Tire bubble disease detection and management method and device

Abstract

The invention discloses a tire bubble disease condition detection and management method and device. The method comprises the steps that a plurality of first phase images of different target areas of a target tire are obtained, and the target areas at least comprise a tread area and a sidewall area; analyzing each first phase image by using a pre-trained target segmentation model to obtain a plurality of second phase images without non-target region information; a pre-trained target detection model is utilized to analyze each second phase image, bubble state information in each second phase image is obtained, and the bubble state information at least comprises the bubble number, the bubble type and the bubble size; and determining a target disease condition grade of the target tire according to the multiple pieces of bubble state information, and executing a management operation corresponding to the target disease condition grade on the target tire. The technical problems that a traditional tire bubble detection scheme is inaccurate in detection result and low in detection efficiency are solved.

Description

Technical Field

[0001] This application relates to the field of tire quality testing technology, and more specifically, to a method and apparatus for detecting and managing tire bubble defects. Background Technology

[0002] In the tire manufacturing industry, ensuring that the quality and distribution of air bubbles inside tires meet high standards is crucial, as it directly affects the tire's safety performance and service life.

[0003] Traditional manual inspection or offline testing methods, such as visual inspection, X-ray, or ultrasound, while playing a vital role in the past, are increasingly showing their limitations. On the one hand, manual inspection is greatly affected by individual differences, lacks uniformity, and is inefficient, making it difficult to meet the demands of large-scale production. On the other hand, offline testing methods, such as ultrasonic testing and handheld PDA (Personal Digital Assistant) devices, while achieving a degree of automation, still suffer from sensitivity to the acoustic properties of specific materials, high equipment failure rates, and operational complexity, making it difficult to achieve high-precision and high-efficiency testing standards. In particular, ultrasonic testing may be affected by signal absorption or reflection when dealing with certain special materials, leading to decreased accuracy. Furthermore, relying on handheld devices for barcode reading and information input increases operational complexity, hindering the automation and intelligent transformation of production lines.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a method and apparatus for detecting and managing tire bubble defects, which at least solves the technical problems of inaccurate detection results and low detection efficiency in traditional tire bubble detection schemes.

[0006] According to one aspect of the embodiments of this application, a method for detecting and managing tire bubble defects is provided, comprising: acquiring multiple first phase images, each including different target regions of a target tire, wherein the target regions include at least: a tread region and a sidewall region; analyzing each first phase image using a pre-trained target segmentation model to obtain multiple second phase images from which non-target region information has been removed; analyzing each second phase image using a pre-trained target detection model to obtain bubble state information in each second phase image, wherein the bubble state information includes at least: the number of bubbles, the bubble type, and the bubble size; determining the target defect level of the target tire based on the multiple bubble state information, and performing management operations corresponding to the target defect level on the target tire.

[0007] Optionally, the target segmentation model is a tire sidewall segmentation model. The pre-trained target segmentation model is used to analyze each first phase image to obtain multiple second phase images after removing non-target region information. This includes: for a first phase image that includes the tread region of the target tire, the first phase image is directly used as the corresponding second phase image; for a first phase image that includes the tire sidewall region of the target tire, the tire sidewall segmentation model is used to analyze the first phase image to determine the tire sidewall detection box and mask region corresponding to the tire sidewall region in the first phase image, and the second phase image corresponding to the first phase image is determined based on the tire sidewall detection box and mask region.

[0008] Optionally, the training process of the tire sidewall segmentation model includes: constructing an initial segmentation model, wherein the initial segmentation model is a YOLO segmentation model with an added efficient channel attention module; acquiring multiple third-phase images including the tire sidewall region, and performing data augmentation processing on each third-phase image, using each obtained fourth-phase image and each third-phase image as a training sample, wherein the data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping; labeling the tire sidewall region in each training sample with detection boxes, and using the labeling information as the sample label of the corresponding training sample; iteratively training the initial segmentation model using multiple training samples and sample labels to obtain the tire sidewall segmentation model, wherein the loss function used in the iterative training process is the enhanced cross-union loss function.

[0009] Optionally, a pre-trained object detection model is used to analyze each second-phase image to obtain bubble state information in each second-phase image, including: for each second-phase image, the object detection model is used to analyze the second-phase image to determine the bubble detection box and bubble type corresponding to each bubble in the second-phase image, wherein the bubble type includes at least one of the following: butterfly-shaped bubble, transitional bubble, solid bubble, single-bubble bubble; the number of bubbles in the second-phase image is counted, and the bubble size corresponding to each bubble is determined according to the bubble detection box and bubble type corresponding to each bubble; the number of bubbles, the bubble type of each bubble, and the bubble size are used as the bubble state information corresponding to the second-phase image.

[0010] Optionally, the training process of the object detection model includes: constructing an initial detection model, wherein the initial detection model is a YOLO detection model with an added efficient channel attention module; acquiring multiple fifth-phase images of the target region including the tire, and performing data augmentation processing on each fifth-phase image, using each sixth-phase image and each fifth-phase image as a training sample, wherein the data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping; labeling the bubbles in each training sample with detection boxes and bubble type, and using the labeling information as the sample label of the corresponding training sample; iteratively training the initial detection model using multiple training samples and sample labels to obtain the object detection model, wherein the loss function used in the iterative training process is the augmented cross-union loss function.

[0011] Optionally, the bubble size corresponding to each bubble is determined based on the bubble detection box and bubble type, including: for each second phase image, determining the size mapping relationship between the pixel distance in the second phase image and the actual tire distance; for butterfly-shaped bubbles in the second phase image, performing edge detection in the bubble detection box corresponding to the butterfly-shaped bubble to obtain two fitted elliptical regions, determining the first pixel distance between the center points of the two elliptical regions, and determining the first bubble size of the butterfly-shaped bubble based on the first pixel distance and the size mapping relationship; for transitional or solid bubbles in the second phase image, performing edge detection in the bubble detection box corresponding to the transitional or solid bubbles to obtain fitted bubble regions, dividing the bubble regions into multiple bubble sub-regions, determining the average brightness value corresponding to each bubble sub-region, determining the second pixel distance between the center point of the first bubble sub-region with the largest average brightness value and the center point of the second bubble sub-region with the smallest average brightness value, and determining the second bubble size of the transitional or solid bubble based on the second pixel distance and the size mapping relationship; for single-unit bubbles in the second phase image, determining the third pixel distance between the two sets of opposite sides in the bubble detection box corresponding to the single-unit bubble, and determining the third bubble size of the single-unit bubble based on the larger third pixel distance and the size mapping relationship.

[0012] Optionally, the target lesion level of the target tire is determined based on multiple bubble state information, including: counting the total number of bubbles in multiple second-phase images; when the total number of bubbles is 0, the target lesion level of the target tire is determined as the lowest lesion level; when the total number of bubbles is 1, the target lesion level of the target tire is determined based on the bubble size of a single bubble; when the total number of bubbles is 2, the target lesion level of the target tire is determined based on the distribution state and bubble size of two bubbles; when the total number of bubbles exceeds 2, the target lesion level of the target tire is determined as the highest lesion level.

