A transparent mark defect detection method and device, electronic equipment and storage medium

Abstract

The application relates to a transparent identification defect detection method and device, electronic equipment and a storage medium, wherein the method adopts a super-resolution system to process video data; when the super-resolution system processes the video data, a real-time processing frame obtained by processing the video data by the super-resolution system is acquired; a transparent identification arranged on each product in the real-time processing frame is acquired, and a characteristic parameter of the transparent identification is determined, the characteristic parameter being used to represent a quality level of the transparent identification; a defect detection result of the transparent identification is generated according to a comparison result of the characteristic parameter and a standard parameter; after the video data is processed by the super-resolution system, the transparent identification in the high-resolution real-time processing frame is automatically recognized, so that the transparent identification on the product is automatically recognized, defects of the transparent identification are checked, the defect checking efficiency of the transparent identification is improved, and the defect detection accuracy of the transparent identification is improved compared with artificial detection.

Description

Technical Field

[0001] This application relates to the field of computer vision, and in particular to a method, apparatus, electronic device and storage medium for detecting defects in transparent markings. Background Technology

[0002] During product manufacturing, labels are typically added to products to provide information and seal the packaging. These labels come in various forms, with transparent labels being a common choice for sealing and explaining products. Therefore, ensuring the quality of these transparent labels during production is crucial. Common defects include misalignment, air bubbles, impurities, wrinkles, and lifting. Detecting and identifying these defects is essential for improving production efficiency and ensuring product quality.

[0003] Traditionally, defects in transparent labels are detected manually. However, due to their transparency, these labels are difficult to extract, making it challenging to inspect for defects and ensure product quality. Inspectors require considerable time to detect any defects. Furthermore, due to customization, the materials, sizes, placement, and content of transparent labels vary, meaning these defects can occur in any part of the label. Visual fatigue and human subjectivity can lead to missed or undetected defects, resulting in low efficiency in transparent label defect detection. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for detecting defects in transparent signs, in order to solve the problem that manual inspection of defects in transparent signs can lead to missed detections or failure to detect defects in a timely manner, resulting in low efficiency in detecting defects in transparent signs.

[0005] In a first aspect, this application provides a method for detecting defects in transparent labels. The method includes: processing video data using a super-resolution system, wherein the video data is collected from the area where the transparent label is located on a product; when the super-resolution system processes the video data, acquiring a real-time processing frame obtained by the super-resolution system processing the video data; acquiring the transparent labels set on each product in the real-time processing frame and determining the feature parameters of the transparent labels, the feature parameters being used to characterize the quality level of the transparent labels; and generating a defect detection result for the transparent labels based on the comparison result between the feature parameters and standard parameters.

[0006] Secondly, this application provides a transparent label defect detection device, comprising: an acquisition module for processing video data using a super-resolution system, wherein the video data is acquired by collecting data from the area where the transparent label is located on a product; acquiring real-time processing frames obtained by the super-resolution system when processing the video data; a determination module for acquiring the transparent labels set on each product in the real-time processing frames and determining the feature parameters of the transparent labels, wherein the feature parameters are used to characterize the quality level of the transparent labels; and a generation module for generating a defect detection result of the transparent labels based on the comparison result of the feature parameters and standard parameters.

[0007] Thirdly, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0008] Memory, used to store computer programs;

[0009] When the processor executes a program stored in the memory, it implements the steps of the transparent marking defect detection method according to any embodiment of the first aspect.

[0010] Fourthly, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the transparent marking defect detection method as described in any embodiment of the first aspect.

