A method, system, and apparatus for detecting bubbles in a cross-section of a tire semi-product

Abstract

The application discloses a kind of tire semi-product section bubble detection method, system and equipment, method includes: by moving visual detection device by motion device, and by visual detection device acquisition shooting image;Photograph image is carried out image splicing, and tire section image is obtained;Tire section image is input to preset tire semi-product section bubble detection model, and by attention mechanism is combined with multi-scale feature fusion module feature extraction is carried out to tire section image, determines bubble feature, and bubble region containing bubble feature is marked, and obtains annotation file;The area of each bubble region in annotation file is calculated based on bubble feature;The sum of the area of all bubble regions and the ratio of tire semi-product section area are calculated, and the bubble rate of tire semi-product section is determined.The application can more effectively and more quickly detect the bubble rate of tire semi-product section, improve the work efficiency of tire semi-product detection.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more specifically, to a method, system, and device for detecting air bubbles in the cross-section of a tire semi-finished product. Background Technology

[0002] In the tire manufacturing process, the quality of semi-finished products directly affects the pass rate of the final finished tires. Among them, the air bubble rate of semi-finished products is an important standard for measuring their quality. Traditional methods for detecting air bubble rate require a long testing time (more than two hours), resulting in low work efficiency.

[0003] Therefore, the existing technology has defects and urgently needs improvement. Summary of the Invention

[0004] In view of the above problems, the purpose of this invention is to provide a method, system and device for detecting air bubbles in the cross-section of tire semi-finished products, which can detect the air bubble rate of the cross-section of tire semi-finished products more effectively and quickly, thereby improving the work efficiency of tire semi-finished product inspection.

[0005] The first aspect of this invention provides a method for detecting air bubbles in the cross-section of a tire semi-finished product, comprising:

[0006] The visual detection device is moved by a motion device, and the visual detection device acquires captured images.

[0007] The captured images are stitched together to obtain a tire cross-section image;

[0008] The tire cross-section image is input into a preset tire semi-finished product cross-section bubble detection model. The tire cross-section image is used to extract features through an attention mechanism combined with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are marked to obtain a labeling file.

[0009] Calculate the area of ​​each bubble region in the annotation file based on bubble features;

[0010] Calculate the ratio of the sum of the areas of all bubble regions to the cross-sectional area of ​​the tire semi-finished product to determine the bubble rate of the tire semi-finished product cross-section.

[0011] This plan also includes:

[0012] Determine whether the stitched image is complete;

[0013] If not, the visual detection device is moved by the motion device to acquire the next captured image, and the image stitching of the next captured image continues;

[0014] If so, the stitched image is determined to be a tire cross-section image.

[0015] In this solution, the step of stitching the captured images includes:

[0016] The captured image is preprocessed; the preprocessing includes noise reduction and illumination normalization.

[0017] Key points and descriptors of the captured image are extracted using a multi-feature fusion method to determine the image features of the captured image; the image features of the captured image include SURF features, edge features, and color features.

[0018] A feature matching algorithm based on geometric constraints is used to perform feature matching on the captured image.

[0019] Based on the analysis of the matched image features, the captured images are aligned using a global optimization method, and multi-band fusion technology is used to stitch the images together to generate the stitched result.

[0020] The stitching results are adjusted for color consistency to obtain a tire cross-section image.

[0021] This plan also includes:

[0022] Auxiliary feature matching is performed using the RANSAC algorithm.

[0023] In this scheme, the feature extraction of the tire cross-section image is performed by combining an attention mechanism with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are then marked to obtain a labeled file, including:

[0024] Feature extraction is performed on the tire cross-section image by combining an attention mechanism with a multi-scale feature fusion module to identify bubble features within the tire cross-section image;

[0025] The bubble feature is selected by using a bounding box;

[0026] The bounding box is filtered using an improved nonmaximum suppression algorithm to identify bubble regions containing bubble features and generate a labeling file.