[0013] Optionally, determining the target disease level of the target tire based on the bubble size of a single bubble includes: determining the target size range to which the bubble size of a single bubble belongs from a preset disease level size table, and determining the disease level corresponding to the target size range as the target disease level of the target tire, wherein the disease level size table stores a preset mapping relationship between multiple size ranges and multiple disease levels.

[0014] Optionally, determining the target lesion level of the target tire based on the distribution state and size of the two bubbles includes: when the two bubbles are distributed in the same second phase image, determining the fourth pixel distance between the edges of the two bubbles, and determining the actual distance between the two bubbles based on the mapping relationship between the fourth pixel distance and the size; when the actual distance is less than a preset distance threshold, determining the sum of the bubble size of the two bubbles and the actual distance as the lesion size; when the actual distance is not less than the preset distance threshold, determining the sum of the bubble size of the two bubbles as the lesion size; when the two bubbles are distributed in different second phase images, determining the sum of the bubble size of the two bubbles as the lesion size; determining the target size range to which the lesion size belongs from a preset lesion level size table, and determining the lesion level corresponding to the target size range as the target lesion level of the target tire, wherein the lesion level size table stores a preset mapping relationship between multiple size ranges and multiple lesion levels.

[0015] Optionally, management operations corresponding to the target symptom level are performed on the target tire, including: determining the target color corresponding to the target symptom level from a preset symptom level color table, and displaying the status of the target tire based on the target color in the tire status display interface, wherein the symptom level color table stores the mapping relationship between different symptom levels and different colors; and sorting the target tire to the tire storage area corresponding to the target symptom level.

[0016] According to another aspect of the embodiments of this application, a device for detecting and managing tire bubble defects is also provided, comprising: an acquisition module, configured to acquire multiple first phase images, each including different target regions of a target tire, wherein the target regions include at least: a tread region and a sidewall region; a segmentation module, configured to analyze each first phase image using a pre-trained target segmentation model to obtain multiple second phase images from which non-target region information has been removed; a detection module, configured to analyze each second phase image using a pre-trained target detection model to obtain bubble state information in each second phase image, wherein the bubble state information includes at least: the number of bubbles, the bubble type, and the bubble size; and a management module, configured to determine the target defect level of the target tire based on the multiple bubble state information, and to perform management operations corresponding to the target defect level on the target tire.

[0017] According to another aspect of the embodiments of this application, a computer program product is also provided, the computer program product comprising: a computer program, wherein when the computer program is executed by a processor, it implements the above-described method for detecting and managing tire bubble defects.

[0018] According to another aspect of the embodiments of this application, an electronic device is also provided, the electronic device including: a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described method for detecting and managing tire bubble defects through the computer program.

[0019] In this embodiment, the target segmentation model can effectively distinguish the tire tread and sidewall, avoiding interference from non-target areas, thus providing a foundation for subsequent bubble detection. The target detection model can identify different types of bubble defects, including but not limited to butterfly-shaped bubbles, transitional bubbles, solid bubbles, and single-unit bubbles. By combining the models, bubble information on the tire can be effectively obtained, including but not limited to the number, type, and size of bubbles. Based on the bubble information, the tire defect level can be determined, thereby achieving automated tire sorting, reducing manual intervention, lowering production costs, and solving the technical problems of inaccurate detection results and low detection efficiency in traditional tire bubble detection schemes. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1 This is a flowchart illustrating an optional method for detecting and managing tire bubble defects according to an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of an optional butterfly-shaped bubble according to an embodiment of this application;

[0023] Figure 3 This is a schematic diagram of an optional solid bubble according to an embodiment of this application;

[0024] Figure 4 This is a schematic diagram of an optional monomeric bubble according to an embodiment of this application;

[0025] Figure 5 This is a schematic diagram illustrating an optional calculation of pathological dimensions according to an embodiment of this application;

[0026] Figure 6 This is a schematic diagram of an optional tire bubble detection and management device according to an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of the structure of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first," "second," etc., used in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] To better understand the embodiments of this application, the following is a translation and explanation of some nouns or terms that appear in the description of the embodiments of this application:

[0031] Phase images, in non-destructive testing (NDT) techniques, are images generated by analyzing the phase differences of signals reflected from within a material. In tire bubble detection, phase images, which may be generated using techniques such as X-rays, ultrasound, or magnetic resonance imaging (MRI), are used to reveal the internal structure of the tire, particularly the presence and distribution of bubbles. These images provide detailed information about internal material defects and are of significant value in detecting bubble defects within tires.

[0032] Butterfly-shaped bubbles: These are bubble defects that form in specific areas inside the tire, resembling the outstretched wings of a butterfly. They typically form on the tire's sidewalls or tread. These bubbles are usually caused by incomplete bonding between material layers or residual air during tire manufacturing. Their unique shape makes them relatively easy to distinguish in image recognition. The presence of butterfly-shaped bubbles can affect tire performance and lifespan, therefore requiring special attention in tire quality control.

[0033] Transitional bubbles: Transitional bubbles refer to a type of bubble inside a tire whose boundaries are not very clear, and may have a gradual transition zone with other material layers or bubbles. These bubbles typically appear at the junctions of tire material layers, where variations in material flow and curing conditions during manufacturing result in a blurred transition zone between the bubble and the surrounding material layers. Detecting transitional bubbles requires more sophisticated image processing techniques because they may not be as obvious as other types of bubbles.

[0034] Solid bubbles: Solid bubbles are another form of bubble inside a tire, characterized by well-defined boundaries, a relatively dense or opaque interior, and a clear contrast with the surrounding material layers. In phase images, these bubbles appear as a closed, solid circular or near-circular area. The presence of solid bubbles can affect the structural integrity and mechanical properties of the tire, therefore requiring identification and handling during the quality control phase.

[0035] Single-unit bubbles: Also known as large single bubbles, single-unit bubbles are typically larger bubbles (over 12mm) and refer to isolated defects within the tire, with clear boundaries and relatively fixed sizes. They usually do not form a network with other bubbles but exist independently. Detection of single-unit bubbles is relatively straightforward, but their impact on tire performance still needs to be evaluated. In image recognition, single-unit bubbles can be identified by analyzing the shape and size of the detection box.

[0036] Example 1

[0037] According to an embodiment of this application, a method for detecting and managing tire bubble defects is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0038] Figure 1 This is a flowchart illustrating a method for detecting and managing tire bubble defects according to an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0039] Step S102: Acquire multiple first phase images that include different target regions of the target tire, wherein the target regions include at least: the tread region and the sidewall region.

[0040] Specifically, the multiple first phase images can be obtained by using X-ray or ultrasound technology through a tire bubble machine. It should be noted that the target areas of the tire contained in each first phase image do not overlap.

[0041] Step S104: Analyze each first phase image using a pre-trained target segmentation model to obtain multiple second phase images from which non-target region information has been removed.