[0011] The technical solutions provided in this application have the following advantages compared with the prior art:

[0012] The method provided in this application embodiment acquires video data and processes the video data using a super-resolution system. When the super-resolution system processes the video data, it acquires real-time processing frames obtained from the processing. It acquires transparent labels set on each product within the real-time processing frames and determines the characteristic parameters of the transparent labels, which characterize the quality level of the transparent labels. Based on the comparison results between the characteristic parameters and standard parameters, it generates defect detection results for the transparent labels. After processing the video data using the super-resolution system, the transparent labels within the high-resolution real-time processing frames are automatically identified, thus automatically identifying transparent labels on products and checking for defects. This avoids the problems of missed detections or failure to detect defects in a timely manner due to manual inspection of transparent labels, resulting in low efficiency in defect detection. Therefore, it improves the efficiency of transparent label defect inspection and, compared with manual inspection, increases the accuracy of transparent label defect detection. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a basic flowchart illustrating a method for detecting defects in transparent markings provided in an embodiment of this application.

[0016] Figure 2 A schematic diagram of the basic process of an optional transparent label defect detection method provided in an embodiment of this application;

[0017] Figure 3 This is a schematic diagram of the basic structure of a transparent label defect detection device provided in an embodiment of this application;

[0018] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Figure 1 This is a flowchart illustrating a transparent label defect detection method provided in an embodiment of this application. The method includes, but is not limited to:

[0021] S101. A super-resolution system is used to process the video data, which is obtained by collecting data from the area where the transparent markings on the product are located.

[0022] It is important to understand that the video data mentioned above includes products, and the products contain logos, including transparent logos. The video data can be obtained from a server by capturing video data of the area where the transparent logo is located on the product, or it can be obtained by capturing video data of the area where the transparent logo is located on the product in real time through a shooting device, including but not limited to: cameras, mobile phones, tablets and other devices that can record video.

[0023] It's important to understand that Video Super-Resolution (VSR) evolved from Image Super-Resolution and is a hot topic in computer vision. VSR technology can reconstruct videos, restore video clarity, and improve subjective visual quality. Generative Adversarial Network (GAN)-based Super-Resolution Systems (EGVSR) can meet the needs of large-scale videos. EGVSR employs various optimization techniques to minimize computation while ensuring improved visual quality, thus meeting the requirements of large-scale videos. Therefore, this embodiment can use EGVSR to process video data and obtain super-resolution images to meet the subsequent needs of transparent label recognition.

[0024] S102. When the super-resolution system processes the video data, the real-time processing frame obtained by the super-resolution system processing the video data is acquired.

[0025] It is important to understand that when a super-resolution system processes video data, it processes each frame of the video data sequentially. Processing one frame of video data results in the output of a high-resolution video image, which is then used as the real-time processing frame.

[0026] S103. Obtain the transparent labels set on each product in the real-time processing frame, and determine the feature parameters of the transparent labels, wherein the feature parameters are used to characterize the quality level of the transparent labels;

[0027] It is important to understand that after processing video data through a super-resolution system, the super-resolution system can also be used to identify the images in real-time processing frames to determine the products in the real-time processing frames. After identifying the products in the real-time processing frames, the transparent labels on the products are obtained, and their corresponding feature parameters are determined based on the obtained transparent labels. The feature parameters are used to characterize the quality level of the transparent labels. Specifically, transparent labels may have defects such as misalignment, bubbles, and impurities, and the feature parameters can be used to reflect whether the transparent labels have these defects.

[0028] S104. Generate the defect detection result of the transparent mark based on the comparison result of the characteristic parameters and the standard parameters.

[0029] It is important to understand that the standard parameter refers to the pre-set quality level of the transparent marker. The feature parameters that characterize the quality level of the transparent marker in the real-time processing frame are compared with the standard parameter, and the defect detection results of the transparent marker can be generated based on the comparison results.

[0030] The transparent label defect detection method provided in this embodiment acquires video data and processes the video data using a super-resolution system. While the super-resolution system processes the video data, it acquires real-time processing frames obtained from the processing. It then acquires the transparent labels set on each product within the real-time processing frames and determines the characteristic parameters of each transparent label, which characterize the quality level of the transparent label. Based on the comparison results between the characteristic parameters and standard parameters, it generates a defect detection result for the transparent label. After processing the video data using the super-resolution system, the transparent labels within the high-resolution real-time processing frames are automatically identified, thus automatically identifying transparent labels on products and checking for defects. This avoids the problems of missed detections or failure to detect defects in a timely manner due to manual inspection, resulting in low efficiency in transparent label defect detection. Therefore, it improves the efficiency of transparent label defect inspection and simultaneously increases the accuracy of transparent label defect detection.