[0027] In this solution, calculating the area of ​​each bubble region in the annotation file based on bubble features includes:

[0028] The number of bubbles is determined by counting the number of bubble targets in the annotation file.

[0029] Obtain the resolution of the tire cross-section image;

[0030] The coordinates of the bubble region are converted into actual pixel coordinates based on the resolution of the tire cross-section image;

[0031] Read the segmentation mask of the bubble region;

[0032] The number of pixels in the segmentation mask whose pixel values ​​are greater than a preset pixel value threshold is counted to determine the pixel area of ​​the bubble region.

[0033] This plan also includes:

[0034] Overlay the segmentation masks of all bubble regions onto the same image;

[0035] Mark the overlapping areas obtained by mask overlay;

[0036] Count the number of independent pixels in bubble region i, and determine the area A of bubble region i. i ;

[0037]

[0038] Where M is the segmentation mask, H and W represent the height and width of the image, respectively. i,j,k The pixel value in the j-th row and k-th column of the segmentation mask for the i-th region;

[0039] Calculate the sum of the areas A of all bubble regions;

[0040]

[0041] Where N is the total number of bubble regions.

[0042] This plan also includes:

[0043] The cross-sectional area of ​​a tire semi-finished product is determined by examining a tire cross-section image.

[0044] A second aspect of the present invention provides a tire semi-finished product cross-section bubble detection system, comprising:

[0045] An image acquisition module is used to move a vision detection device via a motion device, and to acquire captured images via the vision detection device.

[0046] An image stitching module is used to stitch the captured images to obtain a tire cross-section image;

[0047] The image analysis module is used to input the tire cross-section image into a preset tire semi-finished product cross-section bubble detection model. It extracts features from the tire cross-section image through an attention mechanism combined with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are marked to obtain a label file. Based on the bubble features, the area of ​​each bubble region in the label file is calculated. The ratio of the sum of the areas of all bubble regions to the area of ​​the tire semi-finished product cross-section is calculated to determine the bubble rate of the tire semi-finished product cross-section.

[0048] A third aspect of the present invention provides a testing device, which includes a method program for detecting air bubbles in the cross-section of a tire semi-finished product. When the method program for detecting air bubbles in the cross-section of a tire semi-finished product is executed by a processor, it implements the steps of the method for detecting air bubbles in the cross-section of a tire semi-finished product as described above.

[0049] This invention discloses a method, system, and device for detecting air bubbles in the cross-section of a tire semi-finished product. The method includes: moving a vision inspection device via a motion device to acquire captured images; stitching the captured images to obtain a tire cross-section image; inputting the tire cross-section image into a preset tire semi-finished product cross-section air bubble detection model; extracting features from the tire cross-section image using an attention mechanism combined with a multi-scale feature fusion module to determine air bubble features; marking air bubble regions containing air bubble features to obtain a labeling file; calculating the area of ​​each air bubble region in the labeling file based on the air bubble features; and calculating the ratio of the sum of the areas of all air bubble regions to the area of ​​the tire semi-finished product cross-section to determine the air bubble rate of the tire semi-finished product cross-section. This invention can detect the air bubble rate of tire semi-finished product cross-sections more effectively and quickly, improving the efficiency of tire semi-finished product inspection. Attached Figure Description

[0050] Figure 1 A flowchart of a method for detecting air bubbles in the cross-section of a tire semi-finished product provided by the present invention is shown;

[0051] Figure 2 A flowchart of the tire cross-section image acquisition method provided by the present invention is shown;

[0052] Figure 3 A flowchart of the method for calculating the area of ​​a bubble region provided by the present invention is shown;

[0053] Figure 4 A block diagram of a tire semi-finished product cross-section bubble detection system provided by the present invention is shown;

[0054] Figure 5 A schematic diagram of the tire semi-finished product provided by the present invention is shown;

[0055] Figure 6 This diagram illustrates the method of selecting bubble features using a bounding box, as provided by the present invention.

