System and method for inspecting silk-screen quality of PET (Polyethylene Terephthalate) substrate

Abstract

The invention provides a PET substrate silk-screen quality inspection system and method, and belongs to the technical field of silk-screen quality inspection, and the PET substrate silk-screen quality inspection system comprises a membrane material flattening system, a control module, a grabbing module, a sensor module and an image acquisition module. The control module controls the film material flattening system to flatten the film material, and the grabbing module grabs the PET substrate and places the PET substrate on the detection platform; the control module detects the position of the PET substrate through the sensor module; the control module obtains image information of the PET substrate through the image acquisition module; the control module carries out image preprocessing on the collected image information; the control module detects the shape and the thickness of the ink pattern on the PET substrate; the control module calculates the distance between the printing ink area and the edge of the PET substrate; the control module is used for performing appearance inspection on the ink area on the PET substrate; the control module generates an inspection report. According to the invention, the efficiency of silk-screen quality detection is improved, and the cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of screen printing quality inspection. Specifically, the present invention relates to a screen printing quality inspection system and inspection method for PET substrates. Background Art

[0002] Most traditional methods for detecting PET inks after screen printing rely on manual inspection or manual tools (such as calipers and thickness gauges) to measure the size, thickness, etc. of the ink area. These methods are not only time-consuming but also difficult to ensure high precision. Especially in mass production, manual operation often leads to inconsistent measurement results, and the manual inspection process is extremely cumbersome and inefficient.

[0003] Chinese Patent No. 106587650A provides a method for reworking local defects in screen printing on glass, including the following steps: marking a glass panel with ink dots detected on the screen printing hole positions as defective products; adjusting the coding shape of the laser head of the laser coding machine so that the coding shape matches the screen printing hole positions; placing the marked defective glass panel under the laser coding machine and aligning the laser head with the screen printing hole positions; turning on the laser coding machine, and the laser emitted by the laser head codes the screen printing hole positions to complete the removal of the ink dots on the screen printing hole positions; placing the glass panel on the screen printing machine and performing a partition line screen printing on the edges of the screen printing hole positions; and the screen printing machine completes the semi-transparent film screen printing of the screen printing hole positions.

[0004] The prior art fails to improve the efficiency of screen printing quality inspection, which will reduce the inspection efficiency and increase costs. Summary of the Invention

[0005] The present invention aims to provide a screen printing quality inspection system and inspection method for PET substrates to achieve the technical purpose of improving the efficiency of screen printing quality inspection.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a screen printing quality inspection system for PET substrates, including a film material flattening system, a control module, a grasping module, a sensor module, and an image acquisition module. The output end of the control module is connected to the input end of the film material flattening system, the output end of the control module is connected to the input end of the grasping module, the output end of the sensor module is connected to the input end of the control module, and the output end of the image acquisition module is connected to the input end of the control module.

[0008] The control module outputs control instructions to the film material flattening system, the control module outputs control instructions to the grasping module, the sensor module inputs position information to the control module, and the image acquisition module inputs image information to the control module.

[0009] The image acquisition module uses an industrial camera, and the output end of the industrial camera is connected to the input end of the control module.

[0010] The grasping module uses a robotic arm, and the output end of the control module is connected to the input end of the robotic arm.

[0011] The control module uses a PLC controller.

[0012] The sensor module includes a vision sensor, a position sensor, and a pressure sensor. The output ends of the vision sensor, the position sensor, and the pressure sensor are respectively connected to the input end of the control module.

[0013] The present invention provides an inspection method for a PET substrate screen printing quality inspection system, characterized in that:

[0014] Step 1: The control module controls the film flattening system to flatten the film.

[0015] Step 2: The control module controls the grasping module to grasp the PET substrate and place it on the detection platform.

[0016] Step 3: The control module detects the position of the PET substrate through the sensor module.

[0017] Step 4: The control module obtains the image information of the PET substrate through the image acquisition module.

[0018] Step 5: The control module performs image preprocessing on the collected image information.

[0019] Step 6: The control module detects the shape and thickness of the ink pattern on the PET substrate.

[0020] Step 7: The control module calculates the distance between the ink area and the edge of the PET substrate.

[0021] Step 8: The control module performs an appearance inspection on the ink area on the PET substrate.

[0022] Step 9: The control module generates an inspection report.

[0023] In Step 5, the control module performs preprocessing methods such as denoising, enhancing contrast, edge enhancement, and grayscale conversion on the collected image information.