[0031] In some examples of this embodiment, before processing the video data using a super-resolution system, the method further includes: preprocessing a first transparent label training sample set, wherein the preprocessing is used to perform data augmentation on the first transparent label training sample set to obtain a second transparent label training sample set; training a super-resolution training model using the second transparent label training sample set; and using the trained super-resolution training model as the super-resolution system. The transparent label training sample set is a video dataset containing products with transparent labels affixed to them.

[0032] Continuing from the previous example, to ensure the diversity of the transparent label training sample set used to train the super-resolution training model and to better train the super-resolution training model, the first transparent label training sample set can be preprocessed to enhance it. Specifically, the transparent labels in the video dataset are first labeled to obtain labeled images. The labels include, but are not limited to, information such as the model, category, size, color, and key point coordinates of the transparent labels. Data augmentation processing is then performed on the labeled images to obtain enhanced images. Data augmentation processing includes: flipping, rotating, grayscale conversion, cropping, random translation, and stitching. Flipping can be done horizontally or vertically on the labeled image; rotation can be done by randomly rotating the labeled image at a certain angle, such as 90° or 180°; grayscale conversion can be done by randomly converting the labeled image to grayscale values; cropping involves randomly sampling a portion from the labeled image and then adjusting the size of this portion to match the size of the labeled image; the augmented image and the labeled image are mixed to obtain a second transparent label training sample set; data augmentation can increase the diversity of the transparent label training sample set, thereby improving the accuracy and robustness of subsequent super-resolution training model training.

[0033] Following the previous example, after preprocessing the transparent marker training sample set, the super-resolution training model is trained using the preprocessed transparent marker training sample set. The purpose of training the super-resolution training model is to enable the trained super-resolution system to obtain transparent markers in the video data when processing the video data in the future. Therefore, after training, the super-resolution training model can serve as the super-resolution system.

[0034] It's important to understand that the video data is processed using a super-resolution system, EGVSR, which comprises three parts: a recurrent generator, a stream estimation network, and a spatiotemporal discriminator. The recurrent generator primarily generates high-resolution video frames for each frame of the video data. The spatiotemporal discriminator evaluates the generation quality through loss to improve the high-resolution output. The video frame image is treated as a whole and input into the stream estimation network. Instead of dividing individual frames, the network calculates the overall pixel changes to compute motion compensation and learn the motion compensation between frames. Motion compensation reduces redundancy in the video frame sequence by predicting and compensating for the current local image from previous local images. By obtaining high-resolution, detailed video frames with clearer local details and more natural temporal changes, a high-resolution video is ultimately achieved.

[0035] In some examples of this embodiment, the characteristic parameters of the transparent label include, but are not limited to: area coordinates, area defect size, and the size of the area. The area coordinates of the transparent label can indicate whether the transparent label is misaligned; the area defect size can indicate whether the transparent label has bubbles or impurities; and the size of the area can indicate whether the transparent label has wrinkles or curling.

[0036] Continuing with the previous example, the characteristic parameters of the transparent sign include at least one of the following: extracting the transparent sign boundary texture to determine the region coordinates of the transparent sign; performing preset processing on the image of the transparent sign to obtain a defect region and determining the size of the defect region of the transparent sign; obtaining the region where the transparent sign is located and determining the size of the region where the transparent sign is located.

[0037] The process of extracting the transparent label's boundary texture to determine the region coordinates of the transparent label includes, but is not limited to: extracting the text pattern feature parameters on the transparent label from the video data and the transparent label's boundary texture to calculate the region coordinates where the transparent label is pasted on the product. The region coordinates can characterize the coordinates and angles of the transparent label on the product.