[0056] Figure 7 A schematic diagram of the bubble identification region provided by the present invention is shown. Detailed Implementation

[0057] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0058] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0059] Figure 1 A flowchart of a method for detecting air bubbles in the cross-section of a tire semi-finished product provided by the present invention is shown.

[0060] like Figure 1 As shown, this invention discloses a method for detecting air bubbles in the cross-section of a tire semi-finished product, comprising:

[0061] S102, the motion device drives the visual detection device to move, and the visual detection device acquires the captured image.

[0062] S104, stitch the captured images to obtain a tire cross-section image;

[0063] S106, Input the tire cross-section image into the preset tire semi-finished product cross-section bubble detection model, extract features from the tire cross-section image through the attention mechanism combined with the multi-scale feature fusion module, determine bubble features, mark the bubble region containing bubble features, and obtain the annotation file;

[0064] S108, Calculate the area of ​​each bubble region in the annotation file based on bubble features;

[0065] S110, calculate the ratio of the sum of the areas of all bubble regions to the cross-sectional area of ​​the tire semi-finished product, and determine the bubble rate of the tire semi-finished product cross-section.

[0066] According to an embodiment of the present invention, the inspection vision device includes an industrial camera, a lens, and an industrial power supply. Its main function is to photograph the cross-section of the workpiece being inspected, providing digital image support for the data processing section. The motion device includes a linear module, a motor driver, and a motor controller. Figure 5As shown, the shape of a semi-finished tire is generally elongated. To ensure sufficiently high resolution in the photographs, the camera captures only a portion of the tire at a time, using a moving part to guide the camera to capture images from different positions on the tire, ultimately stitching them together to form a complete cross-sectional image of the tire. Image features are extracted using a multi-feature fusion method, feature matching is performed based on a geometrically constrained feature matching algorithm, image features are aligned using a global optimization method, and image stitching is performed using multi-band fusion technology. Finally, color consistency adjustments are made to the stitched result. This method outperforms traditional SURF feature-based methods in terms of feature matching accuracy, stitching quality, and computational efficiency, especially in low-texture areas and complex static scenes. It overcomes the limitations of traditional SURF feature-based image stitching methods in static scenes, such as insufficient feature extraction, low matching efficiency, and low alignment accuracy.

[0067] The photo-taking and image stitching commands are executed by an industrial control computer. The industrial control computer controls the industrial camera to take photos and receives the digital images returned by the camera. At the same time, it performs algorithmic stitching on multiple digital images to finally obtain a complete tire cross-section image.

[0068] The pre-defined tire semi-finished product cross-section bubble detection model is trained using historical tire cross-section images collected during historical inspections and sample tire cross-section images collected through online methods. This model is an improved YOLO model, enhancing small target detection, complex background handling, and computational efficiency. It addresses the limitations of existing YOLO models, such as insufficient small target detection capability, high computational resource requirements, and insufficient robustness to complex backgrounds.

[0069] The tire cross-section image is input into a pre-defined tire semi-finished product cross-section bubble detection model. A focus mechanism (CBAM) is introduced into this model to enhance its ability to extract key features. A multi-scale feature fusion module is designed to enhance the detection capability of small targets. An improved non-maximum suppression algorithm (DIoU-NMS) is used for post-processing of the detection results to reduce missed detections of overlapping targets. Deep learning algorithms are used to analyze the decoded tire image and detect bubbles in the tire semi-finished product cross-section. Figure 6 As shown, the bubble feature is selected by bounding box to identify the bubble region. The identification result is as follows. Figure 7 As shown, the number of pixels within the bubble region is counted, the area of ​​each bubble region is determined, and the ratio of the sum of the areas of all bubble regions to the cross-sectional area of ​​the tire semi-finished product is calculated to determine the bubble rate of the tire semi-finished product cross-section.

