Tire detection method, device, system, program product and storage medium

Abstract

This application provides a tire inspection method, device, system, program product, and storage medium. In this embodiment, an image reconstruction model trained using historical X-ray images of normal tires is used to reconstruct the X-ray image of the tire to be inspected, resulting in a reconstructed image. Since the image reconstruction model is trained using historical X-ray images of normal tires, the reconstructed image and the X-ray image will differ significantly for defective tires. Based on this, the similarity between the X-ray image and the reconstructed image of the tire to be inspected can be calculated; and the quality of the tire to be inspected can be determined based on the similarity between the X-ray image and the reconstructed image, achieving automated tire quality inspection without relying on human experience, thus improving the accuracy of tire quality inspection.

Description

Technical Field

[0001] This application relates to the field of tire quality inspection technology, and in particular to a tire inspection method, equipment, system, program product and storage medium. Background Technology

[0002] Tire manufacturing demands high standards in its processes, and is susceptible to operational errors or equipment malfunctions that can lead to substandard tire quality. Tires may exhibit various defects, such as overlapping cords, sparse joints, foreign matter, and air bubbles on the sidewall or shoulder.

[0003] Currently, in the tire industry, X-rays are frequently used to detect defects in manufactured tires. However, relying on visual inspection of tire X-ray images leads to a significant problem of false positives. Summary of the Invention

[0004] This application provides a tire testing method, equipment, system, program product, and storage medium to improve the accuracy of tire quality testing and reduce the probability of false detection.

[0005] This application provides a tire inspection method, comprising: acquiring an image to be inspected; the image to be inspected being an X-ray image of a tire to be inspected; reconstructing the image to be inspected using an image reconstruction model to obtain a reconstructed image of the image to be inspected; the image reconstruction model being trained using historical X-ray images of normal tires as samples; calculating the similarity between the image to be inspected and the reconstructed image; and determining the quality of the tire to be inspected based on the similarity between the image to be inspected and the reconstructed image.

[0006] This application also provides a detection method, comprising: acquiring an image to be detected; the image to be detected being an image of an object to be detected; reconstructing the image to be detected using an image reconstruction model to obtain a reconstructed image of the image to be detected; wherein the image reconstruction model is trained using historical images of normal objects; calculating the similarity between the image to be detected and the reconstructed image; and determining the quality of the object to be detected based on the similarity between the image to be detected and the reconstructed image.

[0007] This application embodiment also provides a tire inspection system, including: an X-ray machine and a computer device;

[0008] The X-ray machine is used to emit X-rays onto the tire to be inspected; receive the radiation signal generated by the X-rays penetrating the tire to be inspected; convert the radiation signal into an X-ray image; and provide the X-ray image to the computer equipment.

[0009] The computer device is used to reconstruct the X-ray image using an image reconstruction model to obtain a reconstructed image of the X-ray image; calculate the similarity between the X-ray image and the reconstructed image; and determine the quality of the tire to be inspected based on the similarity between the X-ray image and the reconstructed image; wherein the image reconstruction model is trained using historical X-ray images of normal tires as samples.

[0010] This application embodiment also provides an X-ray machine, including: a mechanical body; the mechanical body is provided with a receiving cavity; an X-ray emitter and an X-ray detector are disposed in the receiving cavity;

[0011] The mechanical body is also equipped with a signal acquisition unit and a processing unit; the signal acquisition unit is electrically connected between the X-ray detector and the processing unit.

[0012] The cavity is used to hold the tire to be inspected; the X-ray emitter is used to emit X-rays outward; the X-rays can penetrate the tire to be inspected.

[0013] The X-ray detector is used to receive the radiation signal generated by the X-rays penetrating the tire to be tested;

[0014] The signal acquisition unit is used to acquire the radiation signal, convert the radiation signal into an X-ray image, and provide the X-ray image to the processing unit.

[0015] The processing unit is configured to reconstruct the X-ray image using an image reconstruction model to obtain a reconstructed image of the X-ray image; calculate the similarity between the X-ray image and the reconstructed image; and determine the quality of the tire to be detected based on the similarity between the X-ray image and the reconstructed image; wherein the image reconstruction model is trained using historical X-ray images of normal tires as samples.

[0016] This application embodiment also provides a computer device, including: a memory and a processor; wherein, the memory is used to store computer programs;

[0017] The processor is coupled to the memory and is used to execute the computer program to perform the steps in the above detection methods.

[0018] This application also provides a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps in the above-described tire detection method and / or detection method.

[0019] This application also provides a computer program product, including: a computer program; the computer program is executed by a processor to implement the above-described tire detection method and / or detection method.

[0020] In this embodiment, an image reconstruction model trained using historical X-ray images of normal tires is used to reconstruct the X-ray image of the tire to be inspected, resulting in a reconstructed image. Since the image reconstruction model is trained using historical X-ray images of normal tires, the reconstructed image and the X-ray image will differ significantly for defective tires. Based on this, the similarity between the X-ray image and the reconstructed image of the tire to be inspected can be calculated; and the quality of the tire to be inspected can be determined based on the similarity between the X-ray image and the reconstructed image. This achieves automated tire quality inspection without relying on human experience, thus improving the accuracy of tire quality inspection. Attached Figure Description

[0021] 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:

[0022] Figure 1a This is a schematic diagram of the tire inspection system provided in an embodiment of this application;

[0023] Figure 1b This is a schematic diagram of the internal structure of the image reconstruction model provided in the embodiments of this application;

[0024] Figure 1c This is a schematic diagram of the internal structure of the encoder provided in an embodiment of this application;

[0025] Figure 1d This is a schematic diagram of the internal structure of the decoder provided in an embodiment of this application;

[0026] Figure 1e This is a schematic diagram of the tire inspection process provided in an embodiment of this application;

[0027] Figure 1f and Figure 1g A schematic diagram illustrating the training process of the image reconstruction model provided in this application embodiment;

[0028] Figure 1h This is a schematic diagram of the similarity calculation process provided in the embodiments of this application;

[0029] Figure 2 A schematic flowchart of the tire testing method provided in the embodiments of this application;

[0030] Figure 3 A schematic flowchart of the detection method provided in the embodiments of this application;

[0031] Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of this application;

[0032] Figure 5 This is a structural block diagram of an X-ray machine provided in an embodiment of this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] To address the problem that existing tire quality inspection methods rely on human experience and have low accuracy, some embodiments of this application utilize an image reconstruction model trained using historical X-ray images of normal tires as samples to reconstruct the X-ray image of the tire to be inspected, resulting in a reconstructed image. Since the image reconstruction model is trained on historical X-ray images of normal tires, the reconstructed image and the X-ray image will differ significantly for defective tires. Based on this, the similarity between the X-ray image and the reconstructed image of the tire to be inspected can be calculated; and the quality of the tire to be inspected can be determined based on the similarity between the X-ray image and the reconstructed image, thus achieving automated tire quality inspection without relying on human experience and contributing to improved accuracy.

[0035] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0036] It should be noted that the same reference numerals denote the same object in the following figures and embodiments. Therefore, once an object is defined in one figure or embodiment, it does not need to be discussed further in subsequent figures and embodiments.

[0037] Tire manufacturing demands high standards in its processes, and is susceptible to operational errors or equipment malfunctions that can lead to substandard tire quality. Tires may exhibit various defects, such as overlapping cords, sparse joints, foreign matter, and air bubbles on the sidewall or shoulder.

[0038] X-ray-based non-destructive testing of tire interiors is a common method for tire quality inspection. X-rays have strong penetrating power, and the interior of a tire contains various composite materials, such as steel belt layers, crown belt layers, and carcass plies. Different materials absorb X-rays at different rates due to variations in density and thickness. Therefore, the amount of radiation received by the X-ray detector varies depending on the location penetrating the tire. Based on this, a photoelectric converter transforms the radiation into a corresponding electrical signal, followed by a series of analog-to-digital conversions and noise reduction operations to obtain the corresponding digital image.

[0039] Figure 1a This is a schematic diagram of the tire inspection system provided in an embodiment of this application. Figure 1a As shown, the system includes an X-ray machine 11 and a computer device 12.

[0040] X-ray machine 11 is a device that generates X-rays and converts the radiation signal of an object detected by X-rays into a corresponding X-ray image. In this embodiment, X-ray machine 11 can emit X-rays outward. Because X-rays have strong penetrating power, they can be used for internal flaw detection of objects. In addition to emitting X-rays outward, the X-ray machine 11 provided in this embodiment can also receive the radiation signal generated by the penetration of X-rays into an object and convert the radiation signal into a corresponding X-ray image.

[0041] Computer device 12 refers to an electronic device with communication and data processing functions. Computer device 12 can be a terminal device or a server device. A terminal device can be a smartphone, tablet, personal computer, or wearable device, etc. A server device can be a single server device, a cloud-based server array, or a virtual machine (VM) running within a cloud-based server array. Additionally, server device can also refer to other computing devices with corresponding service capabilities, such as computers or other terminal devices (running service programs), etc.

