unqualified product are recorded.

[0044] Injection molded products marked as defective are transferred to an isolated area through automated sorting equipment; if the sorting equipment detects that the number of defective products exceeds a preset threshold within a preset time period, an alarm signal is triggered.

[0045] The following steps are also included:

[0046] Integrate the misalignment recognition results, correction parameters, and early warning monitoring status of each injection molded product, including bounding box set, misalignment type, misalignment value, correction adjustment amount, defective product marking information, and alarm signal;

[0047] Based on the integrated detection data, a visual display interface is designed, including a dislocation distribution diagram, a correction parameter comparison table, and a warning monitoring status statistics panel;

[0048] Update the test data and correction parameters of each injection molded product to the display interface in real time and synchronize them to the central database through the network; if the test data changes, the interface content will be automatically refreshed;

[0049] The test data and correction parameters are stored in the quality traceability database, a unique identification code is generated for each injection molded product, and an index relationship is established. If you need to query the test record of a specific product, you can quickly retrieve relevant data through the identification code, including misalignment recognition results, correction parameters and early warning monitoring status;

[0050] The stored test data and correction parameters are encrypted and access permissions are set; if the operator needs to modify or export data, they must pass identity authentication and permission review to prevent data leakage or tampering.

[0051] The node features include pixel intensity, texture statistics and spatial position information.

[0052] The second aspect of the present invention discloses a system for identifying and processing the misalignment of silk screen patterns on injection molded products. The system includes a memory and a processor. The memory stores a program for identifying and processing the misalignment of silk screen patterns. When the program for identifying and processing the misalignment of silk screen patterns is executed by the processor, the steps of any one of the methods for identifying and processing the misalignment of silk screen patterns are implemented.

[0053] The present invention solves the technical deficiencies in the background art and has the following beneficial effects: acquiring a silk screen pattern image and performing feature extraction to output a bounding box set containing coordinate information; comparing and analyzing the bounding box set with a preset template to generate a dislocation analysis report that identifies the type, value, and location of the dislocation; generating correction parameters based on the dislocation analysis report and feeding them back to the silk screen equipment to adjust the printing position of the silk screen pattern; if dislocation still exists after correction, marking the injection molded product as defective and performing early warning monitoring to prevent defective products from entering subsequent production links. The present invention improves the accuracy and efficiency of silk screen pattern dislocation identification through full-process automation, effectively enhancing the level of intelligence in the injection molded product production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, they can also obtain drawings of other embodiments based on these drawings.

[0055] Figure 1 A flow chart of a first method for identifying and processing the misalignment of a silk-screen pattern on an injection-molded product;

[0056] Figure 2 A flow chart of a second method of a method for identifying and processing misalignment of a silk-screen pattern on an injection molded product;

[0057] Figure 3 The following is a system block diagram of a system for identifying and processing the misalignment of silk-screen patterns on injection-molded products. DETAILED DESCRIPTION

[0058] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0060] like Figure 1 As shown, the first aspect of the present invention discloses a method for identifying and processing the misalignment of a silk screen pattern on an injection molded product, comprising the following steps:

[0061] S102, obtaining a silk screen pattern image of the injection molded product, performing feature extraction on the silk screen pattern image, and outputting a silk screen pattern bounding box set containing coordinate information;

[0062] S104: Compare and analyze the silk-screen pattern boundary frame set with a preset template to obtain a dislocation analysis report of the injection molded product;

[0063] S106. Generating correction parameters based on the misalignment analysis report and feeding them back to the screen printing equipment to adjust the printing position of the screen printing pattern;

[0064] S108. If misalignment still exists after correction, the injection molded product is marked as unqualified and early warning monitoring is performed.