[0024] In Step 6, the control module obtains the shape of the ink pattern on the PET substrate through image segmentation and feature extraction, and the control module obtains the thickness information of the ink through an algorithm trained with data.

[0025] In Step 8, the control module performs defect detection, defect classification, and defect position identification on the ink area.

[0026] The technical effects of the present invention are as follows:

[0027] (1) The present invention improves the detection accuracy and avoids missed detections. Through image recognition technology, it can comprehensively and meticulously inspect each product, not only detecting tiny defects that are difficult to notice with the naked eye but also analyzing and generating detection reports in real time. The high precision of image recognition ensures the consistency and reliability of the detection process, thus effectively reducing the possibility of missed detections.

[0028] (2) The present invention automatically records product data and reduces manual errors. It is equipped with an automatic data acquisition system that can record the product quantity and detection results in real time, reducing the errors in manual recording. Through an intelligent data management system, production data can be automatically statistically analyzed and archived, greatly improving the accuracy of the data and the transparency of the production process.

[0029] (3) The present invention realizes automatic loading and unloading, reducing product losses. By introducing a robotic arm system, automatic loading and unloading operations are achieved, avoiding product damage that may occur during manual operations. The robotic arm can not only precisely control the product handling process but also work in coordination with other production equipment, effectively reducing the operation risks and the probability of product damage during manual handling.

[0030] (4) The present invention improves production efficiency. The automated operations reduce the downtime and ensure the continuous operation of the equipment, greatly improving the production efficiency.

[0031] (5) The present invention has strong flexible adaptability. It is applicable not only to PET substrates of different shapes and sizes but also can efficiently detect surface defects, having strong versatility. Description of the Drawings

[0032] This specification includes the following drawings, and the shown contents are respectively:

[0033] Figure 1 It is a logical structure block diagram of a silk screen quality inspection system and inspection method for a PET substrate according to the present invention;

[0034] Figure 1 The markings in it are: 1. Film material flattening system; 2. Control module; 3. Gripping module; 4. Sensor module; 5. Image acquisition module. Detailed Embodiments

[0035] The following further details the specific embodiments of the present invention by describing the embodiments with reference to the drawings, aiming to help those skilled in the art have a more complete, accurate, and in-depth understanding of the inventive concept and technical solutions of the present invention and to facilitate its implementation.

[0036] The present invention provides a silk screen quality inspection system for PET substrates, including a film flattening system 1, a control module 2, a grasping module 3, a sensor module 4, and an image acquisition module 5. The output end of the control module 2 is connected to the input end of the film flattening system 1, the output end of the control module 2 is connected to the input end of the grasping module 3, the output end of the sensor module 4 is connected to the input end of the control module 2, and the output end of the image acquisition module 5 is connected to the input end of the control module 2.

[0037] The control module 2 outputs control instructions to the film flattening system 1, the control module 2 outputs control instructions to the grasping module 3, the sensor module 4 inputs position information to the control module 2, and the image acquisition module 5 inputs image information to the control module 2.

[0038] The image acquisition module 5 uses an industrial camera, and the output end of the industrial camera is connected to the input end of the control module 2.

[0039] The grasping module 3 uses a robotic arm, and the output end of the control module 2 is connected to the input end of the robotic arm.

[0040] The control module 2 uses a PLC controller.

[0041] The sensor module 4 includes a vision sensor, a position sensor, and a pressure sensor. The output ends of the vision sensor, the position sensor, and the pressure sensor are respectively connected to the input end of the control module 2.

[0042] The present invention provides an inspection method for a silk screen quality inspection system for PET substrates, characterized in that:

[0043] Step 1: The control module 2 controls the film flattening system 1 to flatten the film;

[0044] Step 2: The control module 2 controls the grasping module 3 to grasp the PET substrate and place it on the detection platform;

[0045] Step 3: The control module 2 detects the position of the PET substrate through the sensor module 4;

[0046] Step 4: The control module 2 obtains the image information of the PET substrate through the image acquisition module 5;

[0047] Step 5: The control module 2 performs image preprocessing on the acquired image information;

[0048] Step 6: The control module 2 detects the shape and thickness of the ink pattern on the PET substrate;

[0049] Step 7: The control module 2 calculates the distance between the ink area and the edge of the PET substrate;

[0050] Step 8: The control module 2 performs an apparent inspection on the ink area on the PET substrate.