[0042] In this embodiment, the computer device 12 is communicatively connected to the X-ray machine 11. The computer device 12 can be deployed in the cloud or at the geographical location where the X-ray machine 11 is located, i.e., locally. The connection between the X-ray machine 11 and the computer device 12 can be wireless or wired. Optionally, the X-ray machine 11 can communicate with the computer device 12 via a mobile network. Accordingly, the mobile network standard can be any one of 2G (GSM), 2.5G (GPRS), 3G (WCDMA, TD-SCDMA, CDMA2000, UTMS), 4G (LTE), 4G+ (LTE+), 5G, WiMax, etc. Optionally, the X-ray machine 11 can also communicate with the computer device 12 via Bluetooth, WiFi, infrared, etc.

[0043] In this embodiment, the X-ray machine 11 emits X-rays towards the tire to be inspected. X-rays can penetrate the tire. Since the tire contains various composite materials, such as steel belt layers, crown belt layers, and carcass ply layers, different materials have different densities and thicknesses, resulting in different absorption rates of X-rays. Therefore, the radiation levels at different parts of the tire vary. Consequently, the amount of radiation generated at different parts of the tire after X-ray penetration is different. Based on this, the X-ray machine 11 can also receive radiation signals from different parts of the tire penetrated by X-rays and convert the radiation signals into corresponding electrical signals. Then, the electrical signals are processed to obtain an X-ray image of the tire. Optionally, the X-ray machine 11 can perform analog-to-digital conversion, noise reduction, and other processing on the electrical signals to obtain a corresponding digital image, i.e., an X-ray image of the tire.

[0044] Tires are susceptible to defects during manufacturing due to operational errors or equipment malfunctions, leading to substandard quality and various flaws such as overlapping cords, loose joints, foreign matter, and air bubbles on the sidewall or shoulder. These defects are visible in tire X-ray images. Therefore, the X-ray machine 11 provides the X-ray image of the tire to be inspected to the computer device 12, which then performs quality inspection based on the image. The process of quality inspection based on the X-ray image of the tire is described below.

[0045] like Figure 1a As shown, computer device 12 uses an image reconstruction model trained with historical X-ray images of normal tires as samples to reconstruct the X-ray image of the tire to be inspected, thereby obtaining a reconstructed image of the X-ray image of the tire to be inspected. In this embodiment, a normal tire refers to a tire with no internal damage, or a defect-free tire. These tires meet tire quality inspection standards and do not have the internal defects listed above.

[0046] Optionally, in the image reconstruction model, feature extraction can be performed on the X-ray image of the tire to be detected to obtain the image features of the X-ray image of the tire to be detected; and the image features of the X-ray image of the tire to be detected can be used to reconstruct the image to obtain the reconstructed image of the X-ray image.

[0047] Optionally, such as Figure 1b As shown, in the image reconstruction model, the encoder can be used to encode the X-ray image (X) of the tire to be detected in order to extract the image features Z of the X-ray image of the tire to be detected; further, the image features Z of the X-ray image are input into the decoder for decoding to obtain the reconstructed image G(X) of the X-ray image.

[0048] The features extracted from the input image by the encoder can be represented as:

[0049] η=En(X)(a)

[0050] In equation (a), X is the input X-ray image, En represents the encoder, and η is the feature map (future map) obtained after downsampling, which is the image feature Z of the X-ray image mentioned above.

[0051] In this application embodiment, the network architecture of the encoder and decoder is not limited. Optionally, as... Figure 1c As shown, the encoder can consist of multiple downsampling modules and pooling layers. "Multiple" refers to two or more. Each downsampling module may include a convolutional layer, a batch normalization layer, and an activation function layer. The activation function layer can use a rule function, a sigmoid function, or a tanh function, etc. In this embodiment, the stride of the convolutional layer in the downsampling module is not limited. Optionally, the stride of the convolutional layer can be 2, etc. Optionally, for an input X-ray image of size 256*256, after encoding by the encoder, it can be downsampled into a 4*4 feature map, and the points in the feature map can be normalized to 0~1.

[0052] The decoder uses upsampling to restore the obtained feature map to the original image size.

[0053] G(X)=De(η)(b)

[0054] In equation (b), De is the decoder, G(X) is the generated reconstructed image, and η is the feature map.

[0055] like Figure 1d As shown, the decoder consists of multiple upsampling modules. "Multiple" means two or more. Among them, Figure 1d The diagram illustrates four upsampling modules. Optionally, an upsampling module may include: a deconvolutional layer, a batch normalization layer, and an activation function layer. The stride of the deconvolutional layer is the same as the stride of the convolutional layer in the downsampling module of the encoder described above. The activation function used in the activation function layer is also the same as the activation function used in the downsampling module of the encoder described above. Optionally, as... Figure 1d As shown, the feature map can be input into other network layers before upsampling. For example... Figure 1d The deconvolutional layers shown include, but are not limited to, batch normalization layers and ReLU activation function layers. Optionally, the deconvolutional layer here can be composed of multiple deconvolutional layers with different strides, such as a deconvolutional layer with a stride of 2 and a deconvolutional layer with a stride of 1.

[0056] Optionally, such as Figure 1d As shown, the upsampled feature map can also be input into other network layers. For example... Figure 1d The diagram shows deconvolutional layers and sigmoid activation function layers, etc. The deconvolutional layer at this point can be a deconvolutional layer with a stride of 2.

[0057] Since the image reconstruction model is trained using historical X-ray images of normal tires as samples, the similarity between the X-ray image and the reconstructed image of a normal tire is high. Conversely, for defective tires, because the image reconstruction model has not learned from X-ray images of defective tires, it cannot reconstruct the defective regions from the X-ray images. Therefore, the difference between the X-ray image and the reconstructed image of a defective tire is large, and the similarity is low.

[0058] Based on the above analysis, computer device 12 can calculate the similarity between the X-ray image of the tire to be inspected and the reconstructed image of the X-ray image; and determine the quality of the tire to be inspected based on the similarity between the X-ray image of the tire to be inspected and the reconstructed image of the X-ray image, thus realizing automated tire quality inspection. This inspection process does not rely on human experience, which helps to improve the accuracy of tire quality inspection.

[0059] On the other hand, detection methods that rely on manual experience to visually identify tire defects from X-ray images are both mechanical and labor-intensive. Furthermore, the results of visual identification are easily affected by factors such as human experience and the operator's condition, leading to significant false positives and false negatives. The detection method provided in this embodiment, however, features a stable detection system, which helps reduce the probability of false negatives.

[0060] Furthermore, for tire quality detection models trained using X-ray images of defective tires (defined as defective X-ray images), the accuracy of the trained model is low because there are relatively few defective samples in tire X-ray images. On the other hand, since tire defects are diverse, the defects in defective X-ray images may not cover all types of tire defects. Therefore, the model cannot detect defects that have not been learned, leading to missed detections of defective tires and affecting detection accuracy. Moreover, the cost of labeling defective samples is extremely high.

[0061] This embodiment utilizes an image reconstruction model trained on historical X-ray images of normal tires (defined as normal X-ray images) for quality detection. On one hand, normal X-ray image samples are readily available and plentiful, resulting in a high accuracy of the trained image reconstruction model. On the other hand, since the image reconstruction model does not learn the image features of defective X-ray images, the reconstructed image of any tire, regardless of its defect, differs significantly from the original X-ray image. Therefore, the tire detection method provided in this embodiment can detect any defect, exhibiting good versatility and further reducing the probability of missed detections. Furthermore, since normal X-ray images do not require sample annotation, sample annotation costs are reduced.

[0062] In this embodiment of the application, before reconstructing the X-ray image of the tire to be inspected using the image reconstruction model, the image reconstruction model can be trained. The training process of the image reconstruction model can be offline or online. The training process of the image reconstruction model is described below as an example.

[0063] In this embodiment, as Figure 1e As shown, computer device 12 can acquire historical X-ray images of normal tires. Optionally, as... Figure 1e As shown, computer device 12 can acquire historical X-ray images of normal tires from a historical image database. Furthermore, computer device 12 can use these historical X-ray images of normal tires to train a model, thereby obtaining an image reconstruction model.

[0064] In this application embodiment, the implementation form of the image reconstruction model is not limited. Optionally, historical X-ray images of normal tires can be used as training samples to perform generative adversarial training on a generative adversarial network (GAN) until a set stopping condition is met; and the generator in the GAN when the set stopping condition is met is used as the image reconstruction model. The following is combined with... Figure 1f An illustrative explanation of the generative adversarial training process is provided. For example... Figure 1f As shown, the generative adversarial training process mainly includes the following steps:

[0065] S1. With minimizing the first loss function as the training objective, and using historical X-ray images of normal tires as training samples, train the current generator to obtain the target generator.