[0065] It should be noted that the silk screen pattern image is acquired and feature extraction is performed, outputting a bounding box set containing coordinate information; the bounding box set is compared and analyzed with a preset template to generate a dislocation analysis report, clarifying the dislocation type, value, and position; correction parameters are generated based on the dislocation analysis report and fed back to the silk screen equipment to adjust the printing position of the silk screen pattern; if dislocation still exists after correction, the injection molded product is marked as unqualified and early warning monitoring is performed to prevent unqualified products from entering subsequent production links. The present invention improves the accuracy and efficiency of silk screen pattern dislocation recognition through full process automation, effectively improving the level of intelligence in the injection molded product production process.

[0066] More specifically, feature extraction is performed on the silk screen pattern image, and a set of silk screen pattern bounding boxes containing coordinate information is output, specifically:

[0067] The silk-screen pattern image is divided into multiple local regions. Each local region is used as a node in the graph structure. The edge connection relationship is constructed based on the color similarity and spatial distance between adjacent local regions to generate the initial graph topology.

[0068] Calculate the Euclidean distance between each node and the rest of the nodes in the initial graph topology, perform weighted aggregation on the Euclidean distance between each node and the rest of the nodes, and obtain the feature vector of each node;

[0069] Based on the feature vector of each node, determine whether each node belongs to the target area of the silk screen pattern and output the confidence score of each node;

[0070] Mark nodes with confidence levels higher than a preset threshold as feature regions, and obtain the centroid coordinates and coverage of each feature region;

[0071] For each feature region, extract the node features of its corresponding node, and obtain the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region based on the node features;

[0072] The coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are collected to output a set of silk screen pattern bounding boxes.

[0073] In summary, this method uses graph structure to model the local correlation of silk screen patterns, combines dynamic feature aggregation with bounding box fusion mechanism, effectively solves the problems of complex background interference and pattern overlap, and effectively improves detection accuracy and robustness.

[0074] More specifically, based on the feature vector of each node, it is determined whether each node belongs to the target area of the silk screen pattern, and the confidence of each node is output, such as Figure 2 As shown, specifically:

[0075] S202. Construct a multi-layer perceptron classifier. The input layer dimension of the classifier is consistent with the dimension of the feature vector, and the output layer is a binary classification structure, which respectively represents the probability of the node belonging to the target area or the non-target area of the silk screen pattern. The hidden layer of the classifier uses a nonlinear activation function to enhance the distinguishing ability of the feature expression.

[0076] It should be noted that the multilayer perceptron (MLP) classifier is a feedforward classification model based on artificial neural networks. Its core structure consists of an input layer, a hidden layer, and an output layer. The input layer has the same dimensions as the node feature vector and is used to receive feature data. The hidden layer transforms and combines input features using nonlinear activation functions (such as ReLU or Sigmoid) to extract higher-level feature expressions, thereby enhancing the model's discriminative ability. The output layer typically uses a Softmax or Sigmoid function to map the hidden layer output to a probability value, which is used to represent the confidence that the node belongs to the target area or non-target area. By stacking multiple layers, the MLP classifier can capture the complex nonlinear relationships between features, thereby achieving accurate classification of the target area.

[0077] S204: Input the feature vector of each node into the classifier, and calculate the probability value of the node belonging to the target area through forward propagation as the confidence value of the node;

[0078] It should be noted that the feature vector of the node is input into the input layer of the multi-layer perceptron classifier, and is processed by weighted summation and nonlinear activation function, and then passed to the hidden layer layer by layer to extract high-order feature expressions; then, the output of the hidden layer is passed to the output layer, and the probability value of the node belonging to the target area and the non-target area is calculated by the Softmax or Sigmoid function; finally, the probability value of the target area is used as the confidence of the node.

[0079] S206: If the confidence level of the node is lower than a preset threshold, the node is determined to be a non-target area node and marked as a background area;

[0080] S208: If the confidence level of the node is higher than a preset threshold, the node is determined to be a target region node, marked as a candidate feature region, and its spatial position information is recorded.

[0081] In summary, this method achieves accurate identification and marking of the target area of the silk-screen pattern, laying the foundation for subsequent misalignment identification and correction. At the same time, it has strong robustness and adaptability, and can effectively cope with challenges such as complex backgrounds and pattern overlap, thereby improving the quality and efficiency of silk-screen pattern inspection of injection molded products.