[0038] The process of performing pre-processing on the image of the transparent label to obtain a defect area, and determining the size of the defect area of ​​the transparent label includes, but is not limited to: extracting the transparent label from the video data, performing morphological calculations to enhance the image and binarize the extracted transparent label to obtain a black area in the mask image, using the black area as the defect area, and obtaining the size of the defect area by obtaining the size of the black area.

[0039] The process of obtaining the area where the transparent marker is located and determining the size of the area where the transparent marker is located includes, but is not limited to: extracting the transparent marker from the video data, obtaining the binary mask of the transparent marker, calculating the area fitting value and convex hull key points of the edge contour of the transparent marker, and then obtaining the size of the area where the transparent marker is located.

[0040] In some examples of this embodiment, generating a defect detection result for the transparent label based on the comparison result between the feature parameters and the standard parameters includes: comparing the feature parameters with the standard parameters to generate a comparison result; when the comparison result indicates that the feature parameters do not match the standard parameters, it is determined that the transparent label has a defect.

[0041] When the parameters of the obtained transparent label include region coordinates, the region coordinates of the transparent label can be compared with the region coordinates in the standard parameters to determine whether the transparent label has a misalignment defect. Specifically, the characteristics of a misaligned transparent label are reflected in the following: under the binary display of the defective image, the angle of the texture area at the boundary of the transparent label image is rotated (0°-360°), or the distance between the coordinate points corresponding to the transparent label coordinates exceeds the specified range (e.g., limited to 15). That is, if any parameter of the angle or coordinate in the region coordinates of the transparent label exceeds the range of the angle and coordinate parameters of the standard parameters, or if any parameter of the angle or coordinate does not conform to the standard parameters, it is determined that the transparent label has a misalignment defect. When neither the angle nor the coordinate in the region coordinates of the transparent label exceeds the range of the angle and coordinate parameters of the standard parameters, that is, if both the angle and the coordinate conform to the standard parameters, it is determined that the transparent label does not have a misalignment defect.

[0042] Specifically, when the parameters of the obtained transparent label include the size of the defect area, the size of the defect area of ​​the transparent label can be compared with the size of the defect area in the standard parameters to determine whether the transparent label has defects such as bubbles or impurities. Specifically, the presence of bubbles or impurities in the transparent label is determined by comparing the size of the defect area of ​​the transparent label with the size of the defect area in the standard parameters. When the size of the defect area of ​​the transparent label exceeds the size of the defect area in the standard parameters, that is, when the feature parameters do not match the standard parameters, the transparent label is determined to have bubbles or impurities. For example, if the size of the defect area in the standard parameters is set to 0.5mm, and the transparent label has a black area whose size exceeds 0.5mm in the standard parameters, then the size of the black defect area of ​​the transparent label does not match the size of the defect area in the standard parameters, and therefore the transparent label is determined to have bubbles or impurities. When the size of the black area of ​​the transparent label does not exceed 0.5mm, or the transparent label has no black area, then the transparent label is determined not to have bubbles or impurities.

[0043] Specifically, when the acquired feature parameters of the transparent label include the size of its region, the size of the transparent label can be compared with the size of a normal transparent label in the standard parameters to determine whether the transparent label has defects such as wrinkles or warping. Specifically, by acquiring the region where the transparent label is located, extracting the binary mask of the transparent label, calculating the area fitting value of the edge contour and convex hull key points, etc., the size of the region where the transparent label is located is obtained. The size of the region where the transparent label is located is then used to determine whether the transparent label has defects such as wrinkles or warping. For example, when the size of the region where the transparent label is located is L1, and the size of the region where a normal transparent label is located in the standard parameters is L, if the difference between the size of L1 and the size of L exceeds a threshold (a preset threshold, such as a threshold of 5), it is determined that the size of the transparent label does not match the size of the transparent label in the standard parameters, and therefore, it is determined that the transparent label has defects such as wrinkles or warping.