[0066] The first loss function is the generator's loss function, which can be determined based on historical X-ray images and the reconstructed image generated by the current generator.

[0067] Optionally, the first loss function may consist of two parts: consistency loss and generative adversarial loss.

[0068] Consistency Loss: The generator aims to produce a reconstructed image that is as similar as possible to the defect-free image, which can be achieved by minimizing the pixel differences between the input X-ray image and the reconstructed image through residual calculation. Accordingly, such as... Figure 1g As shown, the consistency loss can be expressed as:

[0069] Loss res =E X～P ||xG(x)||1 (1)

[0070] In equation (1), x represents the pixel value of the X-ray image, and G(x) represents the pixel value of the reconstructed image of the X-ray image. ||xG(x)||1 represents the 1-norm of the pixel difference between the X-ray image and the reconstructed image of the X-ray image.

[0071] Generative Adversarial Loss: Generative adversarial loss calculates the distance between image features on the discriminator, thereby optimizing the generator to make the generated reconstructed image more similar to the input X-ray image, ultimately generating a reconstructed image that can deceive the discriminator. The inventors discovered that image feature matching can reduce the instability of GAN training. Therefore, they extracted the intermediate layer of the discriminator, matched the image features of the input X-ray image with those of the corresponding reconstructed image, and calculated the distance between the image features of the X-ray image and the corresponding reconstructed image. Optionally, the Euclidean distance or cosine distance between the image features of the X-ray image and the corresponding reconstructed image can be calculated. For example, Figure 1g As shown, the generative adversarial loss can be defined as:

[0072] Loss adv =E x～p ||f(x)-f(G(x))||2 (2)

[0073] In equation (2), f(x) represents the image features of the X-ray image, and f(G(x)) represents the image features of the reconstructed image of the X-ray image. ||f(x)-f(G(x))||2 represents the distance between the image features of the X-ray image and the reconstructed image of the X-ray image, expressed using the 2-norm.

[0074] Furthermore, the generator's loss function, i.e., the first loss function, can be defined as:

[0075] L G =Loss adv +λ1Loss res (3)

[0076] In equation (3), λ1 is the scaling factor.

[0077] Based on the aforementioned first loss function, the encoder in the current discriminator can be used to extract features from the historical X-ray image and the reconstructed image generated by the current generator to obtain the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator; the distance between the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator can be calculated; and the pixel difference between the historical X-ray image and the reconstructed image generated by the current generator can be calculated; and the first loss function can be determined based on the pixel difference between the historical X-ray image and the reconstructed image generated by the current generator, and the distance between the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator.

[0078] S2. Input the historical X-ray image into the target generator to reconstruct the image, so as to obtain the reconstructed image of the historical X-ray image.

[0079] S3. Use the current discriminator in the generative adversarial network to distinguish between historical X-ray images and reconstructed images of these X-ray images.

[0080] S4. Calculate the discrimination probability of the current discriminator for the reconstructed image of the historical X-ray image.

[0081] The discrimination probability of the current discriminator for the reconstructed images of historical X-ray images can be understood as the proportion of the number of images correctly discriminated by the current discriminator for the reconstructed images of historical X-ray images to the total number of reconstructed images of historical X-ray images.

[0082] S5. Determine whether the discrimination probability is less than or equal to the set probability threshold. If the determination result is greater than (no), proceed to step S6; if the determination result is less than or equal to (yes), proceed to step S9.

[0083] In this embodiment, the specific value of the probability threshold is not limited. Preferably, the probability threshold is less than or equal to 50%. For example, it can be 3%, 5%, 10%, or 15%, etc.

[0084] S6. Using minimizing the second loss function as the training objective, and using historical X-ray images and reconstructed images of historical X-ray images as training samples, train the current discriminator to obtain the target discriminator.

[0085] The second loss function is the loss function of the discriminator, which can be determined based on the actual discrimination performance of the current discriminator on historical X-ray images and reconstructed images of historical X-ray images.

[0086] In this embodiment, the discriminator of the GAN is used to distinguish whether the image input to the discriminator is the original X-ray image or a reconstructed image generated by the generator. Therefore, choosing an appropriate loss function can effectively improve the quality of the reconstructed image. In some embodiments, the second loss function can be the cross-entropy loss function. However, cross-entropy will cause false samples (i.e., the reconstructed image generated by the generator) that are classified as true at the decision boundary (the original X-ray image) but are still far from the real data to stop iterating, because they have successfully deceived the discriminator. Therefore, gradient vanishing occurs when updating the generator.

[0087] To address this issue, in this embodiment, a least squares loss function can be used as the second loss function. The least squares loss function penalizes false samples that are classified as true but are still far from the real data, pulling these false samples back to the decision boundary, thereby improving the quality of the reconstructed image generated by the generator. For example, Figure 1g As shown, the least squares loss function can be expressed as:

[0088]

[0089] In equation (4), D(x) represents the probability that the original X-ray image is identified as a reconstructed image by the discriminator; D(G(x)) represents the probability that the reconstructed X-ray image is identified as a reconstructed image by the discriminator. Ideally, the discriminator in equation (4) has a probability of 0 for the original X-ray image and a probability of 1 for the reconstructed X-ray image.

[0090] Based on the aforementioned second loss function, the least squares loss function for the current discriminator to distinguish the historical X-ray image and the reconstructed image of the historical X-ray image can be calculated according to the actual discrimination of the current discriminator on the historical X-ray image and the reconstructed image of the historical X-ray image, and this loss function can be used as the second loss function.

[0091] S7. Determine whether the current loop count has reached the set number of rounds. If the result is no, proceed to step S8; if the result is yes, proceed to step S10.

[0092] S8. Set the target generator as the current generator, set the target discriminator as the current discriminator, increment the current loop count j by 1; that is, j = j + 1, and return to execute step S1.

[0093] S9. Use the target generator as the image reconstruction model; and use the current discriminator as the trained discriminator.

[0094] S10. Use the target generator as the image reconstruction model and the target discriminator as the trained discriminator.

[0095] The historical X-ray images in steps S1-S10 above are all X-ray images of normal tires, that is, X-ray images used as training samples for the image reconstruction model.

[0096] Optionally, such as Figure 1e As shown, after the image reconstruction model is trained, test historical X-ray images can be obtained from the historical image library to test the discrimination effect of the trained image reconstruction model and thus evaluate the detection effect of the model.

[0097] Furthermore, after the image reconstruction model is trained, it can be applied to the detection of actual tire quality. This involves inputting the X-ray image of the tire to be inspected into the image reconstruction model; and then using the model to reconstruct the input X-ray image to obtain a reconstructed image of the tire's X-ray image.

[0098] Furthermore, the computer device 12 can calculate the similarity between the X-ray image of the tire to be inspected and its reconstructed image; and determine the quality of the tire to be inspected based on the similarity between the X-ray image of the tire to be inspected and its reconstructed image.

[0099] In some embodiments, the computer device 12 can calculate the pixel difference between the X-ray image of the tire to be detected and its reconstructed image as the similarity between the X-ray image of the tire to be detected and its reconstructed image. Optionally, the smaller the pixel difference between the X-ray image of the tire to be detected and its reconstructed image, the greater the similarity between the X-ray image of the tire to be detected and its reconstructed image.

[0100] Optionally, such as Figure 1h As shown, the mean pixel difference between the X-ray image of the tire to be inspected and its reconstructed image can be calculated as the similarity between the X-ray image and the reconstructed image. The calculation formula can be expressed as:

[0101] R(x)=Mean(|xG(x)|) (5)

[0102] In equation (5), x represents the pixel value of the X-ray image, and G(x) represents the pixel value of the reconstructed image of the X-ray image. Mean(|xG(x)|) represents the mean of the pixel difference between the X-ray image and the reconstructed image of the X-ray image.

[0103] In other embodiments, such as Figure 1h As shown, the image features of the X-ray image of the tire to be detected and the image features of the reconstructed image of the X-ray image can be obtained. Optionally, the X-ray image of the tire to be detected and its reconstructed image can be input into the encoder of the trained image reconstruction model, and the image features of the X-ray image of the tire to be detected and the reconstructed image of the X-ray image can be encoded in the encoder to obtain the image features of the X-ray image of the tire to be detected and the image features of the reconstructed image of the X-ray image.

[0104] Furthermore, the distance between the image features of the X-ray image of the tire to be detected and the image features of its reconstructed image is calculated as the similarity between the X-ray image of the tire to be detected and its reconstructed image. The shorter the distance between the image features of the X-ray image of the tire to be detected and the image features of its reconstructed image, the greater the similarity between the X-ray image of the tire to be detected and its reconstructed image.