[0082] More specifically, for each feature region, the node features of the corresponding node are extracted, and the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region are obtained according to the node features, specifically:

[0083] Extract the node features of the nodes corresponding to each feature region; use the weighted average method to integrate the node features of all adjacent nodes in each feature region to obtain the region-level feature vector;

[0084] It should be noted that all nodes in the feature area are traversed to extract their node feature vectors, including pixel intensity, texture statistics and spatial position information; weight values are calculated based on the color similarity and spatial distance between adjacent nodes to ensure that more relevant nodes are given higher weights when integrating features; then, the feature vector of each node is multiplied by its weight value, and the weighted feature vectors of all adjacent nodes are summed; finally, the sum is divided by the total weight to obtain the region-level feature vector for subsequent bounding box regression and optimization.

[0085] A multi-layer perceptron regression network was constructed. The input layer dimension was consistent with the region-level feature vector dimension, and the output layer was a four-dimensional structure, representing the center point coordinates, width, and height of the minimum bounding rectangle. The hidden layer of the regression network used a nonlinear activation function.

[0086] It should be noted that the multilayer perceptron (MLP) regression network is a feedforward regression model based on an artificial neural network. Its core structure consists of an input layer, a hidden layer, and an output layer. The input layer, whose dimensions match those of the region-level feature vector, receives feature data. The hidden layer transforms and combines input features using nonlinear activation functions (such as ReLU or Tanh) to extract higher-level feature expressions, thereby enhancing the model's fitting capabilities. The output layer is a four-dimensional structure representing the center point coordinates (x, y), width (w), and height (h) of the minimum enclosing rectangular bounding box, which directly outputs the regression results. By stacking these multiple layers, the MLP regression network can capture the complex nonlinear relationships between features, thereby accurately predicting the bounding box parameters.

[0087] The region-level feature vector is input into the regression network, and the initial center point, initial width, and initial height of the bounding box are calculated through forward propagation.

[0088] It should be noted that the region-level feature vector is input into the input layer of the regression network. It is then processed through weighted summation and nonlinear activation functions, and then passed to the hidden layer layer by layer to extract high-order feature expressions. The output of the hidden layer is passed to the output layer, where the initial center point coordinates, initial width, and initial height of the bounding box are calculated through linear transformation. Finally, the calculation results of the output layer are used as the initial parameters of the bounding box for subsequent calibration and optimization to ensure that the bounding box closely matches the actual coverage of the feature region.

[0089] If the initial center point of the bounding box is within the valid area of the image, the coordinates of the initial center point are used as the center point coordinates; if the initial center point of the bounding box is at the edge or outside the image, the position of the initial center point is corrected according to the spatial position information of the feature area until the corrected center point is within the valid area of the image, and the coordinates of the corrected center point are used as the center point coordinates;

[0090] It should be noted that the minimum distance between the initial center point and the edge of the image is calculated to determine its deviation direction; then, according to the centroid coordinates and coverage range of the feature area, a correction vector is generated, and the initial center point is moved in the opposite direction of the deviation direction until it enters the valid area of the image; then, it is verified whether the corrected center point is within the actual coverage range of the feature area. If not, the correction vector is further adjusted; finally, the coordinates of the corrected center point are used as the final center point coordinates.

[0091] If the initial width or height is smaller than the actual coverage of the feature area, the bounding box size is enlarged according to a preset ratio; if the initial width or height is larger than the actual coverage of the feature area, the bounding box size is reduced according to a preset ratio. This is done until both the initial width and the initial height are within the actual coverage of the feature area, and the width and height of the minimum bounding box are obtained.

[0092] Finally, the coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are obtained, including the center point coordinates, width and height.

[0093] In summary, this method improves the accuracy and stability of bounding box positioning through feature integration and regression network optimization, effectively solves the problems of bounding box size deviation and position offset, and ensures a high degree of matching between the bounding box and the feature area, thereby quickly obtaining the coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area.