[0051] Step Nine: The control module 2 generates an inspection report.

[0052] In Step Five, the control module 2 preprocesses the collected image information by denoising, enhancing contrast, edge enhancement, and grayscale conversion.

[0053] In Step Six, the control module 2 obtains the shape of the ink pattern on the PET substrate through image segmentation and feature extraction, and the control module 2 obtains the thickness information of the ink through an algorithm trained with data.

[0054] In Step Eight, the control module 2 performs defect detection, defect classification, and identification of the defect location on the ink area.

[0055] The following details a screen printing quality inspection system and inspection method for the ink on the surface of a PET substrate according to the present invention.

[0056] A screen printing quality inspection system for the ink on the surface of a PET substrate according to the present invention includes a control module 2, a grasping module 3, a sensor module 4, a film flattening system 1, and an image acquisition module 5. Among them, the control module 2 uses a PLC controller for the coordinated control of the overall system, responsible for the motion control of the grasping module 3, the triggering of the image acquisition module 5, the processing of the image data collected by the image acquisition module 5, and generating a detection report. The PLC processor integrates an embedded processing unit, such as an industrial computer or a high-performance GPU, for the real-time processing of image data and deep learning model inference. The real-time operating system (RTOS) in the PLC processor is used for hardware control to ensure the synchronization and efficiency of image acquisition, processing, and grasping actions. The communication interface in the PLC controller is equipped with a high-speed industrial bus (such as EtherCAT or Profinet) to achieve high-speed data transmission between devices, ensuring the stable connection of the sensor module 4, the image acquisition module 5, the grasping module 3, and the control module 2. In the embodiment of the present invention, the PLC controller uses Beckhoff CX8000.

[0057] The grasping module 3 uses a robotic arm. The high-precision servo motor and linear guide in the robotic arm ensure the precise positioning of the robotic arm in the X, Y, and Z directions, ensuring the stable and reliable picking and placing actions of the PET substrate. Each robotic arm has a large load capacity (for example, a maximum of 5 kg) and can handle PET substrates of different sizes and weights. The movement range of the robotic arm is adjustable to adapt to the space layout of production lines of different sizes. The robotic arm is equipped with a suction cup or a vacuum grasping device with an adsorption function to ensure the stability of the picking and placing process on the PET substrate and prevent damage to the PET substrate.

[0058] The sensor module 4 includes a vision sensor, a position sensor, and a pressure sensor. The vision sensor is configured to be located at the end of the robotic arm and is used to detect whether the PET substrate is accurately placed to ensure precise grasping by the robotic arm. The position sensor uses a high-precision position sensor (such as a laser displacement sensor or an encoder) to detect the placement state and posture of the PET substrate to ensure the accuracy of the operation. A pressure sensor is used in the film flattening system 1 to ensure uniform distribution of the vacuum adsorption force, thereby avoiding deformation of the film caused by excessive local pressure.

[0059] The film flattening system 1 includes a vacuum suction module and an intelligent sensing module. The vacuum suction module uses an efficient air suction device to ensure the flatness of the film on the platform and avoid wrinkles and bubbles. During the film flattening process, the intelligent sensing module monitors the flatness of the film in real time. Once unevenness is detected, the adsorption force is immediately adjusted to avoid the impact of the uneven film on subsequent ink detection. It should be noted that the vacuum suction module includes a vacuum adsorption table and an air extraction system. The air extraction system uses a fan, a vacuum pump, or a variable-frequency fan. The intelligent sensing module uses a pressure sensor.

[0060] The image acquisition module 5 uses an industrial camera to ensure image clarity and detail capture ability. The industrial camera is a high-resolution industrial camera (such as a CMOS camera), supports high-frame-rate shooting (60fps and above), and meets the requirements of a fast production line. The automatic focusing and automatic exposure functions of the industrial camera can meet the shooting requirements under different lighting conditions. The resolution of the industrial camera is above 20 million pixels to ensure clear images and sufficient capture of the tiny details in the ink area. The focal length and aperture are adjustable to ensure clear imaging when shooting PET substrates of different sizes. In the present invention, the robotic arm is equipped with at least one industrial camera, which supports shooting at different angles, and combines multi-camera fusion technology to achieve more comprehensive image acquisition.