[0105] Optionally, such as Figure 1h As shown, the mean distance between the image features of the X-ray image of the tire to be detected and its reconstructed image can be calculated as the similarity between the X-ray image of the tire to be detected and its reconstructed image. The corresponding calculation formula can be expressed as:

[0106] D(x)=Mean(|En(x)-En(G(x))|)(6)

[0107] In equation (6), En(x) represents the image features of the X-ray image, and En(G(x)) represents the image features of the reconstructed image of the X-ray image. Mean(|En(x)-En(G(x))|) represents the mean distance between the image features of the X-ray image and the reconstructed image of the X-ray image.

[0108] In practical applications, either one of the two similarity calculation methods can be chosen, or a combination of both can be used. When combining both methods, a weighted calculation can be performed on the similarities obtained from the two methods to obtain a weighted result; and the quality of the tire to be inspected can be determined based on the weighted result. The formula for the weighted calculation can be expressed as:

[0109] S(x)=λ2R(x)+D(x) (7)

[0110] Where λ2 represents the weighting factor.

[0111] Optionally, a difference threshold can be set; and it can be determined whether the weighted calculation result is greater than or equal to the set difference threshold; if the determination result is yes, then it is determined that the tire to be tested has a defect. Correspondingly, if the determination result is no, then it is determined that the tire to be tested is normal, that is, there is no defect.

[0112] However, since different X-ray images have different pixel counts, a separate difference threshold needs to be set for each image with different pixel counts, and it is difficult to select a suitable difference threshold. To solve this problem, in this embodiment, the weighted calculation results can also be normalized to obtain the anomaly score of the X-ray image of the tire to be detected. The normalization formula can be expressed as:

[0113]

[0114] In equation (8), S′(x) represents the anomaly score of the X-ray image of the tire to be inspected; max(S i (x)) and min(S) i (x) represents the maximum and minimum abnormality scores of the X-ray images of normal tires used in the image reconstruction model training, respectively. Where i = 1, 2, ..., N, and N is the total number of X-ray images of normal tires used in the image reconstruction model training.

[0115] Regarding equation (8), the closer the anomaly score of the tire's X-ray image is to 1, the greater the probability that the X-ray image is a defective image. Based on this, the quality of the tire to be inspected can be determined according to the anomaly score of the tire's X-ray image. Optionally, if the anomaly score of the tire's X-ray image is greater than or equal to a set anomaly threshold, the tire to be inspected is determined to have a defect. Correspondingly, if the anomaly score of the tire's X-ray image is less than the set anomaly threshold, the tire to be inspected is determined to be a normal tire, i.e., without defects.

[0116] In this embodiment, the specific value of the abnormal threshold is not limited. Optionally, 1 / 2 < abnormal threshold < 1.

[0117] In this embodiment, after determining the quality inspection result of the tire to be inspected, the computer device 12 can also output the quality inspection result of the tire to be inspected. In some embodiments, the computer device 12 is an electronic device located locally on the X-ray machine 11. The computer device 12 includes a display screen. Accordingly, the computer device 12 can display the quality inspection result of the tire to be inspected through the display component. Optionally, the computer device 12 can also display an X-ray image of the tire to be inspected through the display component.

[0118] In some embodiments, the computer device 12 is a server-side device located in the cloud, in which case the computer device 12 can provide the quality inspection results of the tire to be inspected to the X-ray machine 11. The X-ray machine 11 may be equipped with a display screen. Accordingly, the X-ray machine 11 can display the quality inspection results of the tire to be inspected through a display component. Optionally, the X-ray machine 11 can also display an X-ray image of the tire to be inspected through a display component. Alternatively, the computer device 12 can provide the quality inspection results of the tire to be inspected to a local computer device of the X-ray machine 11, which will then display the quality inspection results of the tire to be inspected; and so on.

[0119] In some embodiments, in addition to detecting tire quality, the computer device 12 can also identify the type of defect in a defective tire. Accordingly, the computer device 12 can also perform defect identification on the X-ray image of the tire to be inspected to determine the type of defect.

[0120] Optionally, the computer device 12 can input the X-ray image of the tire to be inspected into a defect recognition model. In the defect recognition model, feature extraction is performed on the X-ray image of the tire to be inspected to obtain the image features of the X-ray image. Furthermore, defect recognition can be performed based on the image features of the X-ray image to obtain the defect type of the tire to be inspected. Furthermore, the computer device 12 can also output the defect type of the tire to be inspected.

[0121] If the computer device 12 is an electronic device located locally on the X-ray machine 11, the defect type of the tire to be inspected can be displayed via a display component. If the computer device 12 is a server-side device located in the cloud, the defect type of the tire to be inspected can be provided to the X-ray machine 11. Accordingly, the X-ray machine 11 can display the defect type of the tire to be inspected via a display component. Alternatively, the computer device 12 can provide the defect type of the tire to be inspected to a computer device located locally on the X-ray machine 11, which will then display the defect type of the tire to be inspected; and so on.

[0122] The aforementioned defect identification model can be a multi-class classification model, trained using pre-labeled X-ray images of abnormal tires as samples. These pre-labeled X-ray images can include X-ray images of tires with various defect types. Multiple X-ray images are generated for each defect type. Each X-ray image can be pre-labeled with information such as whether a defect exists and its type.

[0123] In addition to the system embodiments described above, this application also provides a detection method applicable to any computer device, including X-ray machines with data processing capabilities. The detection method provided in this application is described below by way of example.

[0124] Figure 2 This is a schematic flowchart illustrating the tire testing method provided in an embodiment of this application. Figure 2 As shown, the method includes:

[0125] 201. Obtain the image to be inspected; the image to be inspected is an X-ray image of the tire to be inspected.

[0126] 202. Use an image reconstruction model to reconstruct the image to be detected, so as to obtain the reconstructed image of the image to be detected; wherein, the image reconstruction model is trained using historical X-ray images of normal tires as samples.

[0127] 203. Calculate the similarity between the image to be detected and the reconstructed image.

[0128] 204. Determine the quality of the tire to be detected based on the similarity between the image to be detected and the reconstructed image.

[0129] In this embodiment, for the device or equipment for inspecting tires, in step 201, an X-ray image of the tire to be inspected can be acquired.

[0130] Optionally, if the computer device performing the detection method is an X-ray machine or a device deployed locally on the X-ray machine, then an optional implementation of step 201 is: acquiring the X-ray image generated by the X-ray machine as the X-ray image of the tire to be detected.

[0131] In other embodiments, the computer device performing the above detection method may be deployed in the cloud, such as a server-side device of a cloud service provider. Optionally, the detection method provided in this application embodiment may also be deployed in the cloud as a SaaS service. For the server-side device that deploys the SaaS service or the detection method provided in this application embodiment, another optional implementation of step 201 is: receiving an X-ray image sent by an X-ray machine or other devices deployed locally on the X-ray machine as the image to be detected.

[0132] Regardless of the execution entity, in step 202, an image reconstruction model trained using historical X-ray images of normal tires can be used to reconstruct the image to be detected, thereby obtaining a reconstructed image of the image to be detected. Optionally, such as Figure 1a As shown, in the image reconstruction model, feature extraction can be performed on the image to be detected to obtain the image features of the image to be detected; and the image features of the image to be detected can be used to reconstruct the image to obtain the reconstructed image of the image to be detected.

[0133] Since the image reconstruction model is trained using historical X-ray images of normal tires as samples, the similarity between the X-ray image and the reconstructed image of a normal tire is high. Conversely, for defective tires, because the image reconstruction model has not learned from X-ray images of defective tires, it cannot reconstruct the defective regions from the X-ray images. Therefore, the difference between the X-ray image and the reconstructed image of a defective tire is large, and the similarity is low.

[0134] Based on the above analysis, in step 203, the similarity between the X-ray image of the tire to be inspected and the reconstructed image of the X-ray image can be calculated; and based on the similarity between the X-ray image of the tire to be inspected and the reconstructed image of the X-ray image, the quality of the tire to be inspected is determined, thus realizing automated tire quality inspection. This inspection process does not rely on human experience, which helps to improve the accuracy of tire quality inspection. For other implementation effects of the inspection method provided in this application embodiment, please refer to the relevant analysis in the above system embodiment, which will not be repeated here.

[0135] In this embodiment of the application, before reconstructing the X-ray image of the tire to be inspected using the image reconstruction model, the image reconstruction model can be trained. The training process of the image reconstruction model can be offline or online. The training process of the image reconstruction model is described below as an example.

[0136] In this embodiment, historical X-ray images of normal tires can be acquired. Optionally, historical X-ray images of normal tires can be acquired from a historical image database. Furthermore, historical X-ray images of normal tires can be used for model training to obtain an image reconstruction model.