[0094] More specifically, the silk screen pattern boundary box set is compared and analyzed with a preset template to obtain a misalignment analysis report of the injection molded product, specifically:

[0095] Spatially align the preset template with the bounding box set of the silk screen pattern, and use affine transformation to convert the template coordinates to the same coordinate system as the bounding box set;

[0096] A preset template is a standardized reference model used for comparison against a set of bounding boxes for a silkscreen pattern. It contains information such as the ideal position, size, and rotation angle of the silkscreen pattern. Typically generated based on a design drawing or a qualified sample, a preset template accurately reflects the correct layout of the silkscreen pattern on the injection molded part.

[0097] Traverse the set of silk screen pattern bounding boxes and match each bounding box with the corresponding area in the template; when the center point of the bounding box is within the preset range of the preset template area, the two are considered to be matched successfully and a mapping relationship is established;

[0098] For each successfully matched bounding box, calculate the size deviation parameters between it and the corresponding area of the template, including center point coordinate deviation, width deviation, and height deviation. If the rotation angle of the bounding box is inconsistent with the template area, additional angle deviation is calculated.

[0099] Compare the calculated size deviation parameters with the preset deviation thresholds; if all size deviation parameters are less than the preset deviation thresholds, the bounding box is determined to be a normal box; if any size deviation parameter exceeds the preset deviation threshold, the bounding box is marked as a misaligned box, and the specific misalignment type, value, and position are recorded;

[0100] Summarize the size deviation parameters of all bounding boxes and generate a misalignment analysis report, including the normal box, misaligned box, and the misalignment type, value, and position of the misaligned box.

[0101] In summary, this method achieves comprehensive and accurate identification of silk screen pattern misalignment through spatial alignment, deviation calculation and misalignment classification, effectively solving the detection problems of complex backgrounds and multiple types of misalignment. At the same time, it provides a basis for subsequent correction parameter generation and equipment adjustment, and improves the level of intelligent control in the production process of injection molded products.

[0102] More specifically, correction parameters are generated based on the misalignment analysis report and fed back to the screen printing equipment to adjust the printing position of the screen printing pattern, specifically:

[0103] According to the misalignment analysis report, the misalignment types of the misalignment frame are divided into three categories: center point offset, size deviation, and rotation angle deviation. The specific value and position information of each type of misalignment are extracted as the basic data for the correction parameter calculation.

[0104] For the misaligned frame with center point offset, the coordinate difference between it and the center point of the area corresponding to the preset template is calculated and converted into the displacement adjustment of the screen printing equipment; for the misaligned frame with size deviation, the width difference and height difference between it and the area corresponding to the preset template are calculated and converted into the scaling ratio of the screen printing equipment; for the misaligned frame with rotation angle deviation, the angle difference between it and the area corresponding to the preset template is calculated and converted into the rotation adjustment of the screen printing equipment; the generated correction parameters;

[0105] Apply the generated correction parameters to the real-time simulation model of the silk screen pattern, regenerate the virtual bounding box set, and compare it with the preset template;

[0106] The real-time silkscreen pattern simulation model is a 3D virtual simulation system generated using industrial software (such as SolidWorks and UG) to simulate the effect of a corrected silkscreen pattern. The model dynamically generates a set of virtual bounding boxes by applying the generated correction parameters (including displacement adjustment, scaling, and rotation adjustment) to a preset template. This virtual bounding box is then compared with the actual silkscreen pattern to verify the validity of the correction parameters.

[0107] If the deviation parameters between the virtual bounding box and the template are all less than the preset threshold, the correction parameters are considered valid; if there is still a deviation, the correction parameters are readjusted until the accuracy requirements are met;

[0108] The verified correction parameters are fed back to the screen printing equipment through the communication interface, driving the equipment to perform displacement, scaling and rotation adjustments. If the equipment feeds back an adjustment completion signal, the screen printing pattern image of the injection molded product is re-acquired and a second misalignment detection is performed to ensure the correction effect.