[0042] To avoid interference from background information, a target segmentation model needs to be introduced to analyze each first phase image, remove background interference, and obtain multiple second phase images that do not contain information about non-target regions.

[0043] Step S106: Analyze each second phase image using a pre-trained target detection model to obtain bubble state information in each second phase image. The bubble state information includes at least the number of bubbles, bubble type, and bubble size.

[0044] The target detection model can analyze the second phase image to obtain information reflecting the bubble state. The analysis of bubble state information helps to judge and classify tire defects in the later stage, thus facilitating the classification and management of tires in the later stage.

[0045] Step S108: Determine the target disease level of the target tire based on the status information of multiple bubbles, and perform management operations corresponding to the target disease level on the target tire.

[0046] The following describes the steps involved in the detection and management of tire bubble defects, using a specific implementation process as an example.

[0047] Generally speaking, the tread and sidewall of a tire have significant differences in physical structure. The tread is mainly responsible for direct contact with the ground, bearing the weight of the vehicle and friction, while the sidewall connects the tread to the rim, while also providing some support and cushioning.

[0048] Due to differences in structure and function, the distribution of air bubbles on the tread and sidewall differs, as does their impact on tire performance. The tread typically has a relatively uniform and regular structure; even after segmentation model processing, the output is often still a complete image of the tread because there are fewer air bubbles, and the model does not need to perform fine segmentation. Conversely, the sidewall has a more complex structure, and air bubbles may be hidden in folds, seams, or other difficult-to-detect locations. Therefore, the model needs to perform more refined segmentation to highlight sidewall features and ensure accurate air bubble detection. Thus, as a preferred implementation, the target segmentation model is a sidewall segmentation model. After sidewall segmentation, the model can focus on the sidewall image, which helps improve the accuracy of air bubble detection. Training a mature and complete target segmentation model is crucial for improving air bubble detection accuracy.

[0049] As an optional implementation, the training process of the tire side segmentation model includes the following steps:

[0050] Step S1: Construct an initial segmentation model, which is a YOLO segmentation model with an added efficient channel attention module.

[0051] Specifically, we can build upon the YOLOv8 segmentation model and introduce the ECA (Efficient Channel Attention) mechanism into its Backbone layer, which integrates SPP (Spatial Pyramid Pooling) and FPN (Feature Pyramid Network), to form an improved YOLOv8 object segmentation model. The ECA attention mechanism efficiently calculates inter-channel dependencies through one-dimensional convolution, reducing model complexity while improving feature extraction and fusion performance, making it particularly suitable for fine-grained segmentation of tire sidewall targets.

[0052] The principle of the attention mechanism can be summarized as follows: avoiding dimensionality reduction operations in the channel attention module, and achieving efficient channel attention computation through one-dimensional convolution by adopting a local cross-channel interaction strategy. This method maintains performance while significantly reducing model complexity, and determines the coverage of local cross-channel interactions by adaptively selecting the convolution kernel size. The ECA module achieves significant performance improvements with few parameters and low computational cost, and has higher efficiency and better performance compared to other attention modules.

[0053] The operation process of one-dimensional convolution is as follows:

[0054] First, global average pooling is performed on the input feature map of size W×H×C, which means adding up all the pixel values ​​of the feature map and averaging them to obtain a numerical value. This numerical value is used to represent the corresponding feature map, resulting in a 1×1×C feature map, where W represents the width, H represents the height, and C represents the number of channels.

[0055] Secondly, the kernel size k of the one-dimensional convolution is adaptively calculated based on the number of channels. The formula for calculating the kernel size is as follows:

[0056]

[0057] The above formula is used to calculate the kernel size of a one-dimensional convolution, where C is the number of channels in the input features, and γ and b are hyperparameters. Taking the absolute value and rounding down to the nearest odd number ensures that the kernel size is odd.

[0058] Next, a padding is set to ensure that the number of output channels remains unchanged. Then, a one-dimensional convolution with a kernel size of k is used to process the 1×1×C feature map. Finally, a sigmoid activation function is used to activate the final 1×1×C attention weights.

[0059] Finally, the obtained 1×1×C attention weights are applied to the input features to learn the importance of each channel relative to other channels. This process can be represented by the following formula:

[0060] out = Conv1D K (in)

[0061] The above formula represents the transformation of input feature *in* into output feature *out* through a one-dimensional convolution operation (kernel size *k*). Conv1D K This is a one-dimensional convolution operation.

[0062] Step S2: Obtain multiple third-phase images including the tire sidewall region, and perform data augmentation processing on each third-phase image. Use each fourth-phase image and each third-phase image as a training sample. The data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping.

[0063] For the multiple third-phase images containing the tire sidewall region, in order to enhance the generalization ability of the model, the original images need to be augmented. This includes, but is not limited to, random rotation (to adapt to the uncertainty of the tire rotation angle), random scaling (to take into account the changes in tire size), and random flipping (to simulate the diversity of tire mounting orientation). The augmented images are called fourth-phase images, which together with the original images form training samples, expanding the image library required for model training and enabling it to better adapt to various situations in actual production.

[0064] Step S3: Mark the detection box in the tummy side region of each training sample and use the marked information as the sample label of the corresponding training sample; use multiple training samples and sample labels to iteratively train the initial segmentation model to obtain the tummy side segmentation model. The loss function used in the iterative training process is the enhanced crossover ratio loss function.

[0065] For each training sample, we used the Labelme tool to accurately annotate the detection boxes, ensuring accurate labeling of the tire sidewall region. The annotated detection box information, along with the image, was used as input to the training samples, and the annotation information was used as the labels for the corresponding training samples. Next, we used these training samples and their labels to iteratively train the initial segmentation model.

[0066] In actual training, the enhanced cross-union ratio (CUNR) loss function can be used instead of the traditional full CUNR loss function. The enhanced CUNR loss function not only considers the cross-union ratio between the predicted and ground truth boxes but also incorporates distance and orientation factors, making the model more accurate in locating the tire region and achieving faster convergence. The hyperparameters used during model training include, but are not limited to: number of training epochs, batch size, initial learning rate, optimizer weight decay, and learning rate momentum. The enhanced CUNR loss function can be calculated using the following formula:

[0067] L EIoU =L IoU +L dis +L asp

[0068] Among them, L IoU It is the intersection-union loss, used to measure the overlap area between the predicted bounding box and the ground truth bounding box, L. dis It is the distance loss, used to calculate the Euclidean distance between the center point of the detection box and the center point of the ground truth box to evaluate the positional deviation; L asp It is a directional loss, which optimizes the shape of the bounding box by penalizing the difference in width and height.

[0069] In the above process, using a YOLO model with an added efficient channel attention module for tire sidewall segmentation significantly improves the model's ability to capture details, especially in recognizing tire sidewall regions against complex backgrounds. The introduction of data augmentation techniques, such as random rotation, scaling, and flipping, increases the model's generalization ability, enabling it to maintain high detection accuracy when faced with tire images of different angles, sizes, and positions. The use of an enhanced intersection-over-union (IoU) loss function further optimizes the model's training process, improving the accuracy and consistency of segmentation.