[0137] In this embodiment, the implementation form of the image reconstruction model is not limited. Optionally, historical X-ray images of normal tires can be used as training samples to perform generative adversarial training on a generative adversarial network (GAN) until a set stopping condition is met; and the generator in the GAN when the set stopping condition is met is used as the image reconstruction model. For a detailed implementation of the generative adversarial training process, please refer to the above. Figure 1b The relevant descriptions will not be repeated here.

[0138] Once the image reconstruction model is trained, it can be applied to the actual detection of tire quality. This involves inputting the X-ray image of the tire to be inspected into the image reconstruction model; and then using the model to reconstruct the input X-ray image to obtain a reconstructed image of the tire.

[0139] Furthermore, the similarity between the X-ray image of the tire to be inspected (i.e., the image to be inspected) and its reconstructed image can be calculated; and the quality of the tire to be inspected can be determined based on the similarity between the X-ray image of the tire to be inspected and its reconstructed image.

[0140] In some embodiments, the pixel difference between the X-ray image of the tire to be detected and its reconstructed image can be calculated as the similarity between the X-ray image of the tire to be detected and its reconstructed image. Optionally, the smaller the pixel difference between the X-ray image of the tire to be detected and its reconstructed image, the greater the similarity between the X-ray image of the tire to be detected and its reconstructed image.

[0141] Optionally, the mean pixel difference between the X-ray image of the tire to be inspected and its reconstructed image can be calculated as the similarity between the X-ray image and the reconstructed image. The calculation formula can be found in equation (5) above.

[0142] In other embodiments, image features of the X-ray image of the tire to be detected and image features of the reconstructed image of the X-ray image can be obtained. Optionally, the X-ray image of the tire to be detected and its reconstructed image can be input into the encoder of the trained image reconstruction model, whereby the image features of the X-ray image of the tire to be detected and the reconstructed image of the X-ray image are encoded respectively to obtain the image features of the X-ray image of the tire to be detected and the image features of the reconstructed image of the X-ray image.

[0143] Furthermore, the distance between the image features of the X-ray image of the tire to be detected and the image features of its reconstructed image is calculated as the similarity between the X-ray image of the tire to be detected and its reconstructed image. The shorter the distance between the image features of the X-ray image of the tire to be detected and the image features of its reconstructed image, the greater the similarity between the X-ray image of the tire to be detected and its reconstructed image.

[0144] Optionally, the mean distance between the image features of the X-ray image of the tire to be detected and its reconstructed image can be calculated as the similarity between the X-ray image of the tire to be detected and its reconstructed image. The corresponding calculation formula can be found in the above formula (6).

[0145] In practical applications, either one of the two similarity calculation methods can be chosen, or a combination of both can be used. When the two methods are combined, the similarities calculated by the two methods can be weighted to obtain a weighted calculation result; and the quality of the tire to be tested can be determined based on the weighted calculation result. The formula for the weighted calculation can be found in equation (7) above.

[0146] Optionally, a difference threshold can be set; and it can be determined whether the weighted calculation result is greater than or equal to the set difference threshold; if the determination result is yes, then it is determined that the tire to be tested has a defect. Correspondingly, if the determination result is no, then it is determined that the tire to be tested is normal, that is, there is no defect.

[0147] However, since different X-ray images have different pixel counts, a separate difference threshold needs to be set for each X-ray image with different pixel counts, and it is difficult to select a suitable difference threshold. To solve this problem, in this embodiment, the weighted calculation results can also be normalized to obtain the anomaly score of the X-ray image of the tire to be detected. The normalization formula can be found in equation (8) above.

[0148] Regarding equation (8), the closer the anomaly score of the tire's X-ray image is to 1, the greater the probability that the X-ray image is a defective image. Based on this, the quality of the tire to be inspected can be determined according to the anomaly score of the tire's X-ray image. Optionally, if the anomaly score of the tire's X-ray image is greater than or equal to a set anomaly threshold, the tire to be inspected is determined to have a defect. Correspondingly, if the anomaly score of the tire's X-ray image is less than the set anomaly threshold, the tire to be inspected is determined to be a normal tire, i.e., without defects. In this embodiment, the specific value of the anomaly threshold is not limited. Optionally, 1 / 2 < anomaly threshold < 1.

[0149] In this embodiment, after determining the quality inspection result of the tire to be inspected, the quality inspection result of the tire to be inspected can also be output. In some embodiments, if the device performing the above-described inspection method is an electronic device located locally on the X-ray machine, the quality inspection result of the tire to be inspected can be displayed through a display component. Optionally, an X-ray image of the tire to be inspected can also be displayed through the display component.

[0150] In other embodiments, the device performing the above-described detection method is a server-side device located in the cloud. In this case, the quality inspection results of the tire to be inspected can be sent to the provider of the image to be inspected, and the provider of the image to be inspected can output the quality inspection results of the tire to be inspected. The provider of the image to be inspected can be an X-ray machine or other electronic equipment deployed locally on the X-ray machine.

[0151] In some embodiments, in addition to detecting tire quality, for tires with defects, the type of defect can also be identified. Accordingly, defect identification can also be performed on the X-ray image of the tire to be inspected (i.e., the image to be inspected) to determine the type of defect in the tire to be inspected.

[0152] Optionally, the X-ray image of the tire to be inspected can be input into the defect recognition model. In the defect recognition model, feature extraction is performed on the X-ray image of the tire to be inspected to obtain the image features of the X-ray image. Further, defect recognition can be performed based on the image features of the X-ray image to obtain the defect type of the tire to be inspected. Furthermore, the defect type of the tire to be inspected can also be output.

[0153] If the device performing the above detection method is an X-ray machine or an electronic device located locally on the X-ray machine, the defect type of the tire to be detected can be displayed via a display component. If the device performing the above detection method is a server-side device located in the cloud, the defect type of the tire to be detected can be provided to the provider of the image to be detected. Accordingly, the provider of the image to be detected can display the defect type of the tire to be detected via a display component.

[0154] It is worth noting that the detection method provided in this application embodiment can be applied not only to tire inspection but also to the inspection of other objects. For example, it can be used for disease diagnosis using medical imaging, which includes X-rays, CT scans, color Doppler ultrasound, etc., but is not limited to these. Another example is that if the image to be tested is an X-ray image, X-ray images of other industrial parts or products can be used to perform non-destructive testing on the interior of the industrial product. These industrial parts or products can include steel, steel pipes, glass, PVC sheets, ceramics, bearings, automotive parts, vehicle parts, and industrial equipment (such as boilers), but are not limited to these. Yet another example is that if the image to be tested is a visual image, the image of the outer packaging of the object to be tested can be used to detect whether the printing on the outer packaging is qualified. Yet another example is that an image of printed fabric or wallpaper can be used to detect whether the pattern of the printed fabric or wallpaper is qualified, etc. Accordingly, this application embodiment also provides a detection method such as... Figure 3 As shown, the detection method includes:

[0155] 301. Obtain the image to be detected; the image to be detected is the image of the object to be detected.

[0156] 302. Use an image reconstruction model to reconstruct the image to be detected, so as to obtain the reconstructed image of the image to be detected; wherein, the image reconstruction model is trained using historical images of normal objects as samples.

[0157] 303. Calculate the similarity between the image to be detected and the reconstructed image.

[0158] 304. Determine the quality of the object to be detected based on the similarity between the image to be detected and the reconstructed image.

[0159] In this embodiment, the application scenario and the object to be detected differ. In some embodiments, the object to be detected may be human tissue. Accordingly, the image to be detected may be a medical image of human tissue. In other embodiments, the object to be detected may be the outer packaging of an item, such as an outer packaging bag or outer packaging box. Accordingly, the image to be detected is an image of the outer packaging; and so on, but not limited to these. In still other embodiments, the object to be detected is an industrial product, which can be non-destructively inspected using X-rays. Accordingly, the image to be detected is an X-ray image of the industrial product.

[0160] Whether the image to be detected and the historical images are visual images or X-ray images can be determined by the purpose of the detection. For example, in some embodiments, in order to detect the external quality of an object, the image to be detected and the historical images can be visual images; in other embodiments, in order to detect the internal quality of an object, the image to be detected and the historical images can be X-ray images; and so on.

[0161] In this embodiment, an image reconstruction model trained using historical images of normal objects as samples is used to reconstruct the image of the object to be detected, resulting in a reconstructed image of the object to be detected. Since the image reconstruction model is trained using historical images of normal objects as samples, the reconstructed image of a defective object will differ significantly from the original image. Based on this, the similarity between the image to be detected and the reconstructed image can be calculated; and based on the similarity between the two images, the quality of the object to be detected can be determined, achieving automated detection without relying on human experience, thus helping to improve the accuracy of quality detection.