[0070] According to an embodiment of the present invention, it further includes:

[0071] Determine whether the stitched image is complete;

[0072] If not, the motion device will move the visual detection device to acquire the next captured image and continue to stitch the next captured image together.

[0073] If so, the stitched image will be identified as a tire cross-section image.

[0074] It should be noted that by analyzing the stitched images, when the edge of the stitched image is a tire cross-section, it is determined that the stitched image is incomplete. The motion device is then controlled to move the vision detection device to acquire the next image. The next image is then stitched together with the stitched image until the edge of the stitched image is no longer a tire cross-section, at which point the tire cross-section image is determined.

[0075] Figure 2 A flowchart of the tire cross-section image acquisition method provided by the present invention is shown.

[0076] like Figure 2 As shown, according to an embodiment of the present invention, image stitching of captured images includes:

[0077] S202, preprocesses the captured image; preprocessing includes noise reduction and illumination normalization;

[0078] S204. Use a multi-feature fusion method to extract key points and descriptors from the captured image to determine the image features of the captured image; the image features of the captured image include SURF features, edge features, and color features.

[0079] S206, a feature matching algorithm based on geometric constraints is used to match the image features of the captured image;

[0080] S208 analyzes the matched image features, aligns the captured images using a global optimization method, and performs image stitching using multi-band fusion technology to generate the stitching result;

[0081] S210, adjust the color consistency of the splicing results to obtain the tire cross-section image.

[0082] It should be noted that the captured images are denoised using methods such as wavelet thresholding preset by the system to eliminate high-frequency noise in the frequency domain while preserving the integrity of the image edge structure. Illumination normalization, as a preprocessing technique, can largely eliminate the impact of changes in illumination conditions on image feature recognition. By adjusting the brightness and contrast of the image, image standardization is achieved, allowing subsequent image feature extraction to be performed under relatively stable conditions.

[0083] Using a multi-feature fusion method and a parallel feature extraction architecture, key points and descriptors of the captured image are extracted through SURF feature extraction, edge feature extraction, and color feature extraction. The key points and descriptors extracted by the three methods are then fused to obtain the image features of the captured image.

[0084] A geometrically constrained feature matching algorithm is employed to improve the accuracy and robustness of matching by utilizing the geometric relationships between feature points to determine the matching point. Using this matching point as a constraint, global optimization methods such as bundle adjustment are used to optimize camera pose parameters and 3D point coordinates, minimizing the reprojection error function to complete image alignment. Multi-band fusion techniques, such as the Laplacian Pyramid, are used for multi-band fusion. Weighted averaging is applied to eliminate seams in the low-frequency band, while detail texture is preserved in the high-frequency band, thus completing image stitching. Finally, color consistency adjustment is performed on the stitched result. Color consistency adjustment can be achieved in the CIE Lab color space by using thin-plate spline interpolation to establish color mapping relationships between adjacent images and correct color differences in the stitched area. Furthermore, for ghosting artifacts present in the stitched result, a motion region detection algorithm can be used to identify moving objects (such as residual images of detection tools), and the corresponding weights in the fusion weight map are reset to zero, completely eliminating ghosting artifacts and ultimately outputting a complete tire cross-section image.

[0085] According to an embodiment of the present invention, it further includes:

[0086] Auxiliary feature matching is performed using the RANSAC algorithm.

[0087] It should be noted that the RANSAC (Random Sample Consensus) algorithm is an algorithm that estimates mathematical model parameters from a set of observed data containing outliers through an iterative process. In the image feature matching process, using the RANSAC algorithm to filter image features can effectively reduce false matches that occur during feature matching.

[0088] According to an embodiment of the present invention, feature extraction is performed on a tire cross-section image using an attention mechanism combined with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are then labeled to obtain a labeled file, including:

[0089] Feature extraction is performed on tire cross-section images by combining an attention mechanism with a multi-scale feature fusion module to identify bubble features within the tire cross-section images;

[0090] Select the bubble feature using a bounding box;

[0091] An improved nonmaximum suppression algorithm is used to filter bounding boxes, identify bubble regions containing bubble features, and generate annotation files.