[0070] Furthermore, after determining the second phase image containing the target region, it is necessary to locate and identify the bubbles in the second phase image. This requires a mature target detection model. Therefore, training a mature and complete target detection model is of great significance for accurately locating and identifying bubble state information.

[0071] As an optional implementation, the training process of the object detection model includes the following steps: constructing an initial detection model, wherein the initial detection model is a YOLO detection model with an added efficient channel attention module; acquiring multiple fifth-phase images of the target region including the tire, and performing data augmentation processing on each fifth-phase image, using each sixth-phase image and each fifth-phase image as a training sample, wherein the data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping; labeling the bubbles in each training sample with detection boxes and bubble type, and using the labeling information as the sample label of the corresponding training sample; iteratively training the initial detection model using multiple training samples and sample labels to obtain the object detection model, wherein the loss function used in the iterative training process is the enhanced cross-union loss function.

[0072] As can be seen from the above training process of the object detection model, the training process of the object detection model and the training process of the tire side segmentation model are roughly the same. Both first build an initial model, then obtain sample data, and finally use the sample data to iteratively train the initial model. Therefore, the training process of the object detection model will not be described in detail here. The main point is to explain the differences between the object detection model and the tire side segmentation model in the training process. The differences mainly include two aspects: the type of initial model and the labeled objects.

[0073] During the training of the object detection model, the initial model can be the YOLOv8 detection model. When generating sample data for specific annotation information, the Labelme tool can be used for precise bounding box annotation. The bounding boxes correspond to the bubble type. For example, transition bubbles are labeled as Transition bubble, solid bubbles as Solid bubble, butterfly bubbles as Butterfly bubble, individual bubbles as Individual bubble, and other interference information as Other. The annotated images can be used as sample data for subsequent iterative training.

[0074] Target detection models can accurately identify the type and size of tire bubbles, which is crucial for the classification and assessment of tire defects. By statistically analyzing the number and type of bubbles, combined with measurements of bubble size, a comprehensive assessment of tire defects can be achieved, providing data support for subsequent defect grading and management.

[0075] As an optional implementation, each first phase image is analyzed using a pre-trained target segmentation model to obtain multiple second phase images after removing non-target region information. This can be achieved in the following way:

[0076] For a first phase image that includes the tread area of ​​the target tire, the first phase image is directly used as the corresponding second phase image.

[0077] Specifically, a first phase image including the tread region of the target tire can be input into the target segmentation model. The target segmentation model does not process it and directly outputs the first phase image, which is then used as the second phase image. Alternatively, the first phase image can be used directly as the second phase image without going through the target segmentation model.

[0078] For a first phase image including the sidewall region of the target tire, the sidewall segmentation model is used to analyze the first phase image to determine the sidewall detection box and mask region corresponding to the sidewall region in the first phase image, and the second phase image corresponding to the first phase image is determined based on the sidewall detection box and mask region.

[0079] Specifically, the model first performs preliminary localization of the tire sidewall based on an improved YOLOv8 network architecture. This localization process produces bounding boxes that surround the outline of the tire sidewall. Although these bounding boxes may not be perfectly precise at this stage, they provide a general location of the sidewall region. Building upon this preliminary localization, the model further refines the feature extraction process using the ECA attention mechanism, generating a more accurate segmentation mask. This mask not only defines the exact extent of the tire sidewall but also removes non-sidewall portions of the image, such as the background and other tire components, ensuring that the segmented sidewall represents the purest region of interest.

[0080] By combining the detection bounding box and the segmentation mask, we can accurately extract the lateral region from the first phase image to obtain the second phase image. The detection bounding box provides a rough location of the lateral region, while the mask ensures that pixels within the region belong only to the lateral region; the combination of the two yields the precise boundary of the lateral region.

[0081] The resulting second-phase image is then fed into a pre-trained target detection model for accurate detection of bubble lesions.

[0082] As an optional implementation, the pre-trained target detection model is used to analyze each second-phase image to obtain the bubble state information in each second-phase image. This can be achieved through the following steps:

[0083] Step S1: For each second phase image, the target detection model is used to analyze the second phase image to determine the bubble detection box and bubble type corresponding to each bubble in the second phase image. The bubble type includes at least one of the following: butterfly-shaped bubble, transition bubble, solid bubble, and single-bubble bubble.

[0084] Step S2: Count the number of bubbles in the second phase image, and determine the bubble size corresponding to each bubble based on the bubble detection box and bubble type.

[0085] Specifically, step S2 can be achieved through the following steps S21 to S24:

[0086] Step S21: For each second phase image, determine the size mapping relationship between the pixel distance in the second phase image and the actual tire distance.

[0087] Specifically, when determining the mapping relationship, the ratio of the circumferential length of a certain tire sidewall region to the pixel width of that sidewall region in the corresponding second phase image can be selected as a scale to calculate the actual distance represented by each pixel. For example, if the circumferential length of a tire sidewall region is 100 cm and the pixel width of the sidewall image is 200 pixels, then the scale is 100 cm / 200 pixels = 0.5 cm / pixel, that is, each pixel represents 0.5 cm. The solved scale relationship can be used as a size mapping relationship for the subsequent mapping calculation of the bubble size to the actual size of the bubble in the relevant target region in the second phase image.

[0088] Step S22: For the butterfly-shaped bubble in the second phase image, perform edge detection in the bubble detection box corresponding to the butterfly-shaped bubble to obtain two fitted elliptical regions, determine the first pixel distance between the center points of the two elliptical regions, and determine the first bubble size of the butterfly-shaped bubble based on the first pixel distance and the size mapping relationship.

[0089] Figure 2 The diagram shows an optional butterfly-shaped bubble. As can be seen, by detecting and fitting the edge of the bubble, two fitted elliptical regions can be obtained. These two elliptical regions are adjacent. The center points ① and ② of these two elliptical regions are further obtained. Connecting these two center points yields a straight line A. The straight line A is used as the pixel distance of the butterfly-shaped bubble. Furthermore, by combining the pixel distance with the size mapping relationship determined in step S21, the actual size of the butterfly-shaped bubble can be obtained.

[0090] Step S23: For transitional or solid bubbles in the second phase image, perform edge detection in the bubble detection box corresponding to the transitional or solid bubble to obtain the fitted bubble region. Divide the bubble region into multiple bubble sub-regions, determine the average brightness value corresponding to each bubble sub-region, determine the second pixel distance between the center point of the first bubble sub-region with the largest average brightness value and the center point of the second bubble sub-region with the smallest average brightness value, and determine the second bubble size of the transitional or solid bubble based on the second pixel distance and size mapping relationship.

[0091] While transitional bubbles may exhibit more elongated or irregular shapes, and solid bubbles may be closer to circular or elliptical shapes, their dimensions can be approximated by the straight-line distance connecting the two ends of the bubble. This method applies to any bubble morphology with clearly defined two ends, regardless of its shape, as long as the two ends of the bubble can be accurately detected, allowing the bubble's size and position information to be characterized by calculating the straight-line distance. Therefore, in this application, the bubble size is calculated in the same way for transitional and solid bubbles. The following uses a solid bubble as an example to illustrate the size calculation process for these two types of bubbles.