[0061] The image acquisition module 5 needs to work under the condition of a high-quality light source. The light source of the present invention uses an LED ring light source to ensure uniform illumination and reduce shadows or reflections caused by uneven illumination. The brightness of the LED ring light source is adjustable to ensure that the image acquisition module 5 can acquire high-quality images under different ambient lighting conditions. The LED ring light source of the present invention is a polarized light source, which can reduce the interference of reflected light on image acquisition when detecting high-reflection surfaces (such as PET substrates) and improve the contrast of the ink area.

[0062] Next, a method for inspecting the screen printing quality of the ink on the surface of a PET substrate according to the present invention will be described in detail.

[0063] The control module 2 controls the film flattening system 1 to flatten the film material, ensuring that there are no wrinkles on the surface of the PET substrate and that it does not affect subsequent inspections. Specifically, the vacuum suction module in the film flattening system 1 flattens the film material through vacuum suction technology, ensuring that it has no wrinkles or bubbles, thereby avoiding affecting subsequent ink inspections. During the process of flattening the film material, the intelligent induction module in the film flattening system 1 has intelligent sensors that continuously monitor the flatness of the film material. Once unevenness is detected, the adsorption force is immediately adjusted to prevent the uneven film material from affecting subsequent ink inspections.

[0064] The control module 2 controls the gripping module 3 to grip the PET substrate and place it on the inspection platform.

[0065] The control module 2 detects the position of the PET substrate through the sensor module 4 and proceeds to the next step after determining that the position of the PET substrate is accurate. Specifically, the control module 2 uses the vision sensor and position sensor in the sensor module 4 to inspect the position of the PET substrate. The vision sensor detects whether the PET substrate is placed accurately to ensure precise gripping by the robotic arm. The position sensor uses a high-precision position sensor (such as a laser displacement sensor or an encoder) to detect the placement state and posture of the PET substrate, ensuring the accuracy of the operation.

[0066] The control module 2 obtains the image information of the PET substrate through the image acquisition module 5. Specifically, the control module 2 controls the triggering of the image acquisition module 5 and takes a photo of the PET substrate after its placement position is accurate.

[0067] The control module 2 performs image preprocessing on the acquired image information. Specifically, the control module 2 performs preprocessing methods such as denoising, contrast enhancement, edge enhancement, and grayscale conversion on the acquired image to improve the accuracy of image analysis. Grayscale conversion is to convert the RGB image into a grayscale image, reducing the computational complexity by removing color information, thereby improving the processing speed. Denoising is to use Gaussian filtering, mean filtering, or median filtering to remove noise in the image, ensuring data stability during subsequent processing. Contrast enhancement applies the Contrast Limited Adaptive Histogram Equalization (CLAHE) technique to perform local contrast enhancement on the image, enhancing the distinguishability of the ink area, especially in the case of uneven illumination. Edge enhancement uses the Laplacian transform or Sobel operator to enhance the edges of the ink area and extract more accurate contour information.

[0068] After preprocessing the acquired images, the control module 2 detects the shape and thickness of the ink pattern on the PET substrate. Specifically, the control module 2 performs image segmentation and feature extraction on the images, uses various segmentation algorithms (such as threshold segmentation, edge detection, deep learning models) to extract the ink region, and obtains the geometric features of the ink. The measurement of the thickness, width, and length of the ink is based on image analysis technology, combined with a deep learning model, to automatically measure the thickness, width, and length parameters of the ink region.

[0069] Image segmentation and feature extraction include threshold segmentation, edge detection, deep learning segmentation, and region extraction. Threshold segmentation is performed by thresholding the gray value or color channel of the image to segment the ink region. This method can use a fixed threshold or the Otsu automatic threshold selection algorithm. Edge detection uses the Canny edge detection algorithm to obtain an accurate contour by identifying the edges of the ink region. To avoid the influence of noise, a multi-scale detection method is adopted. Deep learning segmentation uses a convolutional neural network (CNN) to segment the image, combined with a pre-trained model (such as U-Net or Mask R-CNN), to automatically identify the boundary between the ink region and the background region. Region extraction performs connectivity analysis on the extracted ink region in the image, and uses morphological operations (such as dilation and erosion) to optimize the boundary of the ink region and reduce detection errors.

[0070] The applications of deep learning models in feature extraction include: Convolutional Neural Network (CNN) application: Using CNN for image feature extraction to process complex image information. During training, the network automatically extracts effective features from the image (such as the shape, edges, texture, etc. of the ink region), and then applies these features to the prediction of parameters such as ink thickness, width, and length. U-Net architecture: Using the U-Net network for the segmentation of the ink region. U-Net is a classic fully convolutional neural network, especially suitable for the segmentation of fine regions such as medical images, capable of retaining more detailed information and efficiently processing irregular shapes in the image. Mask R-CNN: Used for instance segmentation, capable of accurately segmenting the contour of each ink region and providing a corresponding mask for each region, suitable for multi-region detection tasks.