[0162] In some embodiments, the image to be detected can be input into an image reconstruction model; in the image reconstruction model, feature extraction is performed on the image to be detected to obtain image features of the image to be detected; image reconstruction is performed using the image features of the image to be detected to obtain a reconstructed image of the image to be detected. The training process of the image reconstruction model can be found in the relevant content of the above embodiments, and will not be repeated here.

[0163] It is worth noting that in the embodiments of this application, a normal object refers to a defect-free object, that is, an object that meets the quality standards being inspected, and belongs to the same product as the image to be inspected. Accordingly, the historical images of normal objects and the images of the objects to be inspected are images of the same nature. For example, if the object to be inspected is an industrial product, and the purpose of the inspection is to detect whether there is damage or defects inside the industrial product, and the object to be inspected is an X-ray image of the industrial product, then a normal object is a work product that meets the quality inspection standards, which can also be understood as a work product without internal defects; accordingly, the historical images of normal objects are X-ray images of normal objects.

[0164] For details on the training process of the reconstructed image model and the processing of the detection results, please refer to the relevant content in the above implementation method for tire detection, which will not be repeated here.

[0165] In some embodiments, the similarity between the image to be detected and its reconstructed image can be calculated, including at least one of the following calculation methods:

[0166] The pixel difference between the image to be detected and its reconstructed image is calculated as the similarity between the image to be detected and its reconstructed image.

[0167] And / or,

[0168] Obtain the image features of the image to be detected and the image features of the reconstructed image of the image to be detected; calculate the distance between the image features of the image to be detected and the image features of the reconstructed image, which is used as the similarity between the image to be detected and the reconstructed image of the image to be detected.

[0169] Furthermore, the similarity calculated by at least one method can be weighted to obtain a weighted calculation result; and the quality of the object to be detected can be determined based on the weighted calculation result.

[0170] Optionally, the weighted calculation result is normalized to obtain the anomaly score of the image to be detected; if the anomaly score of the image to be detected is greater than or equal to the set anomaly threshold, it is determined that the object to be detected has a defect.

[0171] It should be noted that the execution subject of each step of the method provided in the above embodiments can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 201 and 202 can be device A; or the execution subject of step 201 can be device A, and the execution subject of step 202 can be device B; and so on.

[0172] Furthermore, some processes described in the above embodiments and accompanying drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or they may be executed in parallel. The operation numbers, such as 201, 202, etc., are merely used to distinguish different operations and do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel.

[0173] Accordingly, embodiments of this application also provide a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, cause one or more processors to perform the steps in the above-described tire detection method and / or detection method.

[0174] This application also provides a computer program product, including: a computer program; the computer program is executed by a processor to implement the above-described tire detection method and / or detection method.

[0175] Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of this application. For example... Figure 4 As shown, the computer device includes a memory 40a and a processor 40b. The memory 40a is used to store computer programs.

[0176] The processor 40b is coupled to the memory 40a for executing a computer program to: acquire an image to be detected; the image to be detected is an X-ray image of the tire to be detected; reconstruct the image to be detected using an image reconstruction model to obtain a reconstructed image of the image to be detected; the image reconstruction model is trained using historical X-ray images of normal tires as samples; calculate the similarity between the image to be detected and the reconstructed image; and determine the quality of the tire to be detected based on the similarity between the image to be detected and the reconstructed image.

[0177] Optionally, when the processor 40b uses the image reconstruction model to reconstruct the image to be detected, it specifically performs the following: inputting the image to be detected into the image reconstruction model; extracting features from the image to be detected in the image reconstruction model to obtain image features of the image to be detected; and using the image features of the image to be detected to reconstruct the image to obtain a reconstructed image of the image to be detected.

[0178] Optionally, before using the image reconstruction model to reconstruct the image to be detected, the processor 40b is also used to: train the generative adversarial network using historical X-ray images of normal tires as training samples until a set stopping condition is met; and use the generator in the generative adversarial network when the set stopping condition is met as the image reconstruction model.

[0179] Optionally, when the processor 40b performs generative adversarial training on the generator and discriminator, it specifically performs the following: training the current generator with the first loss function as the training objective and historical X-ray images as training samples to obtain a target generator; the first loss function is determined based on the historical X-ray images and the reconstructed images generated by the current generator; inputting the historical X-ray images into the target generator for image reconstruction to obtain a reconstructed image of the historical X-ray images; using the current discriminator in the generative adversarial network to distinguish between the historical X-ray images and the reconstructed images of the historical X-ray images; and calculating the discrimination probability of the current discriminator for the reconstructed images of the historical X-ray images; and determining whether the discrimination probability is... If the probability is less than or equal to a set probability threshold, or greater than a set probability threshold, the current discriminator is trained using the second loss function as the training objective and historical X-ray images and their reconstructed images as training samples to obtain a target discriminator. The second loss function is determined based on the actual discrimination performance of the current discriminator on historical X-ray images and their reconstructed images. The system also determines whether the current iteration count has reached a set number of rounds. If not, the target generator is used as the current generator, and the target discriminator is used as the current discriminator. The system then returns to the previous operation of training the current generator until the discrimination probability is less than or equal to the set probability threshold, or the iteration count reaches the set number of rounds.

[0180] Optionally, the processor 40b is further configured to: extract features from historical X-ray images and reconstructed images generated by the current generator using the encoder in the current discriminator during the training of the current generator, so as to obtain image features of historical X-ray images and image features of reconstructed images generated by the current generator; calculate the distance between image features of historical X-ray images and image features of reconstructed images generated by the current generator; calculate the pixel difference between historical X-ray images and reconstructed images generated by the current generator; and determine a first loss function based on the pixel difference between historical X-ray images and reconstructed images generated by the current generator, and the distance between image features of historical X-ray images and image features of reconstructed images generated by the current generator.

[0181] Optionally, the processor 40b is further configured to: during the training of the current generator, calculate the least squares loss function of the current discriminator in distinguishing the historical X-ray image and the reconstructed image of the historical X-ray image, based on the actual discrimination of the current discriminator in distinguishing the historical X-ray image and the reconstructed image of the historical X-ray image, as a second loss function.

[0182] In some embodiments, the processor 40b calculates the similarity between the image to be detected and its reconstructed image by performing at least one of the following calculation methods: calculating the pixel difference between the image to be detected and its reconstructed image as the similarity between the image to be detected and its reconstructed image; and / or, obtaining the image features of the image to be detected and the image features of the reconstructed image; and calculating the distance between the image features of the image to be detected and the image features of its reconstructed image as the similarity between the image to be detected and its reconstructed image.

[0183] Accordingly, when determining the quality of the tire to be inspected, the processor 40b is specifically used to: perform a weighted calculation on the similarity calculated by at least one calculation method to obtain a weighted calculation result; and determine the quality of the tire to be inspected based on the weighted calculation result.

[0184] Furthermore, when determining the quality of the tire to be inspected, the processor 40b specifically performs the following: normalizes the weighted calculation results to obtain the anomaly score of the X-ray image; and determines the quality of the tire to be inspected based on the anomaly score of the image to be inspected.

[0185] Furthermore, when determining the quality of the tire to be inspected, the processor 40b specifically determines that the tire to be inspected has a defect if the abnormality score of the image to be inspected is greater than or equal to a set abnormality threshold.

[0186] In some embodiments, the processor 40b is further configured to: display the quality inspection results for the tire to be inspected via the display component 40c; or, send the quality inspection results for the tire to be inspected to a provider of the image to be inspected via the communication component 40d, so that the provider can output the quality inspection results.

[0187] In other embodiments, the processor 40b is further configured to: if the tire to be inspected has a defect, perform defect identification on the image to be inspected to determine the type of defect in the tire to be inspected.

[0188] Optionally, the processor 40b is also configured to: display the defect type of the tire to be inspected via the display component 40c; or, send the defect type of the tire to be inspected to the provider of the X-ray image via the communication component 40c, so that the provider can output the defect type of the tire to be inspected.

[0189] The computer equipment provided in this embodiment can use an image reconstruction model trained with historical X-ray images of normal tires as samples to reconstruct the X-ray image of the tire to be inspected, obtaining a reconstructed image of the X-ray image. Since the image reconstruction model is trained with historical X-ray images of normal tires as samples, the reconstructed image and the X-ray image of a defective tire will differ significantly. Based on this, the similarity between the X-ray image and the reconstructed image of the tire to be inspected can be calculated; and based on the similarity between the X-ray image and the reconstructed image, the quality of the tire to be inspected can be determined, realizing automated tire quality inspection without relying on human experience, thus helping to improve the accuracy of tire quality inspection.

[0190] In this embodiment of the application, the processor 40b is further configured to: acquire an image to be detected; the image to be detected is an image of an object to be detected; use an image reconstruction model to reconstruct the image to be detected to obtain a reconstructed image of the image to be detected; wherein the image reconstruction model is trained on historical images of normal objects; calculate the similarity between the image to be detected and the reconstructed image; and determine the quality of the object to be detected based on the similarity between the image to be detected and the reconstructed image.