[0044] In some examples of this embodiment, after generating the defect detection result of the transparent marker based on the feature parameters, the method further includes: after the high-resolution video data is obtained by processing the video data by the super-resolution system, tracking the identified transparent markers through the super-resolution system, and labeling the defective transparent markers when outputting the high-resolution video data. It should be understood that the depth information, size information, and template coordinates of the transparent markers contained in the video data are used as input to train the network, thereby detecting and tracking multiple transparent marker targets within the detection range, dynamically tracking their changes as their states move, and labeling the defective parts of the transparent markers during output. It is necessary to count the passes through the detection position and perform real-time labeling, tracking, and statistics on the targets appearing in the monitoring video.

[0045] In some examples of this embodiment, after generating the defect detection results of the transparent markers based on the feature parameters, the method further includes: when a defect is detected in any of the transparent markers in the video data, issuing a defect alert to remind relevant personnel that the transparent markers are defective. It should be understood that in some examples, after generating the detection results of the transparent markers, the detection results can also be saved locally, including the type of transparent marker appearing, the display image (the detection results are displayed based on the coordinate system and the predicted coordinates of the target key points), the defect type (offset, bubbles, impurities, etc.), the marked defect location, the detected data, the predicted coordinates, and other results, for subsequent traceability and querying. The three-dimensional coordinates of the transparent markers are calculated and transformed by combining the acquired world coordinates and camera coordinates. Combining the detected defect area, a fixed real-world location, such as the origin, is selected to establish an xyz three-dimensional spatial coordinate system, facilitating the display of the detection results in three-dimensional coordinates.

[0046] In some examples of this embodiment, after generating the defect detection result of the transparent label based on the feature parameters, the method further includes: automatically removing the product corresponding to the transparent label when the transparent label has a defect. Specifically, after alerting and alarming for the detected defective transparent label, the method tracks the movement trajectory of the defective transparent label using collected video time-series information, predicts its movement trend based on the movement state pattern, and automatically removes the corresponding product if the product with the defective transparent label is not manually removed in a timely manner. Removal is primarily based on defective parts requiring rework due to defects, supplemented by manual removal.

[0047] To better understand the present invention, this embodiment provides a more specific example to illustrate the invention. This embodiment also provides a method for detecting defects in transparent markings, such as... Figure 2 As shown, it includes, but is not limited to:

[0048] S201. Train the super-resolution training model using a transparent labeled training sample set to obtain the super-resolution system;

[0049] In some examples of this embodiment, the transparent label training sample set is first preprocessed. The transparent label training sample set includes video data samples containing transparent labels. The transparent label model, category, size, color, key point coordinates and other information of the transparent label are marked in the video data samples. The video frame data is then preprocessed by random translation, flipping, cropping and splicing. The preprocessed transparent label training sample set is then input into the super-resolution training model for training to obtain the super-resolution system.

[0050] S202. Use a super-resolution system to process video data;

[0051] It's important to understand that the video data is collected from the area containing the transparent markings on the product. This video data can be obtained from a server or directly from the shooting device. The super-resolution system comprises three parts: a recurrent generator, a stream estimation network, and a spatiotemporal discriminator. The recurrent generator primarily generates high-resolution video frames, while the spatiotemporal discriminator assesses the generated quality through loss analysis to improve the high-resolution effect. The video frame image is treated as a whole and input into the stream estimation network. Instead of dividing it into individual frames, the network calculates the overall pixel changes to compute motion compensation and learn the motion compensation between frames. Motion compensation reduces redundant information in the video frame sequence by predicting and compensating for the current local image from previous local images. By obtaining high-resolution, detailed video frames with clearer local areas and more natural temporal changes, a high-resolution video is ultimately achieved.