[0071] The implementation principle of the deep learning model is described in detail below.

[0072] Principle of Convolutional Neural Network (CNN): Through a series of operations including convolution, pooling, and activation functions, CNN gradually extracts features from the original image. Each convolutional kernel in each layer is responsible for extracting different information in the image, such as edges, textures, shapes, etc. By stacking multiple convolutional layers, the network can capture the detailed features of the ink area and perform effective classification or regression tasks. Convolutional layer: Responsible for extracting local features from the image, performing a sliding window operation on the image through the convolutional kernel to generate a feature map. Pooling layer: Reduces the resolution of the feature map, retains the most important features, and reduces the computational amount. Fully connected layer: Synthesizes the features extracted by the convolutional layer and the pooling layer, and finally is used to predict the size, thickness, and other geometric parameters of the ink area.

[0073] U-Net architecture: U-Net is a deep learning architecture specifically for image segmentation, especially suitable for defect detection in medical images and manufacturing. Through its symmetric encoder and decoder structure, the network can maintain the spatial information of the image while extracting features, thus accurately segmenting the ink area.

[0074] Mask R-CNN: Mask R-CNN combines the detection part of Faster R-CNN and the segmentation part of the fully convolutional network, and can generate high-quality masks (i.e., pixel-level segmentation of each object) for each object in the image, suitable for complex multi-object detection and segmentation tasks.

[0075] Specifically for thickness measurement and dimension analysis, first perform thickness prediction. Based on the deep learning model of CNN, learn the non-linear relationship between the gray value in the image and the ink thickness through training data, and use the fully connected layer with a convolutional layer to output the ink thickness. Then perform dimension measurement. Use a network based on the YOLO (You Only Look Once) detection algorithm to quickly and accurately locate the length and width of the ink area and estimate the dimensions. Finally, perform precision optimization. Adopt a regression model to further refine the prediction results of the ink thickness, width, length, and other dimensions to ensure the accuracy of the measurement values.

[0076] The formula for ink thickness measurement is:

[0077] T = f(I(x, y), θ)

[0078] Where f is the deep learning model, I(x, y) is the image gray value, and θ is the training parameter.

[0079] The calculation formulas for ink width and length are:

[0080] Ink width W = max(x) - min(X)

[0081] Ink length L = max(y) - min(Y)

[0082] Among them, (x, y) and (X, Y) are the coordinates of the edge points of the ink area.

[0083] The control module 2 ensures the accuracy of the ink position by calculating the distance between the ink area and the edge of the PET substrate. The calculation of the distance from the ink to the edge of the PET substrate is based on the edge point (x0, y0) of the ink area and the edge point (x p , y p ) to calculate the shortest distance from the ink area to the PET substrate, and the Euclidean distance formula is used for calculation. The calculation formula is:

[0084]

[0085] Among them, (x0, y0) is the edge point of the ink area, and (x p , y p ) is the edge point of the PET substrate.

[0086] In addition to calculating the distance between the ink area and the edge of the PET substrate through the above calculation formula, deep learning can also be used for assistance. A deep learning model (such as an instance segmentation network based on Mask R-CNN) is used to extract the edges of the ink area and the PET substrate, and the relative position relationship between the two is learned through the network.

[0087] The control module 2 conducts an appearance inspection on the ink area on the PET substrate. Specifically, first, defect detection is carried out. Defect detection uses a pre-trained deep learning model (such as ResNet, VGG, etc.) to automatically identify the defects on the surface of the PET substrate. By comparing the defect area and the normal area in the image, the quality problems of the ink area are automatically detected. Then, defect classification is carried out. Based on a deep learning classifier (such as a support vector machine or a convolutional neural network), the surface defects are classified, including scratches, stains, bubbles, etc., and the classification information and position coordinates of the defects are generated. Finally, precise defect location identification is carried out. The convolutional neural network is used to precisely locate the defect area and evaluate whether it will affect the measurement accuracy of the ink area. The appearance inspection uses an image recognition algorithm to detect the defects on the surface of the PET substrate, such as scratches, bubbles, and stains, to ensure that the ink detection is not interfered.