[0092] Figure 3 The diagram illustrates an optional solid bubble. Based on the target detection results, the detection box corresponding to the solid bubble is located. Edge detection processing is performed on this bubble. Based on the pixel size, points ① and ② are determined as the center of the lighter-colored position and the center of the brighter-colored position within the bubble, respectively. Connecting these two points yields a straight line A. The distance of line A is the pixel distance of the solid bubble. Furthermore, combining the pixel distance with the size mapping relationship determined in step S21, the actual size of the solid bubble can be obtained.

[0093] Since the calculation method for the bubble size is the same for solid bubbles and transitional bubbles, the calculation of the size for transitional bubbles will not be elaborated further here.

[0094] Step S24: For a single-unit bubble in the second phase image, determine the third pixel distance between the two pairs of opposite sides in the bubble detection frame corresponding to the single-unit bubble, and determine the third bubble size of the single-unit bubble based on the larger third pixel distance and the size mapping relationship.

[0095] Figure 4 A schematic diagram of an optional single-unit bubble is shown. The third pixel distance between the two sets of opposite sides of the detection box corresponding to the single-unit bubble is line segment A and line segment B, respectively. The pixel distance corresponding to the larger of line A and line B is directly used as the pixel distance of the single-unit bubble. Furthermore, by combining the pixel distance and the size mapping relationship determined in step S21, the actual size of the single-unit bubble can be obtained.

[0096] Step S3: The number of bubbles, the bubble type of each bubble, and the bubble size are used as the bubble state information corresponding to the second phase image.

[0097] Finally, the number of bubbles, the bubble type of each bubble, and the corresponding bubble size in the second phase image are used as the bubble state information of the second phase image.

[0098] After obtaining the bubble status information, the target lesion level of the target tire can be further determined based on multiple bubble status information. The specific determination method is as follows: count the total number of bubbles in multiple second-phase images; when the total number of bubbles is 0, the target lesion level of the target tire is determined as the lowest lesion level; when the total number of bubbles is 1, the target lesion level of the target tire is determined based on the size of a single bubble; when the total number of bubbles is 2, the target lesion level of the target tire is determined based on the distribution and size of the two bubbles; when the total number of bubbles exceeds 2, the target lesion level of the target tire is determined as the highest lesion level.

[0099] Specifically, when the total number of bubbles in all second images is 0, it indicates that there are no defects caused by bubbles in the tire. Therefore, the defect level of the target tire can be directly determined as the lowest defect level. If the total number of bubbles is 1, the defect level is further determined based on the size of a single bubble. If the total number of bubbles is 2, the defect level is determined based on the distribution and size of the two bubbles. If the total number of bubbles exceeds 2, it indicates that the bubbles on the target tire are very serious, and the defect level of the target tire can be directly determined as the highest defect level.

[0100] As an optional implementation, the target disease level of the target tire can be determined based on the size of a single bubble. This can be achieved by: determining the target size range to which the size of a single bubble belongs from a preset disease level size table, and determining the disease level corresponding to the target size range as the target disease level of the target tire. The disease level size table stores a preset mapping relationship between multiple size ranges and multiple disease levels.

[0101] When the total number of bubbles is 1, the size of the bubble can be used as the tire's defect size L. Table 1 shows an optional defect size table. It should be noted that the interval division in Table 1 is only an example and does not constitute a specific limitation.

[0102] Table 1

[0103] Interval Name Dimensions of the lesion (unit: mm) Pathological grade Interval 1 0<L≤10 Grade A Interval 2 10<L<20 Class C Interval 3 20≤L Class D

[0104] It can be seen that when the size of a single bubble belongs to interval 1, the condition grade of the target tire is A; when the size of a single bubble belongs to interval 2, the condition grade of the target tire is C; and when the size of a single bubble belongs to interval 3, the condition grade of the target tire is D.

[0105] Correspondingly, if the total number of bubbles in multiple second-phase images reflecting the target tire is 0, the condition level of the target tire can be directly classified as Grade A; if the total number of bubbles in multiple second-phase images reflecting the target tire exceeds 2, the condition level of the target tire can be directly classified as Grade D.

[0106] As an optional implementation, determining the target lesion level of the target tire based on the distribution and size of two bubbles can be achieved in the following way: when the two bubbles are distributed in the same second phase image, the fourth pixel distance between the edges of the two bubbles is determined, and the actual distance between the two bubbles is determined based on the mapping relationship between the fourth pixel distance and the size; when the actual distance is less than a preset distance threshold, the sum of the bubble size and the actual distance of the two bubbles is determined as the lesion size; when the actual distance is not less than the preset distance threshold, the sum of the bubble size of the two bubbles is determined as the lesion size; when the two bubbles are distributed in different second phase images, the sum of the bubble size of the two bubbles is determined as the lesion size; the target size range to which the lesion size belongs is determined from a preset lesion level size table, and the lesion level corresponding to the target size range is determined as the target lesion level of the target tire, wherein the lesion level size table stores a preset mapping relationship between multiple size ranges and multiple lesion levels.

[0107] When the total number of bubbles is 2, we not only consider the individual size of each bubble, but also evaluate their relative positions. Figure 5 A schematic diagram of an optional pathological size calculation is shown.

[0108] If two bubbles appear in the same second-phase image, assuming the fourth pixel distance between the two bubbles and their edges is determined by a size mapping relationship, if the actual distance is less than a preset distance threshold (assuming the preset distance threshold is 30mm), the actual dimensions of the two bubbles are directly summed with the actual distance, i.e., the actual dimensions D2 and D3 of bubble Q and bubble W are summed with the actual distance D1. The sum is taken as the defect size of the target tire. If the actual distance is not less than the preset distance threshold, the actual dimensions of the two bubbles are directly summed, i.e., the actual dimensions D2 and D3 of bubble Q and bubble W are summed. The sum is taken as the defect size of the target tire. After obtaining the defect size, the target interval to which the defect size of the target tire belongs can be found according to the defect grade size table mentioned above, thereby further determining the defect grade of the target tire.

[0109] If two bubbles appear in different second-phase images, the sum of the actual sizes of the two bubbles is directly determined as the defect size of the target tire. Similarly, after obtaining the defect size, the target interval to which the defect size of the target tire belongs can be found according to the defect grade size table mentioned above, thereby further determining the defect grade of the target tire.

[0110] By comprehensively considering the number, type, and size of air bubbles, the system automates the determination of tire condition levels. This method not only simplifies the determination process but also improves the accuracy and consistency of the assessment, avoiding the subjectivity and errors inherent in manual determination.

[0111] As an optional implementation, the management operation corresponding to the target symptom level is performed on the target tire, which can be achieved in the following way: determine the target color corresponding to the target symptom level from the preset symptom level color table, and display the status of the target tire based on the target color in the tire status display interface, wherein the symptom level color table stores the mapping relationship between different symptom levels and different colors; sort the target tire to the tire storage area corresponding to the target symptom level.