[0109] In summary, this method achieves precise correction of the printing position of the silk-screen pattern through misalignment classification, parameter calculation and simulation verification, effectively solves the adjustment problem of complex misalignment types, improves the production quality and process level of injection molded products, and at the same time has strong automation and intelligent capabilities, which can adapt to the needs of large-scale production.

[0110] More specifically, if misalignment still exists after correction, the injection molded product is marked as unqualified and early warning monitoring is carried out, specifically:

[0111] Re-acquire the silk screen pattern image of the corrected injection molded product, perform feature extraction and bounding box set generation, and compare and analyze it with the preset template;

[0112] If there are still misaligned boxes in the corrected bounding box set, the injection molded product is determined to be unqualified. At the same time, the specific misalignment value, location information and correction history of the unqualified product are recorded.

[0113] Injection molded products marked as defective are transferred to an isolated area through automated sorting equipment; if the sorting equipment detects that the number of defective products exceeds a preset threshold within a preset time period, an alarm signal is triggered.

[0114] In summary, through secondary inspection and automated sorting, we can accurately identify and isolate defective products, effectively preventing them from flowing into subsequent production links. At the same time, through the early warning monitoring mechanism, we can promptly detect production anomalies, avoid the production of large quantities of defective products in the automated production process, and reduce scrap costs.

[0115] The following steps are also included:

[0116] Integrate the misalignment recognition results, correction parameters, and early warning monitoring status of each injection molded product, including bounding box set, misalignment type, misalignment value, correction adjustment amount, defective product marking information, and alarm signal;

[0117] Based on the integrated detection data, a visual display interface is designed, including a dislocation distribution diagram, a correction parameter comparison table, and a warning monitoring status statistics panel;

[0118] Update the test data and correction parameters of each injection molded product to the display interface in real time and synchronize them to the central database through the network; if the test data changes, the interface content will be automatically refreshed;

[0119] The test data and correction parameters are stored in the quality traceability database, a unique identification code is generated for each injection molded product, and an index relationship is established. If you need to query the test record of a specific product, you can quickly retrieve relevant data through the identification code, including misalignment recognition results, correction parameters and early warning monitoring status;

[0120] The stored test data and correction parameters are encrypted and access permissions are set; if the operator needs to modify or export data, they must pass identity authentication and permission review to prevent data leakage or tampering.

[0121] It should be noted that the bounding box set, misalignment type, misalignment value, correction adjustment amount, defective product marking information, and alarm signal are integrated to form complete inspection data. Based on the integrated data, a visual display interface is designed, including a misalignment distribution diagram, a correction parameter comparison table, and an early warning monitoring status statistics panel, which are updated in real time and synchronized to the central database. Then, the inspection data and correction parameters are stored in the quality traceability database, a unique identification code is generated for each injection molded product, and an index relationship is established to support rapid retrieval and traceability. Finally, the stored data is encrypted and access rights are set to ensure data security and integrity. Through data integration and visual display, real-time monitoring and efficient management of production quality are achieved. At the same time, through quality traceability and data encryption, the traceability and security of inspection data are ensured, providing comprehensive and reliable data support for process optimization and quality control, and improving the production management level and quality assurance capabilities of injection molded products.

[0122] The node features include pixel intensity, texture statistics and spatial position information.

[0123] In this embodiment, the method for identifying and processing the misalignment of a silk screen pattern further includes the following steps:

[0124] Use infrared thermal imaging technology to capture the temperature field distribution data of injection molded products in real time and obtain the temperature gradient data of each area;

[0125] Establish a temperature gradient model based on the temperature gradient data, and import the temperature gradient model into the finite element deformation simulation model of the injection molded product;

[0126] Dynamically select the shrinkage calculation mode based on the phase transition temperature of the injection molded product material: when the product temperature is higher than the glass transition temperature, the viscoelastic constitutive equation is used to calculate the molecular chain relaxation effect; when the temperature is lower than the crystallization point, the lattice deformation model is switched to calculate the anisotropic shrinkage, and the final output is the thermal shrinkage compensation vector field of the screen printing pattern area;