[0088] The present invention improves the detection accuracy and avoids the phenomenon of missed detection. Through image recognition technology, it can comprehensively and meticulously inspect each product. It can not only detect the tiny defects that are difficult to detect by the human eye, but also analyze and generate a detection report in real time. The high precision of image recognition ensures the consistency and reliability of the detection process, thus effectively reducing the possibility of missed detection.

[0089] This invention automatically records product data, reducing manual errors. It is equipped with an automatic data acquisition system that can record the quantity of products and test results in real time, reducing the errors in manual recording. Through an intelligent data management system, production data can be automatically statistically analyzed and archived, greatly improving the accuracy of data and the transparency of the production process.

[0090] This invention realizes automatic loading and unloading, reducing product losses. Through the introduction of a robotic arm system, the operation of automatic loading and unloading is achieved, avoiding product damage that may occur during manual operation. The robotic arm can not only precisely control the product handling process but also work in coordination with other production equipment, effectively reducing the operation risks and the probability of product damage during manual handling.

[0091] This invention improves production efficiency. The automated operation reduces downtime and ensures continuous operation of the equipment, greatly improving production efficiency.

[0092] This invention has strong flexible adaptability. It is applicable not only to PET substrates of different shapes and sizes but also can efficiently detect surface defects, with strong versatility.

[0093] The above has described the present invention by way of example in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above methods. As long as various non-substantive improvements are made by adopting the method concept and technical solution of the present invention; or without improvement, the above concept and technical solution of the present invention are directly applied to other occasions, they are all within the protection scope of the present invention.

Claims

1. A PET substrate screen printing quality inspection system, characterized by: It includes a film material flattening system, a control module, a grabbing module, a sensor module and an image acquisition module. The output end of the control module is connected to the input end of the film material flattening system, the output end of the control module is connected to the input end of the grabbing module, the output end of the sensor module is connected to the input end of the control module, and the output end of the image acquisition module is connected to the input end of the control module.

2. A PET substrate screen printing quality inspection system as claimed in claim 1, characterized in that: The control module outputs control instructions to the film material flattening system, the control module outputs control instructions to the grabbing module, the sensor module inputs position information to the control module, and the image acquisition module inputs image information to the control module.

3. A PET substrate screen printing quality inspection system as claimed in claim 1, characterized in that: The image acquisition module adopts an industrial camera, and the output end of the industrial camera is connected to the input end of the control module.

4. A PET substrate screen printing quality inspection system as claimed in claim 1, characterized in that: The grabbing module adopts a mechanical arm, and the output end of the control module is connected to the input end of the mechanical arm.

5. A PET substrate screen printing quality inspection system as claimed in claim 1, characterized in that: The control module adopts a PLC controller.

6. A PET substrate screen printing quality inspection system as claimed in claim 1, characterized in that: The sensor module comprises a visual sensor, a position sensor and a pressure sensor, and the output end of the visual sensor, the output end of the position sensor and the output end of the pressure sensor are respectively connected to the input end of the control module.

7. An inspection method for a PET substrate screen printing quality inspection system according to any one of claims 1 to 6, characterized in that: Step 1: The control module controls the membrane material flattening system to flatten the membrane material; Step 2: The control module controls the grabbing module to grab the PET substrate and place it on the detection platform; Step 3: The control module detects the position of the PET substrate through the sensor module; Step 4: The control module obtains the image information of the PET substrate through the image acquisition module; Step 5: The control module performs image preprocessing on the collected image information; Step 6: The control module detects the shape and thickness of the ink pattern on the PET substrate; Step 7: The control module calculates the distance between the ink area and the edge of the PET substrate; Step 8: The control module performs a surface inspection on the ink area on the PET substrate; Step 9: The control module generates an inspection report.

8. The inspection method of a PET substrate screen printing quality inspection system as claimed in claim 7, characterized in that: In step five, the control module pre-processes the collected image information by denoising, contrast enhancement, edge enhancement and grayscale conversion.

9. The inspection method of a PET substrate screen printing quality inspection system as claimed in claim 7, characterized in that: In step six, the control module obtains the shape of the ink pattern on the PET substrate through image segmentation and feature extraction, and obtains the thickness information of the ink through an algorithm trained with data.

10. The inspection method of a PET substrate screen printing quality inspection system as claimed in claim 7, characterized in that: In step eight, the control module performs defect detection, defect classification and identification of defect locations on the ink area.

Images