[0191] Optionally, the image to be detected is an X-ray image of the object to be detected; the historical image is a historical X-ray image of a normal object.

[0192] In some embodiments, when the processor 40b performs image reconstruction on the image to be detected using an image reconstruction model, it specifically performs the following: inputting the image to be detected into the image reconstruction model; extracting features from the image to be detected in the image reconstruction model to obtain image features of the image to be detected; and performing image reconstruction using the image features of the image to be detected to obtain a reconstructed image of the image to be detected.

[0193] Optionally, when calculating the similarity between the image to be detected and the reconstructed image of the image to be detected, the processor 40b specifically performs at least one of the following calculation methods:

[0194] The pixel difference between the image to be detected and its reconstructed image is calculated as the similarity between the image to be detected and its reconstructed image.

[0195] Obtain the image features of the image to be detected and the image features of the reconstructed image of the image to be detected; calculate the distance between the image features of the image to be detected and the image features of the reconstructed image, which is used as the similarity between the image to be detected and the reconstructed image of the image to be detected.

[0196] Accordingly, when determining the quality of the tire to be inspected, the processor 40b is specifically used to: perform a weighted calculation on the similarity calculated by at least one calculation method to obtain a weighted calculation result; and determine the quality of the object to be inspected based on the weighted calculation result.

[0197] Optionally, when determining the quality of the object to be detected, the processor 40b specifically performs the following: normalizes the weighted calculation result to obtain the anomaly score of the image to be detected; if the anomaly score of the image to be detected is greater than or equal to a set anomaly threshold, then it is determined that the object to be detected has a defect.

[0198] The computer device provided in this embodiment can use an image reconstruction model trained on historical images of normal objects as samples to reconstruct the image of the object to be detected, obtaining a reconstructed image of the object to be detected. Since the image reconstruction model is trained on historical images of normal objects as samples, the reconstructed image of a defective object will differ significantly from the original image. Based on this, the similarity between the image to be detected and the reconstructed image can be calculated; and based on the similarity between the image to be detected and the reconstructed image, the quality of the object to be detected can be determined, achieving automated detection without relying on human experience, thus helping to improve the accuracy of quality detection.

[0199] In some alternative implementations, such as Figure 4 As shown, the computer device may also include components such as a power supply component 40e and an audio component 40f. Figure 4 The diagram only shows some components and does not mean that the computer device must contain them. Figure 4 The inclusion of all components does not imply that a computer device can only include... Figure 4 The components shown.

[0200] In this embodiment, the memory is used to store computer programs and can be configured to store various other data to support operation on its host device. The processor can execute the computer programs stored in the memory to implement corresponding control logic. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0201] In the embodiments of this application, the processor can be any hardware processing device capable of executing the above-described method logic. Optionally, the processor can be a central processing unit (CPU), a graphics processing unit (GPU), or a microcontroller unit (MCU); it can also be a field-programmable gate array (FPGA), a programmable array logic (PAL), a general array logic (GAL), a complex programmable logic device (CPLD), or other programmable devices; or it can be an advanced reduced instruction set (RISC) processor (ARM) or a system on chip (SOC), etc., but is not limited thereto.

[0202] In this embodiment, the communication component is configured to facilitate wired or wireless communication between its host device and other devices. The device housing the communication component can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In another exemplary embodiment, the communication component may also be implemented based on Near Field Communication (NFC), Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), Bluetooth (BT), or other technologies.

[0203] In embodiments of this application, the display component may include a liquid crystal display (LCD) and a touch panel (TP). If the display component includes a touch panel, the display component can be implemented as a touchscreen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation.

[0204] In this embodiment, a power supply component is configured to provide power to various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which the power supply component resides.

[0205] In embodiments of this application, the audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), which is configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals. For example, in devices with voice interaction capabilities, voice interaction with the user can be achieved through the audio component.

[0206] Figure 5 This is a structural block diagram of an X-ray machine provided in an embodiment of this application. Figure 5 As shown, the X-ray machine 50 includes: a mechanical body 51; the mechanical body is provided with a receiving cavity 52; and an X-ray emitter 53 and an X-ray detector 54 are provided in the receiving cavity.

[0207] The mechanical body 51 is also equipped with a signal acquisition unit 55 and a processing unit 56; the signal acquisition unit 55 is electrically connected between the X-ray detector 54 and the processing unit 56.

[0208] In this embodiment, the signal acquisition unit 55 and the processing unit 56 can be integrated on the same PCB board or on different PCB boards. The processing unit 56 may include a processor and its peripheral circuitry. The implementation details of the processor can be found in the relevant content of the above embodiments and will not be repeated here.

[0209] In this embodiment, the receiving cavity 52 is used to hold the tire to be inspected. The X-ray emitter 53 is used to emit X-rays outward; the X-rays can penetrate the tire to be inspected.

[0210] X-ray detector 54 is used to receive radiation signals generated by X-rays penetrating the tire to be inspected.

[0211] The signal acquisition unit 55 is used to acquire radiation signals and convert the radiation signals into X-ray images; and to provide the X-ray images to the processing unit 56.

[0212] Optionally, such as Figure 5 As shown, the signal acquisition unit 55 may include a photoelectric converter 55a and an analog-to-digital converter 55b. The photoelectric converter 55a can convert radiation signals into electrical signals. The electrical signals are analog signals. The analog-to-digital converter 55b can convert electrical signals into data signals, and the matrix formed by these digital signals can constitute a digital image, i.e., an X-ray image.

[0213] Processing unit 56 is used to reconstruct the X-ray image using an image reconstruction model to obtain a reconstructed image of the X-ray image; wherein, the image reconstruction model is trained using historical X-ray images of normal tires as samples; further, the similarity between the X-ray image and the reconstructed image is calculated; and the quality of the tire to be detected is determined based on the similarity between the X-ray image and the reconstructed image.

[0214] Optionally, when processing unit 56 performs image reconstruction on the image to be detected using the image reconstruction model, it specifically performs the following: inputting the image to be detected into the image reconstruction model; extracting features from the image to be detected in the image reconstruction model to obtain image features of the image to be detected; and using the image features of the image to be detected to perform image reconstruction to obtain a reconstructed image of the image to be detected.

[0215] Optionally, before using the image reconstruction model to reconstruct the image to be detected, the processing unit 56 is further configured to: use historical X-ray images of normal tires as training samples to perform generative adversarial training on the generative adversarial network until a set stopping condition is met; and use the generator in the generative adversarial network when the set stopping condition is met as the image reconstruction model.

[0216] Optionally, when performing generative adversarial training on the generator and discriminator, the processing unit 56 specifically performs the following: training the current generator with minimizing the first loss function as the training objective and using historical X-ray images as training samples to obtain a target generator; the first loss function is determined based on the historical X-ray images and the reconstructed images generated by the current generator; inputting the historical X-ray images into the target generator for image reconstruction to obtain a reconstructed image of the historical X-ray images; using the current discriminator in the generative adversarial network to distinguish between the historical X-ray images and the reconstructed images of the historical X-ray images; and calculating the discrimination probability of the current discriminator for the reconstructed images of the historical X-ray images; and determining whether the discrimination probability is... If the probability is less than or equal to a set probability threshold, or greater than a set probability threshold, the current discriminator is trained using the second loss function as the training objective and historical X-ray images and their reconstructed images as training samples to obtain a target discriminator. The second loss function is determined based on the actual discrimination performance of the current discriminator on historical X-ray images and their reconstructed images. The system also determines whether the current iteration count has reached a set number of rounds. If not, the target generator is used as the current generator, and the target discriminator is used as the current discriminator. The system then returns to the previous operation of training the current generator until the discrimination probability is less than or equal to the set probability threshold, or the iteration count reaches the set number of rounds.

[0217] Optionally, the processing unit 56 is further configured to: extract features from historical X-ray images and reconstructed images generated by the current generator using the encoder in the current discriminator during the training process of the current generator, so as to obtain image features of historical X-ray images and image features of reconstructed images generated by the current generator; calculate the distance between image features of historical X-ray images and image features of reconstructed images generated by the current generator; calculate the pixel difference between historical X-ray images and reconstructed images generated by the current generator; and determine a first loss function based on the pixel difference between historical X-ray images and reconstructed images generated by the current generator, and the distance between image features of historical X-ray images and image features of reconstructed images generated by the current generator.

[0218] Optionally, the processing unit 56 is further configured to: during the training of the current generator, calculate the least squares loss function of the current discriminator in distinguishing the historical X-ray image and the reconstructed image of the historical X-ray image, based on the actual discrimination of the current discriminator in distinguishing the historical X-ray image and the reconstructed image of the historical X-ray image, as the second loss function.