[0112] Assume the preset symptom level color table is as shown in Table 2.

[0113] Table 2

[0114] Pathological grade color Grade A green Class C light red Class D red

[0115] By matching the colors with a preset disease level color table, the color reflecting the disease level of the target tire can be determined. Finally, based on the disease level of the target tire, the tire status can be displayed on the tire status display interface. For the detected A-level tires, the system will mark them in green; for C-level tires, they will be marked in light red; and for D-level tires, they will be marked in red, thus intuitively displaying the quality status of each tire on the interface.

[0116] Furthermore, the programmable logic controller (PLC) can be invoked to sort target tires according to their corresponding defect levels and place them on the appropriate conveyor paths. When a tire is marked as Grade A, the PLC will control the tire to continue running on the qualified tire production line until it enters the qualified tire storage area. The system front-end page will display a "Qualified" status and store the grading result, corresponding timestamps, device IDs, and other information in the database for traceability and analysis. When a tire is marked as Grade C, the PLC will automatically adjust the tire's conveying direction, directing it to the area requiring further inspection. The system front-end page will display a "Conditions Qualified, Inspection Required" status, marked in light red, and the grading result will be stored. In addition, the system will automatically send a warning message to the operator, prompting them to pay attention to and handle Grade C tires. When a tire is marked as Grade D, the PLC will immediately activate an emergency procedure, controlling the tire to fall off the current production line and enter the unqualified tire storage area. The system front-end page will display a "Unqualified, Significant Defects Exist" status, marked in red, and the grading result will be stored. At the same time, the system will issue an emergency warning to ensure that substandard tires are not mishandled or enter the subsequent production process.

[0117] The rating result and processing procedure for each tire will be recorded and stored in a database. This includes the tire's barcode information, disease level, rating time, operator feedback, etc. Through data analysis, potential problems on the production line can be identified. For example, frequent occurrences of grade D tires may indicate that equipment maintenance or algorithm optimization is needed, thereby continuously improving production efficiency and product quality.

[0118] Linking tire condition levels with management operations and using color coding to visually display tire status allows for rapid identification of tire condition levels, facilitating management and operation on the production floor. This color coding method, combined with automated sorting technology, achieves efficient tire classification and management, reduces manual intervention, and lowers production costs.

[0119] Through the above steps, automated detection and management of tire bubble defects can be achieved. Specifically, the target segmentation model can effectively distinguish between the tire tread and sidewall, avoiding interference from non-target areas, thus providing a foundation for subsequent bubble detection. The target detection model can identify different types of bubble defects, including but not limited to butterfly-shaped bubbles, transitional bubbles, solid bubbles, and single-celled bubbles. By combining these models, information on bubbles on the tire can be effectively obtained, including but not limited to the number, type, and size of bubbles. Based on the bubble information, the tire defect level can be determined, thereby achieving automated tire sorting, reducing manual intervention, lowering production costs, and solving the technical problems of inaccurate detection results and low detection efficiency in traditional tire bubble detection solutions.

[0120] Example 2

[0121] According to an embodiment of this application, a device for detecting and managing tire bubble defects, used to implement the method for detecting and managing tire bubble defects in Embodiment 1, is also provided, such as... Figure 6 As shown, the tire bubble detection and management device includes at least: an acquisition module 61, a segmentation module 62, a detection module 63, and a management module 64, wherein:

[0122] The acquisition module 61 is used to acquire multiple first phase images that include different target regions of the target tire, wherein the target regions include at least: the tread region and the sidewall region;

[0123] The segmentation module 62 is used to analyze each first phase image using a pre-trained target segmentation model to obtain multiple second phase images after removing non-target region information.

[0124] The detection module 63 is used to analyze each second phase image using a pre-trained target detection model to obtain bubble state information in each second phase image. The bubble state information includes at least the number of bubbles, bubble type, and bubble size.

[0125] The management module 64 is used to determine the target disease level of the target tire based on multiple bubble status information, and to perform management operations on the target tire corresponding to the target disease level.

[0126] The following section describes the functions of each module of the tire bubble detection and management device, based on a specific implementation process.

[0127] Optionally, the device for detecting and managing tire bubble defects may also include a training module.

[0128] As an optional implementation, the training module's training process for the tire sidewall segmentation model includes: constructing an initial segmentation model, wherein the initial segmentation model is a YOLO segmentation model with an added efficient channel attention module; acquiring multiple third-phase images including the tire sidewall region, and performing data augmentation processing on each third-phase image, using each obtained fourth-phase image and each third-phase image as a training sample, wherein the data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping; labeling the tire sidewall region in each training sample with detection boxes, and using the labeling information as the sample label of the corresponding training sample; iteratively training the initial segmentation model using multiple training samples and sample labels to obtain the tire sidewall segmentation model, wherein the loss function used in the iterative training process is the enhanced cross-union loss function.

[0129] As an optional implementation, the target segmentation model is a tire sidewall segmentation model. The segmentation module uses a pre-trained target segmentation model to analyze each first phase image to obtain multiple second phase images after removing non-target region information. This can be achieved in the following way: for a first phase image that includes the tread region of the target tire, the first phase image is directly used as the corresponding second phase image; for a first phase image that includes the tire sidewall region of the target tire, the tire sidewall segmentation model is used to analyze the first phase image to determine the tire sidewall detection box and mask region corresponding to the tire sidewall region in the first phase image, and the second phase image corresponding to the first phase image is determined based on the tire sidewall detection box and mask region.

[0130] As an optional implementation, the detection module analyzes each second-phase image using a pre-trained target detection model to obtain bubble state information in each second-phase image. This can be achieved as follows: For each second-phase image, the target detection model is used to analyze the second-phase image to determine the bubble detection box and bubble type corresponding to each bubble in the second-phase image. The bubble type includes at least one of the following: butterfly-shaped bubble, transitional bubble, solid bubble, and single-bubble bubble; the number of bubbles in the second-phase image is counted, and the bubble size corresponding to each bubble is determined based on the bubble detection box and bubble type; the number of bubbles, the bubble type of each bubble, and the bubble size are used as the bubble state information corresponding to the second-phase image.

[0131] As an optional implementation, the training module's training process for the object detection model includes: constructing an initial detection model, wherein the initial detection model is a YOLO detection model with an added efficient channel attention module; acquiring multiple fifth-phase images including the target region of the tire, and performing data augmentation processing on each fifth-phase image, using each sixth-phase image and each fifth-phase image as a training sample, wherein the data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping; labeling the bubbles in each training sample with detection boxes and bubble type, and using the labeling information as the sample label of the corresponding training sample; iteratively training the initial detection model using multiple training samples and sample labels to obtain the object detection model, wherein the loss function used in the iterative training process is the enhanced cross-union loss function.