[0092] It should be noted that the YOLO series of deep learning models have limitations in current technology, such as insufficient small target detection capability, high computational resource requirements, and insufficient robustness to complex backgrounds. The preset tire semi-finished product cross-section bubble detection model is an improved YOLO model. It introduces a focus mechanism (CBAM) to enhance the model's ability to extract key features; designs a multi-scale feature fusion module to enhance the detection capability of small targets; employs domain-specific data augmentation methods to improve the model's generalization ability in specific scenarios (such as industrial inspection scenarios); and uses an improved non-maximum suppression algorithm (DIoU-NMS) to post-process the detection results, reducing missed detections of overlapping targets.

[0093] Figure 3 A flowchart of the method for calculating the area of ​​the bubble region provided by the present invention is shown.

[0094] like Figure 3 As shown in the embodiment of the present invention, calculating the area of ​​each bubble region in the annotation file based on bubble features includes:

[0095] S302, Count the number of bubble targets in the annotation file to determine the number of bubbles;

[0096] S304, obtain the resolution of the tire cross-section image;

[0097] S306, convert the coordinates of the bubble region into actual pixel coordinates based on the resolution of the tire cross-section image;

[0098] S308, Read the segmentation mask of the bubble region;

[0099] S310, count the number of pixels in the segmentation mask whose pixel value is greater than a preset pixel value threshold, and determine the pixel area of ​​the bubble region.

[0100] It's important to note that the number of bubbles and the pixel area of ​​each bubble's region are calculated using the bubble features annotated in the annotation file. YOLO model annotation files are typically '.txt' files, with each line containing the category number and bounding box information (center point coordinates, width, and height, all normalized values) for a single bubble. The number of bubbles can be directly obtained by counting the number of targets in the YOLO annotation file. The total number of lines in the annotation file, with each line corresponding to one bubble target, represents the total number of bubbles.

[0101] The resolution (width W and height H) of the tire cross-section image is determined by the camera parameters and image resolution of the industrial camera in the vision inspection device. Based on the resolution of the tire cross-section image, the coordinates of the normalized bubble region are converted into actual pixel coordinates.

[0102] The final annotation in the annotation file is a pixel-level segmentation mask, which is usually stored in the form of a polygon point set. First, the polygon point set is converted into a binary mask image. By analyzing the binary mask image, the pixel value of each pixel in the binary mask image is compared with a preset pixel value threshold. The number of pixels with pixel values ​​greater than the preset pixel value threshold is counted to determine the pixel area of ​​the bubble region. The preset pixel value threshold is set by those skilled in the art according to actual needs.

[0103] According to an embodiment of the present invention, it further includes:

[0104] Overlay the segmentation masks of all bubble regions onto the same image;

[0105] Mark the overlapping areas obtained by mask overlay;

[0106] Based on the system's preset constraints, the pixels in the overlapping region are bound to a unique bubble region. The number of independent pixels in bubble region i is counted, and the area A of bubble region i is determined. i ;

[0107]

[0108] Where M is the segmentation mask, H and W represent the height and width of the image, respectively. i,j,k The pixel value in the j-th row and k-th column of the segmentation mask for the i-th region;

[0109] Calculate the sum of the areas A of all bubble regions;

[0110]

[0111] Where N is the total number of bubble regions.

[0112] It should be noted that, assuming the segmentation mask is M, containing N independent bubble regions, each bubble region is represented by a unique label i (i = 1, 2, ..., N). The segmentation masks of all bubble regions are superimposed onto the same image, and the number of masks corresponding to each pixel is counted. Pixels corresponding to multiple segmentation masks are labeled. After all pixels have been analyzed, the labeled pixels are integrated to determine the overlapping regions. For pixels in the overlapping regions, based on system-preset constraints (such as distance from the center point of the bubble region), the pixels in the overlapping regions are bound to the unique bubble regions, becoming independent pixels of those bubble regions.