[0052] S203. Obtain the real-time processing frame obtained by the super-resolution system processing video data, and perform defect detection on the transparent markers in the real-time processing frame;

[0053] In some examples of this embodiment, defect detection is performed on the acquired real-time processing frames. It's important to understand that the super-resolution system processes video data frame by frame, processing each frame sequentially. Each processed frame outputs a high-resolution video image, which is then used as the real-time processing frame. After identifying the transparent marker target in the real-time processing frame, defects such as misalignment, bubbles, and impurities are detected. Bubble and impurity areas appear as spots or stripes. Similarity calculations are used to label the corresponding defect categories. Color features of the transparent marker are extracted to detect bubble and impurity defects, and black area extraction is used to detect and determine defects in the missing parts. The main purpose is to confirm whether there are defects and their categories.

[0054] The detection of defects such as misaligned transparent labels involves extracting the region coordinates of the transparent labels. This is achieved by using text and pattern feature representations on the transparent labels in video data and obtaining boundary details of the transparent labels through super-resolution analysis. The coordinates of the region where the transparent label is located are calculated, representing the coordinates and angles of the transparent label on the product. The characteristics of misaligned transparent labels are that, under binary display of the defective image, the boundary region of the transparent label image appears rotated (0°-360°), or the distance between corresponding feature points exceeds a specified range (e.g., limited to 15). Acceptable labels have angles and coordinates within the normal threshold range. The judgment is mainly based on the calculation of corresponding feature point parameters, marking and displaying abnormal areas, and the target size feature parameters can also serve as an important basis for judging the extent of the defect.

[0055] The process includes detecting bubbles and impurities in transparent labels. Transparent labels are extracted from video data. Morphological calculations, image enhancement, and image binarization are performed on the extracted labels to obtain a black area in the mask image. This black area is considered the defect area, and its size is determined by measuring its size. Defects in transparent labels are characterized by black circular or elliptical dots larger than 0.5mm. Acceptable defects are those with uniformly distributed black or elliptical dots, or dots smaller than 0.5mm. Specifically, the size of the defect area in the transparent label image is obtained by subtracting the size of the black area in the standard image. The judgment is primarily based on the difference in area and size of the black portion in the resulting image calculated using image differencing. The coordinates of the bubble and impurity defect areas are then marked on the test image.

[0056] Among these, the detection of wrinkles and curling of transparent markings is performed. For transparent marking targets, the binary mask of the transparent marking is extracted, and the area fitting value of the edge contour and the convex hull key points are calculated to obtain the size of the area where the transparent marking is located. For the same position area of ​​normal and defective parts, the size L of the normal part and the size L1 of the defective part at the same position are calculated. If the difference between the two size values ​​is within an allowable range (e.g., the difference is less than the threshold of 5), it is considered qualified.

[0057] S204. Track and record any defective transparent markings;

[0058] In some examples of this embodiment, the acquired video data, along with information including depth, size of the transparent markers, and template coordinates of the transparent markers, is used as input to the super-resolution system to train the network. This allows the system to detect and track multiple transparent marker targets within the detection range, dynamically tracking their movement. The output includes annotations of defects in the transparent markers. It is necessary to count the number of targets passing through the detection location and to perform real-time annotation, tracking, and statistics on targets appearing in the monitoring video.

[0059] S205. When any transparent label has a defect, a defect warning shall be issued;

[0060] When a defect is detected in any of the transparent markers in the video data, an error alert is issued to notify relevant personnel that the transparent marker is defective. It should be understood that in some examples, after generating the detection results for the transparent markers, the results can be saved locally, including the type of transparent marker, the display image (the detection results are displayed based on the coordinate system and the predicted coordinates of the target key points), the defect type (offset, bubbles, impurities, etc.), the marked defect location, the detected data, and the predicted coordinates. These results are saved locally for subsequent traceability and querying. The three-dimensional coordinates of the transparent marker are calculated and transformed from the acquired world coordinates and camera coordinates. Combining the detected defect area, a fixed real-world location is selected as the origin to establish an xyz three-dimensional spatial coordinate system, facilitating the display of the detection results in three dimensions.

[0061] S206. Remove any defective transparent labels.