[0132] As an optional implementation, the detection module determines the bubble size of each bubble based on the bubble detection frame and bubble type. This can be achieved in the following way: For each second phase image, determine the size mapping relationship between the pixel distance in the second phase image and the actual tire distance; for butterfly-shaped bubbles in the second phase image, perform edge detection in the bubble detection frame corresponding to the butterfly-shaped bubble to obtain two fitted elliptical regions, determine the first pixel distance between the center points of the two elliptical regions, and determine the first bubble size of the butterfly-shaped bubble based on the first pixel distance and the size mapping relationship; for transitional or solid bubbles in the second phase image, determine the size of the transitional bubble... Edge detection is performed within the bubble detection frame corresponding to a solid bubble to obtain a fitted bubble region. This bubble region is then divided into multiple bubble sub-regions. The average brightness value of each bubble sub-region is determined. The second pixel distance between the center point of the first bubble sub-region with the largest average brightness value and the center point of the second bubble sub-region with the smallest average brightness value is determined. The second bubble size of the transitional or solid bubble is determined based on the mapping relationship between the second pixel distance and the size. For single-unit bubbles in the second phase image, the third pixel distance between the two pairs of opposite edges within the bubble detection frame corresponding to the single-unit bubble is determined. The third bubble size of the single-unit bubble is determined based on the larger third pixel distance and the size mapping relationship.

[0133] As an optional implementation, the management module determines the target lesion level of the target tire based on multiple bubble status information, which can be achieved in the following ways: count the total number of bubbles in multiple second-phase images; when the total number of bubbles is 0, determine the target lesion level of the target tire as the lowest lesion level; when the total number of bubbles is 1, determine the target lesion level of the target tire based on the size of a single bubble; when the total number of bubbles is 2, determine the target lesion level of the target tire based on the distribution state and size of two bubbles; when the total number of bubbles exceeds 2, determine the target lesion level of the target tire as the highest lesion level.

[0134] As an optional implementation, the management module determines the target disease level of the target tire based on the size of a single bubble. This can be achieved by: determining the target size range to which the size of a single bubble belongs from a preset disease level size table, and determining the disease level corresponding to the target size range as the target disease level of the target tire. The disease level size table stores a preset mapping relationship between multiple size ranges and multiple disease levels.

[0135] As an optional implementation, the management module determines the target lesion level of the target tire based on the distribution and size of the two bubbles. This can be achieved in the following ways: when the two bubbles are distributed in the same second phase image, the fourth pixel distance between the edges of the two bubbles is determined, and the actual distance between the two bubbles is determined based on the mapping relationship between the fourth pixel distance and the size; when the actual distance is less than a preset distance threshold, the sum of the bubble size and the actual distance of the two bubbles is determined as the lesion size; when the actual distance is not less than the preset distance threshold, the sum of the bubble size of the two bubbles is determined as the lesion size; when the two bubbles are distributed in different second phase images, the sum of the bubble size of the two bubbles is determined as the lesion size; the target size range to which the lesion size belongs is determined from a preset lesion level size table, and the lesion level corresponding to the target size range is determined as the target lesion level of the target tire. The lesion level size table stores a preset mapping relationship between multiple size ranges and multiple lesion levels.

[0136] As an optional implementation, the management module performs management operations on the target tire corresponding to the target symptom level, which can be achieved in the following way: determine the target color corresponding to the target symptom level from a preset symptom level color table, and display the status of the target tire based on the target color in the tire status display interface, wherein the symptom level color table stores the mapping relationship between different symptom levels and different colors; sort the target tire to the tire storage area corresponding to the target symptom level.

[0137] It should be noted that each module in the tire bubble detection and management device in this embodiment corresponds one-to-one with each implementation step of the tire bubble detection and management method in Embodiment 1. Since Embodiment 1 has been described in detail, some details not shown in this embodiment can be referred to Embodiment 1, and will not be elaborated further here.

[0138] Example 3

[0139] According to an embodiment of this application, a computer program product is also provided, which includes a computer program, wherein when the computer program is executed by a processor, it implements the tire bubble disease detection and management method in Embodiment 1.

[0140] According to an embodiment of this application, a non-volatile storage medium is also provided, which includes a stored computer program, wherein the device containing the non-volatile storage medium executes the tire bubble detection and management method in Embodiment 1 by running the computer program.

[0141] According to an embodiment of this application, a processor is also provided for running a computer program, wherein the computer program executes the tire bubble disease detection and management method in Embodiment 1.

[0142] According to an embodiment of this application, an electronic device is also provided, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the tire bubble disease detection and management method of Embodiment 1 through the computer program.

[0143] Specifically, the computer program executes the following steps during runtime: acquiring multiple first-phase images, each including different target regions of the target tire, wherein the target regions include at least: the tread region and the sidewall region; analyzing each first-phase image using a pre-trained target segmentation model to obtain multiple second-phase images from which non-target region information has been removed; analyzing each second-phase image using a pre-trained target detection model to obtain bubble state information in each second-phase image, wherein the bubble state information includes at least: the number of bubbles, the bubble type, and the bubble size; determining the target condition level of the target tire based on the multiple bubble state information, and performing management operations corresponding to the target condition level on the target tire.

[0144] As an alternative implementation, the above-mentioned electronic device may exist in the form of a mobile terminal, a computer terminal, or a similar computing device. Figure 7 A hardware block diagram of an electronic device for implementing a method for detecting and managing tire bubble defects is shown. Figure 7 As shown, the electronic device 70 may include one or more processors 702 (shown as 702a, 702b, ..., 702n in the figure) 702 (processor 702 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 704 for storing data, and a transmission device 706 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, electronic device 70 may also include components that are more... Figure 7 The more or fewer components shown, or having the same Figure 7 The different configurations shown.

[0145] It should be noted that the aforementioned one or more processors 702 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element of the electronic device 70. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0146] The memory 704 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the tire bubble detection and management method in this embodiment. The processor 702 executes various functional applications and data processing by running the software programs and modules stored in the memory 704, thereby implementing the aforementioned application vulnerability detection method. The memory 704 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 704 may further include memory remotely located relative to the processor 702, and these remote memories can be connected to the electronic device 70 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0147] The transmission device 706 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the electronic device 70. In one example, the transmission device 706 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 706 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0148] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the electronic device 70.

[0149] The sequence numbers of the above embodiments are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0150] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0151] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0152] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0153] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0154] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, 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 steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0155] The above are merely preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for detecting and managing tire bubble defects, characterized in that, include: Acquire multiple first phase images, each including a different target region of the target tire, wherein the target region includes at least: a tread region and a sidewall region; Each first phase image is analyzed using a pre-trained target segmentation model to obtain multiple second phase images from which non-target region information has been removed. Each second phase image is analyzed using a pre-trained target detection model to obtain bubble state information in each second phase image, wherein the bubble state information includes at least: number of bubbles, bubble type, and bubble size; The target disease level of the target tire is determined based on multiple bubble status information, and management operations corresponding to the target disease level are performed on the target tire.