[0113] According to an embodiment of the present invention, it further includes:

[0114] The cross-sectional area of ​​a tire semi-finished product is determined by examining a tire cross-section image.

[0115] It should be noted that the cross-sectional area of ​​the tire is determined by counting the number of pixels occupied by the tire cross-section in the tire cross-sectional image.

[0116] Figure 4 A block diagram of a tire semi-finished product cross-section bubble detection system provided by the present invention is shown.

[0117] like Figure 4 As shown, a second aspect of the present invention provides a tire semi-finished product cross-section bubble detection system, comprising:

[0118] The image acquisition module is used to move the visual inspection device via a motion device, and to acquire captured images via the visual inspection device.

[0119] The image stitching module is used to stitch together captured images to obtain a tire cross-section image;

[0120] The image analysis module is used to input tire cross-section images into a preset tire semi-finished product cross-section bubble detection model. It extracts features from the tire cross-section image through an attention mechanism combined with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are marked to obtain a label file. Based on the bubble features, the area of ​​each bubble region in the label file is calculated. The ratio of the sum of the areas of all bubble regions to the area of ​​the tire semi-finished product cross-section is calculated to determine the bubble rate of the tire semi-finished product cross-section.

[0121] A third aspect of the present invention provides a testing device, which includes a method program for detecting air bubbles in the cross-section of a tire semi-finished product. When the method program for detecting air bubbles in the cross-section of a tire semi-finished product is executed by a processor, it implements the steps of the method for detecting air bubbles in the cross-section of a tire semi-finished product as described above.

[0122] All information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals (including but not limited to signals transmitted between user terminals and other devices) involved in this application have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the "captured images" involved in this disclosure were obtained under full authorization.

[0123] This invention discloses a method, system, and device for detecting air bubbles in the cross-section of a tire semi-finished product. The method includes: moving a vision inspection device via a motion device to acquire captured images; stitching the captured images to obtain a tire cross-section image; inputting the tire cross-section image into a preset tire semi-finished product cross-section air bubble detection model; extracting features from the tire cross-section image using an attention mechanism combined with a multi-scale feature fusion module to determine air bubble features; marking air bubble regions containing air bubble features to obtain a labeling file; calculating the area of ​​each air bubble region in the labeling file based on the air bubble features; and calculating the ratio of the sum of the areas of all air bubble regions to the area of ​​the tire semi-finished product cross-section to determine the air bubble rate of the tire semi-finished product cross-section. This invention can detect the air bubble rate of tire semi-finished product cross-sections more effectively and quickly, improving the efficiency of tire semi-finished product inspection.

[0124] 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 illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0125] The units described above as separate components may or may not be physically separate. The components shown 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 to achieve the purpose of this embodiment according to actual needs.

[0126] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0127] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A method for detecting air bubbles in the cross-section of a tire semi-finished product, characterized in that, include: The visual detection device is moved by a motion device, and the visual detection device acquires captured images. The captured images are stitched together to obtain a tire cross-section image; The tire cross-section image is input into a preset tire semi-finished product cross-section bubble detection model. The tire cross-section image is used to extract features through an attention mechanism combined with a multi-scale feature fusion module to determine bubble features. Bubble regions containing bubble features are marked to obtain a labeling file. Calculate the area of ​​each bubble region in the annotation file based on bubble features; Calculate the ratio of the sum of the areas of all bubble regions to the cross-sectional area of ​​the tire semi-finished product to determine the bubble rate of the tire semi-finished product cross-section; The calculation of the area of ​​each bubble region in the annotation file based on bubble features includes: The number of bubbles is determined by counting the number of bubble targets in the annotation file. Obtain the resolution of the tire cross-section image; The coordinates of the bubble region are converted into actual pixel coordinates based on the resolution of the tire cross-section image; Read the segmentation mask of the bubble region; The number of pixels in the segmentation mask whose pixel values ​​are greater than a preset pixel value threshold is counted to determine the pixel area of ​​the bubble region; Also includes: Overlay the segmentation masks of all bubble regions onto the same image; Mark the overlapping areas obtained by mask overlay; Count the number of independent pixels in bubble region i, and determine the area A of bubble region i. i ; ； Where M is the segmentation mask, H and W represent the height and width of the image, respectively. i,j,k The pixel value in the j-th row and k-th column of the segmentation mask for the i-th region; Calculate the sum of the areas A of all bubble regions; ； Where N is the total number of bubble regions.