[0062] In some examples of this embodiment, after generating the defect detection result of the transparent label based on the feature parameters, the method further includes: automatically removing the product corresponding to the transparent label when the transparent label has a defect. Specifically, after alerting and alarming for the detected defective transparent label, the method tracks the movement trajectory of the defective transparent label using collected video time-series information, predicts its movement trend based on the movement state pattern, and automatically removes the corresponding product if the product with the defective transparent label is not manually removed in a timely manner. Removal is primarily based on defective parts requiring rework due to defects, supplemented by manual removal.

[0063] like Figure 3 As shown in the figure, this application embodiment also provides a transparent label defect detection device, the transparent label defect detection device comprising:

[0064] Acquisition module 1 is used to process video data using a super-resolution system, wherein the video data is acquired by collecting data from the area where the transparent markings on the product are located; when the super-resolution system processes the video data, it acquires the real-time processing frames obtained by the super-resolution system in processing the video data.

[0065] The determining module 2 is used to obtain the transparent labels set on each product in the real-time processing frame and determine the feature parameters of the transparent labels, wherein the feature parameters are used to characterize the quality level of the transparent labels;

[0066] The generation module 3 is used to generate the defect detection result of the transparent mark based on the comparison result of the feature parameters and the standard parameters.

[0067] The transparent label defect detection device provided in this embodiment acquires video data and processes the video data using a super-resolution system. When the super-resolution system processes the video data, it acquires real-time processing frames obtained from the processing. It then acquires the transparent labels set on each product within the real-time processing frames and determines the characteristic parameters of each transparent label, which characterize the quality level of the transparent label. Based on the comparison results between the characteristic parameters and standard parameters, it generates a defect detection result for the transparent label. After processing the video data using the super-resolution system, the device automatically identifies the transparent labels within the high-resolution real-time processing frames, thus automatically identifying the transparent labels on the products and checking for defects. This avoids the problems of missed detections or failure to detect defects in a timely manner due to manual inspection, resulting in low efficiency in transparent label defect detection. Therefore, it improves the efficiency of transparent label defect inspection and increases the accuracy of transparent label defect detection.

[0068] like Figure 4 As shown in the figure, this application provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114.

[0069] Memory 113 is used to store computer programs;

[0070] In one embodiment of this application, when the processor 111 executes the program stored in the memory 113, it implements the transparent identification defect detection method provided in any of the foregoing method embodiments, including:

[0071] A super-resolution system is used to process the video data, wherein the video data is obtained by collecting data from the area where the transparent markings on the product are located;

[0072] When the super-resolution system processes the video data, the real-time processing frame obtained by the super-resolution system processing the video data is acquired.

[0073] Obtain the transparent labels set on each product in the real-time processing frame, and determine the feature parameters of the transparent labels, which are used to characterize the quality level of the transparent labels;

[0074] The defect detection result of the transparent label is generated based on the comparison result between the characteristic parameters and the standard parameters.

[0075] This application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the transparent marking defect detection method provided in any of the foregoing method embodiments.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0077] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for detecting defects in transparent markings, characterized in that, The method for detecting defects in transparent markings includes: A super-resolution system is used to process the video data, wherein the video data is obtained by collecting data from the area where the transparent markings on the product are located; When the super-resolution system processes the video data, the real-time processing frame obtained by the super-resolution system processing the video data is acquired. Obtain the transparent labels set on each product in the real-time processing frame, and determine the feature parameters of the transparent labels, which are used to characterize the quality level of the transparent labels; The defect detection result of the transparent label is generated based on the comparison result between the characteristic parameters and the standard parameters; The characteristic parameters of the transparent identifier include at least one of the following: Extract the transparent label's boundary texture to determine the transparent label's region coordinates. These region coordinates are used to indicate whether the transparent label has any misalignment defects. The image of the transparent label is processed to obtain the defect area, and the size of the defect area of ​​the transparent label is determined. The size of the defect area is used to reflect whether there are bubbles or impurities in the transparent label. The area where the transparent label is located is obtained, and the size of the area where the transparent label is located is determined. The size of the area is used to indicate whether the transparent label has wrinkles or curling. Before processing the video data using a super-resolution system, the method further includes: The first transparent marker training sample set is preprocessed to perform data augmentation on the first transparent marker training sample set to obtain a second transparent marker training sample set. The preprocessing includes: labeling the transparent markers in the video dataset to obtain labeled images, wherein the labels include the model, category, size, color, and keypoint coordinates of the transparent markers; performing data augmentation on the labeled images to obtain augmented images, wherein the data augmentation includes: flipping, rotating, grayscale conversion, cropping, random translation, and stitching; and mixing the augmented images and the labeled images to obtain the second transparent marker training sample set. The super-resolution training model is trained using the second transparent identifier training sample set; The trained super-resolution model is used as the super-resolution system.