2. The method according to claim 1, characterized in that, The target segmentation model is a tire side segmentation model. A pre-trained target segmentation model is used to analyze each of the first phase images to obtain multiple second phase images from which non-target region information has been removed, including: For a first phase image that includes the tread area of ​​the target tire, the first phase image is directly used as the corresponding second phase image; For a first phase image including the sidewall region of the target tire, the sidewall segmentation model is used to analyze the first phase image to determine the sidewall detection box and mask region corresponding to the sidewall region in the first phase image, and a second phase image corresponding to the first phase image is determined based on the sidewall detection box and the mask region.

3. The method according to claim 2, characterized in that, The training process of the tire side segmentation model includes: Construct an initial segmentation model, wherein the initial segmentation model is a YOLO segmentation model with an added efficient channel attention module; Multiple third-phase images, including the tire sidewall region, are acquired, and data augmentation processing is performed on each of the third-phase images. Each of the resulting fourth-phase images and each of the third-phase images are used as training samples. The data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping. A detection box is marked for the lateral region of the tire in each training sample, and the marked information is used as the sample label for the corresponding training sample. The initial segmentation model is iteratively trained using multiple training samples and sample labels to obtain the tire side segmentation model, wherein the loss function used in the iterative training process is the enhanced crossover ratio loss function.

4. The method according to claim 1, characterized in that, Each second phase image is analyzed using a pre-trained object detection model to obtain bubble state information in each second phase image, including: For each second phase image, the target detection model is used to analyze the second phase image to determine the bubble detection box and bubble type corresponding to each bubble in the second phase image. The bubble type includes at least one of the following: butterfly-shaped bubble, transitional bubble, solid bubble, and single-unit bubble. The number of bubbles in the second phase image is counted, and the bubble size corresponding to each bubble is determined based on the bubble detection box and bubble type. The number of bubbles, the bubble type of each bubble, and the bubble size are used as the bubble state information corresponding to the second phase image.

5. The method according to claim 4, characterized in that, The training process of the target detection model includes: Construct an initial detection model, wherein the initial detection model is a YOLO detection model with an added efficient channel attention module; Multiple fifth-phase images including the target area of ​​the tire are acquired, and data augmentation processing is performed on each of the fifth-phase images. Each of the resulting sixth-phase images and each of the fifth-phase images are used as training samples. The data augmentation processing includes at least one of the following: random rotation, random scaling, and random flipping. For each training sample, a detection bounding box and bubble type are labeled, and the labeling information is used as the sample label for the corresponding training sample. The initial detection model is iteratively trained using multiple training samples and sample labels to obtain the target detection model, wherein the loss function used in the iterative training process is the enhanced cross-union loss function.

6. The method according to claim 4, characterized in that, The bubble size is determined based on the bubble detection frame and bubble type corresponding to each bubble, including: For each second phase image, determine the dimensional mapping relationship between the pixel distance in the second phase image and the actual tire distance; For the butterfly-shaped bubble in the second phase image, edge detection is performed in the bubble detection box corresponding to the butterfly-shaped bubble to obtain two fitted elliptical regions. The first pixel distance between the center points of the two elliptical regions is determined, and the first bubble size of the butterfly-shaped bubble is determined based on the first pixel distance and the size mapping relationship. For transitional or solid bubbles in the second phase image, edge detection is performed in the bubble detection box corresponding to the transitional or solid bubble to obtain a fitted bubble region. The bubble region is divided into multiple bubble sub-regions, and the average brightness value corresponding to each bubble sub-region is determined. The second pixel distance between the center point of the first bubble sub-region with the largest average brightness value and the center point of the second bubble sub-region with the smallest average brightness value is determined. The second bubble size of the transitional or solid bubble is determined based on the second pixel distance and the size mapping relationship. For a single-unit bubble in the second phase image, the third pixel distance between two pairs of opposite sides in the bubble detection frame corresponding to the single-unit bubble is determined, and the third bubble size of the single-unit bubble is determined based on the larger third pixel distance and the size mapping relationship.

7. The method according to claim 6, characterized in that, The target disease level of the target tire is determined based on multiple bubble state information, including: Count the total number of bubbles in multiple second-phase images; When the total number of bubbles is 0, the target disease level of the target tire is determined to be the lowest disease level; When the total number of bubbles is 1, the target disease level of the target tire is determined based on the size of a single bubble; When the total number of bubbles is 2, the target disease level of the target tire is determined based on the distribution and size of the two bubbles. If the total number of bubbles exceeds two, the target disease level of the target tire is determined to be the highest disease level.

8. The method according to claim 7, characterized in that, Determining the target disease level of the target tire based on the size of individual bubbles includes: The target size range to which the size of the single bubble belongs is determined from a preset disease level size table, and the disease level corresponding to the target size range is determined as the target disease level of the target tire. The disease level size table stores a preset mapping relationship between multiple size ranges and multiple disease levels.

9. The method according to claim 7, characterized in that, The target defect level of the target tire is determined based on the distribution and size of the two bubbles, including: When the two bubbles are distributed in the same second phase image, the fourth pixel distance between the edges of the two bubbles is determined, and the actual distance between the two bubbles is determined based on the fourth pixel distance and the size mapping relationship; If the actual distance is less than a preset distance threshold, the sum of the bubble size of the two bubbles and the actual distance is determined as the lesion size; If the actual distance is not less than a preset distance threshold, the sum of the bubble sizes of the two bubbles is determined as the lesion size; When the two bubbles are distributed in different second-phase images, the sum of the bubble sizes of the two bubbles is determined as the lesion size; The target size range to which the lesion size belongs is determined from a preset lesion size table, and the lesion level corresponding to the target size range is determined as the target lesion level of the target tire. The lesion size table stores a preset mapping relationship between multiple size ranges and multiple lesion levels.

10. The method according to claim 9, characterized in that, Perform management operations corresponding to the target disease level on the target tire, including: The target color corresponding to the target symptom level is determined from the preset symptom level color table, and the status of the target tire is displayed in the tire status display interface based on the target color. The symptom level color table stores the mapping relationship between different symptom levels and different colors. The target tires are sorted into the tire storage area corresponding to the target symptom level.

11. A device for detecting and managing tire bubble defects, characterized in that, include: The acquisition module is used to acquire multiple first phase images, each including different target regions of the target tire, wherein the target regions include at least: a tread region and a sidewall region; The segmentation module is used to analyze each of the first phase images using a pre-trained target segmentation model to obtain multiple second phase images from which non-target region information has been removed. The detection module is used to analyze each of the second phase images using a pre-trained target detection model to obtain bubble state information in each of the second phase images, wherein the bubble state information includes at least: number of bubbles, bubble type and bubble size; The management module is used to determine the target disease level of the target tire based on multiple bubble status information, and to perform management operations on the target tire corresponding to the target disease level.

12. A computer program product, characterized in that, include: A computer program, wherein when executed by a processor, the computer program implements the method for detecting and managing tire bubble defects as described in any one of claims 1 to 10.

13. An electronic device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute, via the computer program, the method for detecting and managing tire bubble defects according to any one of claims 1 to 10.

Images