2. The method for detecting air bubbles in the cross-section of a tire semi-finished product according to claim 1, characterized in that, Also includes: Determine whether the stitched image is complete; If not, the visual detection device is moved by the motion device to acquire the next captured image, and the image stitching of the next captured image continues; If so, the stitched image is determined to be a tire cross-section image.

3. The method for detecting air bubbles in the cross-section of a tire semi-finished product according to claim 1, characterized in that, The step of stitching the captured images includes: The captured image is preprocessed; the preprocessing includes noise reduction and illumination normalization. Key points and descriptors of the captured image are extracted using a multi-feature fusion method to determine the image features of the captured image; the image features of the captured image include SURF features, edge features, and color features. A feature matching algorithm based on geometric constraints is used to perform feature matching on the captured image. Based on the analysis of the matched image features, the captured images are aligned using a global optimization method, and multi-band fusion technology is used to stitch the images together to generate the stitched result. The stitching results are adjusted for color consistency to obtain a tire cross-section image.

4. The method for detecting air bubbles in the cross-section of a tire semi-finished product according to claim 3, characterized in that, Also includes: Auxiliary feature matching is performed using the RANSAC algorithm.

5. The method for detecting air bubbles in the cross-section of a tire semi-finished product according to claim 1, characterized in that, The process involves extracting features from the tire cross-section image using an attention mechanism combined with a multi-scale feature fusion module, identifying bubble features, and marking bubble regions containing these features to obtain a labeled file, including: Feature extraction is performed on the tire cross-section image by combining an attention mechanism with a multi-scale feature fusion module to identify bubble features within the tire cross-section image; The bubble feature is selected by using a bounding box; The bounding box is filtered using an improved nonmaximum suppression algorithm to identify bubble regions containing bubble features and generate a labeling file.

6. The method for detecting air bubbles in the cross-section of a tire semi-finished product according to claim 1, characterized in that, Also includes: The cross-sectional area of ​​a tire semi-finished product is determined by examining a tire cross-section image.

7. A tire semi-finished product cross-section bubble detection system, used to implement the tire semi-finished product cross-section bubble detection method as described in any one of claims 1-6, characterized in that, include: An image acquisition module is used to move a vision detection device via a motion device, and to acquire captured images via the vision detection device. An image stitching module is used to stitch the captured images to obtain a tire cross-section image; The image analysis module is used to input the tire cross-section image into a preset tire semi-finished product cross-section bubble detection model. It extracts features from the tire cross-section image through an attention mechanism combined with a multi-scale feature fusion module, determines bubble features, and marks bubble regions containing bubble features to obtain a labeling file. Calculate the area of ​​each bubble region in the annotation file based on bubble features; Calculate the ratio of the sum of the areas of all bubble regions to the cross-sectional area of ​​the tire semi-finished product to determine the bubble rate of the tire semi-finished product cross-section.

8. A testing device, characterized in that, The testing equipment includes a method program for detecting air bubbles in the cross-section of a tire semi-finished product. When the method program for detecting air bubbles in the cross-section of a tire semi-finished product is executed by a processor, it implements the steps of the method for detecting air bubbles in the cross-section of a tire semi-finished product as described in any one of claims 1 to 6.

Images