[0127] It should be noted that when the product temperature is above the glass transition temperature, the viscoelastic parameters of the injection molded product in the high-temperature region, including relaxation time and elastic modulus, are extracted based on the temperature field distribution data. The relaxation behavior of the molecular chain at high temperatures is simulated using the viscoelastic constitutive equation to calculate the thermal shrinkage of the material. When the temperature is below the crystallization point, the anisotropic shrinkage is calculated using the lattice deformation model: the lattice structure of the crystallization region is determined based on the temperature gradient data, and the lattice deformation parameters, including the lattice constant and anisotropy coefficient, are extracted. The lattice deformation model is used to calculate the shrinkage of the material in different directions. Finally, the thermal shrinkage of the high and low temperature regions is integrated to generate a thermal shrinkage compensation vector field for the screen printing pattern area, providing accurate data support for the subsequent dynamic adjustment of the deformed template.

[0128] The screen printing pattern area is divided into a rigid translation area, an elastic distortion area, and a free deformation area according to the thermal shrinkage compensation vector field. When the displacement difference between adjacent areas exceeds the preset tolerance standard, the thermal shrinkage compensation is vector-superimposed with the preset template to generate a dynamically adjusted deformation template.

[0129] It should be noted that the displacement characteristics of each area in the vector field are analyzed. If the displacement is uniform and the direction is consistent, it is divided into a rigid translation area; if the displacement changes continuously and the direction changes gradually, it is divided into an elastic distortion area; if the displacement is irregular and the direction is random, it is divided into a free deformation area; then, the displacement difference between adjacent areas is detected. If it exceeds the preset tolerance standard, the thermal shrinkage compensation amount is vector-superimposed with the preset template to generate a dynamically adjusted deformation template; finally, the matching degree of the deformation template and the silk-screen pattern is verified to ensure that it can accurately reflect the actual deformation state of the injection molded product.

[0130] If the matching degree between the deformed template and the preset template is lower than a preset threshold, the preset template is replaced with the deformed template.

[0131] In summary, this method accurately identifies the thermal shrinkage behavior of injection-molded products through the cross-domain collaboration of thermodynamic models and image processing technology, thereby dynamically adjusting the silk-screen pattern template, improving printing accuracy and adaptability, and solving the technical pain point that traditional image comparison cannot predict subsequent deformation, thereby effectively improving product qualification rate and process accuracy.

[0132] like Figure 3 As shown, the second aspect of the present invention discloses a system 6 for identifying and processing the misalignment of a silk screen pattern of an injection molded product. The system includes a memory 41 and a processor 52. The memory 41 stores a program for identifying and processing a silk screen pattern misalignment. When the program for identifying and processing a silk screen pattern misalignment is executed by the processor 52, the steps of any one of the methods for identifying and processing a silk screen pattern misalignment are implemented.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0134] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0135] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0136] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0137] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0138] The above are only specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for identifying and processing the misalignment of silk screen patterns on injection molded products, characterized in that: The following steps are involved: Obtain the silk screen pattern image of the injection molded product, perform feature extraction on the silk screen pattern image, and output a set of silk screen pattern bounding boxes containing coordinate information; Comparing and analyzing the silk-screen pattern boundary box set with a preset template to obtain a dislocation analysis report of the injection molded product; Generate correction parameters based on the misalignment analysis report and feed them back to the screen printing equipment to adjust the printing position of the screen printing pattern; If misalignment still exists after correction, the injection molded product will be marked as unqualified and early warning monitoring will be carried out.

2. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 1, characterized in that: Perform feature extraction on the silk screen pattern image and output a set of silk screen pattern bounding boxes containing coordinate information, specifically: The silk-screen pattern image is divided into multiple local regions. Each local region is used as a node in the graph structure. The edge connection relationship is constructed based on the color similarity and spatial distance between adjacent local regions to generate the initial graph topology. Calculate the Euclidean distance between each node and the rest of the nodes in the initial graph topology, perform weighted aggregation on the Euclidean distance between each node and the rest of the nodes, and obtain the feature vector of each node; Based on the feature vector of each node, determine whether each node belongs to the target area of the silk screen pattern and output the confidence score of each node; Mark nodes with confidence levels higher than a preset threshold as feature regions, and obtain the centroid coordinates and coverage of each feature region; For each feature region, extract the node features of its corresponding node, and obtain the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region based on the node features; The coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are collected to output a set of silk screen pattern bounding boxes.

3. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 2, wherein: Based on the feature vector of each node, determine whether each node belongs to the target area of the silk screen pattern and output the confidence of each node, specifically: A multi-layer perceptron classifier was constructed. The input layer dimension of the classifier was consistent with the dimension of the feature vector, and the output layer was a binary classification structure, representing the probability of a node belonging to the target area or non-target area of the silk screen pattern. The hidden layer of the classifier used a nonlinear activation function to enhance the distinguishing ability of the feature expression. The feature vector of each node is input into the classifier, and the probability value of the node belonging to the target area is calculated through forward propagation as the confidence value of the node; If the confidence of a node is lower than the preset threshold, the node is determined to be a non-target area node and marked as a background area; If the confidence of the node is higher than the preset threshold, the node is determined to be a target area node, marked as a candidate feature area, and its spatial location information is recorded.

4. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 2, wherein: For each feature region, extract the node features of its corresponding node, and obtain the coordinate parameters of the minimum circumscribed rectangular bounding box of the corresponding feature region based on the node features, specifically: Extract the node features of the nodes corresponding to each feature region; use the weighted average method to integrate the node features of all adjacent nodes in each feature region to obtain the region-level feature vector; A multi-layer perceptron regression network was constructed. The input layer dimension was consistent with the region-level feature vector dimension, and the output layer was a four-dimensional structure, representing the center point coordinates, width, and height of the minimum bounding rectangle. The hidden layer of the regression network used a nonlinear activation function. The region-level feature vector is input into the regression network, and the initial center point, initial width, and initial height of the bounding box are calculated through forward propagation. If the initial center point of the bounding box is within the valid area of the image, the coordinates of the initial center point are used as the center point coordinates; if the initial center point of the bounding box is at the edge or outside the image, the position of the initial center point is corrected according to the spatial position information of the feature area until the corrected center point is within the valid area of the image, and the coordinates of the corrected center point are used as the center point coordinates; If the initial width or initial height is smaller than the actual coverage of the feature area, the bounding box size is enlarged according to the preset ratio; if the initial width or initial height is larger than the actual coverage of the feature area, the bounding box size is reduced according to the preset ratio until both the initial width and initial height are within the actual coverage of the feature area; Finally, the coordinate parameters of the minimum circumscribed rectangular bounding box of each feature area are obtained, including the center point coordinates, width and height.

5. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 1, characterized in that: Compare and analyze the silk screen pattern boundary box set with the preset template to obtain a dislocation analysis report of the injection molded product, specifically: Spatially align the preset template with the bounding box set of the silk screen pattern, and use affine transformation to convert the template coordinates to the same coordinate system as the bounding box set; Traverse the set of silk screen pattern bounding boxes and match each bounding box with the corresponding area in the template; when the center point of the bounding box is within the preset range of the preset template area, the two are considered to be matched successfully and a mapping relationship is established; For each successfully matched bounding box, calculate the size deviation parameters between it and the corresponding area of the template, including center point coordinate deviation, width deviation, and height deviation. If the rotation angle of the bounding box is inconsistent with the template area, additional angle deviation is calculated. Compare the calculated size deviation parameters with the preset deviation thresholds; if all size deviation parameters are less than the preset deviation thresholds, the bounding box is determined to be a normal box; if any size deviation parameter exceeds the preset deviation threshold, the bounding box is marked as a misaligned box, and the specific misalignment type, value, and position are recorded; Summarize the size deviation parameters of all bounding boxes and generate a misalignment analysis report, including the normal box, misaligned box, and the misalignment type, value, and position of the misaligned box.

6. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 1, characterized in that: Generate correction parameters based on the misalignment analysis report and feed them back to the screen printing equipment to adjust the printing position of the screen printing pattern. Specifically: According to the misalignment analysis report, the misalignment types of the misalignment frame are divided into three categories: center point offset, size deviation, and rotation angle deviation. The specific value and position information of each type of misalignment are extracted as the basic data for the correction parameter calculation. For the misaligned frame with center point offset, the coordinate difference between it and the center point of the area corresponding to the preset template is calculated and converted into the displacement adjustment of the screen printing equipment; for the misaligned frame with size deviation, the width difference and height difference between it and the area corresponding to the preset template are calculated and converted into the scaling ratio of the screen printing equipment; for the misaligned frame with rotation angle deviation, the angle difference between it and the area corresponding to the preset template is calculated and converted into the rotation adjustment of the screen printing equipment; the generated correction parameters; Apply the generated correction parameters to the real-time simulation model of the silk screen pattern, regenerate the virtual bounding box set, and compare it with the preset template; If the deviation parameters between the virtual bounding box and the template are all less than the preset threshold, the correction parameters are considered valid; if there is still a deviation, the correction parameters are readjusted until the accuracy requirements are met; The verified correction parameters are fed back to the screen printing equipment through the communication interface, driving the equipment to perform displacement, scaling and rotation adjustments. If the equipment feeds back an adjustment completion signal, the screen printing pattern image of the injection molded product is re-acquired and a second misalignment detection is performed to ensure the correction effect.

7. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 6, characterized in that: If misalignment still exists after correction, the injection molded product will be marked as unqualified and early warning monitoring will be carried out, specifically: Re-acquire the silk screen pattern image of the corrected injection molded product, perform feature extraction and bounding box set generation, and compare and analyze it with the preset template; If there are still misaligned boxes in the corrected bounding box set, the injection molded product is determined to be unqualified. At the same time, the specific misalignment value, location information and correction history of the unqualified product are recorded. Transfer the injection molded products marked as unqualified to the isolation area through automated sorting equipment; If the sorting equipment detects that the number of defective products exceeds a preset threshold within a preset time period, an alarm signal is triggered.

8. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 1, wherein: The following steps are also included: Integrate the misalignment recognition results, correction parameters, and early warning monitoring status of each injection molded product, including bounding box set, misalignment type, misalignment value, correction adjustment amount, defective product marking information, and alarm signal; Based on the integrated detection data, a visual display interface is designed, including a dislocation distribution diagram, a correction parameter comparison table, and a warning monitoring status statistics panel; Update the test data and correction parameters of each injection molded product to the display interface in real time and synchronize them to the central database through the network; if the test data changes, the interface content will be automatically refreshed; The test data and correction parameters are stored in the quality traceability database, a unique identification code is generated for each injection molded product, and an index relationship is established. If you need to query the test record of a specific product, you can quickly retrieve relevant data through the identification code, including misalignment recognition results, correction parameters and early warning monitoring status; The stored test data and correction parameters are encrypted and access permissions are set; if the operator needs to modify or export data, they must pass identity authentication and permission review to prevent data leakage or tampering.

9. The method for identifying and processing the misalignment of a silk-screen pattern on an injection molded product according to claim 2, wherein: The node features include pixel intensity, texture statistics and spatial position information.

10. A system for identifying and processing the misalignment of silk-screen patterns on injection-molded products, characterized in that: The silk screen pattern misalignment recognition and processing system includes a memory and a processor, wherein the memory stores a silk screen pattern misalignment recognition and processing method program. When the silk screen pattern misalignment recognition and processing method program is executed by the processor, the steps of the silk screen pattern misalignment recognition and processing method as described in any one of claims 1 to 9 are implemented.

Images