2. The method according to claim 1, characterized in that, The defect detection results for the transparent label are generated based on the comparison results between the characteristic parameters and the standard parameters, including: The feature parameters are compared with the standard parameters to generate a comparison result; When the comparison result shows that the feature parameter does not match the standard parameter, the transparent identifier is determined to have a defect.

3. The method according to claim 2, characterized in that, After generating the defect detection result of the transparent identifier based on the feature parameters, the method further includes: After the super-resolution system processes the video data to obtain high-resolution video data, the super-resolution system tracks the identified transparent markers and marks any defective transparent markers when outputting the high-resolution video data.

4. The method according to claim 2, characterized in that, After generating the defect detection result of the transparent identifier based on the feature parameters, the method further includes: When a defect is detected in any of the transparent identifiers in the video data, a defect alert is issued.

5. The method according to claim 2, characterized in that, After generating the defect detection result of the transparent identifier based on the feature parameters, the method further includes: When the transparent label has a defect, the product corresponding to the transparent label is automatically removed.

6. A transparent label defect detection device, characterized in that, The transparent label defect detection device includes: The acquisition module is used to process video data using a super-resolution system, wherein the video data is acquired by capturing the area where the transparent markings on the product are located; when the super-resolution system processes the video data, the module acquires the real-time processing frames obtained by the super-resolution system processing the video data. The determination module is used to obtain the transparent labels set on each product in the real-time processing frame and determine the feature parameters of the transparent labels, wherein the feature parameters are used to characterize the quality level of the transparent labels; The generation module is used to generate the defect detection result of the transparent label based on the comparison result of the feature parameters and the standard parameters; The determining module is used to achieve at least one of the following: Extract the transparent label's boundary texture to determine the transparent label's region coordinates. These region coordinates are used to indicate whether the transparent label has any misalignment defects. The image of the transparent label is processed to obtain the defect area, and the size of the defect area of ​​the transparent label is determined. The size of the defect area is used to reflect whether there are bubbles or impurities in the transparent label. The area where the transparent label is located is obtained, and the size of the area where the transparent label is located is determined. The size of the area is used to indicate whether the transparent label has wrinkles or curling. Before processing the video data using a super-resolution system, the process further includes: The first transparent marker training sample set is preprocessed to perform data augmentation on the first transparent marker training sample set to obtain a second transparent marker training sample set. The preprocessing includes: labeling the transparent markers in the video dataset to obtain labeled images, wherein the labels include the model, category, size, color, and keypoint coordinates of the transparent markers; performing data augmentation on the labeled images to obtain augmented images, wherein the data augmentation includes: flipping, rotating, grayscale conversion, cropping, random translation, and stitching; and mixing the augmented images and the labeled images to obtain the second transparent marker training sample set. The super-resolution training model is trained using the second transparent identifier training sample set; The trained super-resolution model is used as the super-resolution system.

7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in a memory, it implements the steps of the transparent marking defect detection method according to any one of claims 1-5.

8. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the transparent label defect detection method as described in any one of claims 1-5.

Images