[0219] In some embodiments, the processing unit 56 calculates the similarity between the image to be detected and its reconstructed image by performing at least one of the following calculation methods: calculating the pixel difference between the image to be detected and its reconstructed image as the similarity between the image to be detected and its reconstructed image; and / or, obtaining the image features of the image to be detected and the image features of the reconstructed image; and calculating the distance between the image features of the image to be detected and the image features of its reconstructed image as the similarity between the image to be detected and its reconstructed image.

[0220] Accordingly, when determining the quality of the tire to be tested, the processing unit 56 is specifically used to: perform a weighted calculation on the similarity calculated by at least one calculation method to obtain a weighted calculation result; and determine the quality of the tire to be tested based on the weighted calculation result.

[0221] Furthermore, when determining the quality of the tire to be inspected, the processing unit 56 specifically performs the following: normalizes the weighted calculation results to obtain the anomaly score of the X-ray image; and determines the quality of the tire to be inspected based on the anomaly score of the image to be inspected.

[0222] Furthermore, when determining the quality of the tire to be inspected, the processing unit 56 is specifically used to: determine that the tire to be inspected has a defect if the abnormal score of the image to be inspected is greater than or equal to the set abnormal threshold.

[0223] In some embodiments, the processing unit 56 is further configured to: display the quality inspection results for the tire to be inspected via the display component 57.

[0224] In other embodiments, the processing unit 56 is further configured to: if the tire to be inspected has a defect, perform defect identification on the image to be inspected to determine the type of defect in the tire to be inspected.

[0225] Optionally, the processing unit 56 is also configured to: display the type of defect in the tire to be inspected via the display component 57.

[0226] In some alternative implementations, such as Figure 5 As shown, the X-ray machine may also include components such as a power supply assembly 58 and an audio assembly 59. Figure 5 The diagram only shows some components and does not mean that the computer device must contain them. Figure 5 The inclusion of all components does not imply that a computer device can only include... Figure 5 The components shown.

[0227] The X-ray machine provided in this embodiment can use an image reconstruction model trained with historical X-ray images of normal tires as samples to reconstruct the X-ray image of the tire to be inspected, obtaining a reconstructed image. Since the image reconstruction model is trained with historical X-ray images of normal tires as samples, the reconstructed image and the X-ray image of a defective tire will differ significantly. Based on this, the similarity between the X-ray image and the reconstructed image of the tire to be inspected can be calculated; and based on the similarity between the X-ray image and the reconstructed image, the quality of the tire to be inspected can be determined, achieving automated tire quality inspection without relying on human experience, thus helping to improve the accuracy of tire quality inspection.

[0228] It should be noted that the terms "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0229] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0230] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0231] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0232] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0233] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0234] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0235] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0236] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0237] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A tire inspection method, characterized in that, include: Acquire the image to be detected; the image to be detected is an X-ray image of the tire to be detected; The image to be detected is reconstructed using an image reconstruction model to obtain a reconstructed image of the image to be detected; the image reconstruction model is trained using historical X-ray images of normal tires as samples. Calculate the similarity between the image to be detected and the reconstructed image; The quality of the tire to be detected is determined based on the similarity between the image to be detected and the reconstructed image; The image reconstruction model is the generator used when the generative adversarial network stops training; the generator used when training stops is trained using the following method: The current generator is trained with the goal of minimizing the first loss function and the historical X-ray images as training samples to obtain the target generator; the first loss function is determined based on the historical X-ray images and the reconstructed images generated by the current generator. The historical X-ray image is input into the target generator for image reconstruction to obtain the reconstructed image of the historical X-ray image; The current discriminator in the generative adversarial network is used to distinguish between the historical X-ray image and the reconstructed image of the historical X-ray image; Calculate the discrimination probability of the reconstructed image of the historical X-ray image by the current discriminator; If the discrimination probability is greater than the set probability threshold, then the current discriminator is trained with the goal of minimizing the second loss function and with the historical X-ray images and their reconstructed images as training samples to obtain the target discriminator; the second loss function is determined based on the actual discrimination performance of the current discriminator on the historical X-ray images and their reconstructed images. If the current loop count has not reached the set number of rounds, then the target generator is used as the current generator; and the target discriminator is used as the current discriminator, and the operation of training the current generator is returned until the discrimination probability is less than or equal to the set probability threshold, or the loop count reaches the set number of rounds.

2. The method according to claim 1, characterized in that, The process of reconstructing the image to be detected using an image reconstruction model includes: The image to be detected is input into the image reconstruction model; In the image reconstruction model, feature extraction is performed on the image to be detected to obtain the image features of the image to be detected; Image reconstruction is performed using the image features of the image to be detected to obtain the reconstructed image of the image to be detected.

3. The method according to claim 1, characterized in that, The training process for the current generator also includes: The encoder in the current discriminator is used to extract features from the historical X-ray image and the reconstructed image generated by the current generator, so as to obtain the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator; Calculate the distance between the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator; Calculate the pixel differences between the historical X-ray image and the reconstructed image generated by the current generator; The first loss function is determined based on the pixel differences between the historical X-ray image and the reconstructed image generated by the current generator, and the distance between the image features of the historical X-ray image and the image features of the reconstructed image generated by the current generator.

4. The method according to claim 1, characterized in that, The training process for the current generator also includes: Based on the actual discrimination results of the current discriminator on the historical X-ray image and the reconstructed image of the historical X-ray image, the least squares loss function of the current discriminator in discriminating the historical X-ray image and the reconstructed image of the historical X-ray image is calculated as the second loss function.

5. The method according to claim 1, characterized in that, The calculation of the similarity between the image to be detected and the reconstructed image of the image to be detected includes at least one of the following calculation methods: The pixel difference between the image to be detected and its reconstructed image is calculated as the similarity between the image to be detected and its reconstructed image. Obtain the image features of the image to be detected and the image features of the reconstructed image of the image to be detected; calculate the distance between the image features of the image to be detected and the image features of the reconstructed image, which is used as the similarity between the image to be detected and the reconstructed image of the image to be detected.

6. The method according to claim 5, characterized in that, Determining the quality of the tire to be detected based on the similarity between the image to be detected and its reconstructed image includes: The similarity calculated by at least one of the above calculation methods is weighted to obtain a weighted calculation result; The mass of the tire to be tested is determined based on the weighted calculation results.

7. The method according to claim 6, characterized in that, Determining the mass of the tire to be tested based on the weighted calculation result includes: The weighted calculation results are normalized to obtain the anomaly score of the image to be detected; The quality of the tire to be inspected is determined based on the anomaly score of the image to be inspected.

8. The method according to claim 7, characterized in that, Determining the quality of the tire to be inspected based on the anomaly score of the image to be inspected includes: If the anomaly score of the image to be detected is greater than or equal to the set anomaly threshold, then the tire to be detected is determined to have a defect.

9. The method according to any one of claims 1-8, characterized in that, Also includes: The quality inspection results for the tire under test are displayed via a display component. or, The quality inspection results for the tire to be inspected are sent to the provider of the image to be inspected, so that the provider can output the quality inspection results.

10. The method according to any one of claims 1-8, characterized in that, Also includes: If the tire to be inspected has a defect, then the image to be inspected is used for defect identification to determine the type of defect in the tire to be inspected.

11. The method according to claim 10, characterized in that, Also includes: The type of defect in the tire to be inspected is displayed by a display component; or, The defect type of the tire to be inspected is sent to the provider of the X-ray image so that the provider can output the defect type of the tire to be inspected.

12. A detection method, characterized in that, include: Acquire the image to be detected; the image to be detected is an image of the object to be detected. The image to be detected is reconstructed using an image reconstruction model to obtain a reconstructed image of the image to be detected; wherein, the image reconstruction model is trained using historical images of normal objects; Calculate the similarity between the image to be detected and the reconstructed image; The quality of the object to be detected is determined based on the similarity between the image to be detected and the reconstructed image; The image reconstruction model is the generator used when the generative adversarial network stops training; the generator used when training stops is trained using the following method: The current generator is trained with the first loss function as the training objective and the historical images as training samples to obtain the target generator; the first loss function is determined based on the historical images and the reconstructed images generated by the current generator. The historical image is input into the target generator for image reconstruction to obtain the reconstructed image of the historical image; The current discriminator in the generative adversarial network is used to distinguish between the historical image and the reconstructed image of the historical image; Calculate the discrimination probability of the reconstructed image of the historical image by the current discriminator; If the discrimination probability is greater than the set probability threshold, then the current discriminator is trained with the goal of minimizing the second loss function and with the historical images and their reconstructed images as training samples to obtain the target discriminator; the second loss function is determined based on the actual discrimination performance of the current discriminator on the historical images and their reconstructed images. If the current loop count has not reached the set number of rounds, then the target generator is used as the current generator; and the target discriminator is used as the current discriminator, and the operation of training the current generator is returned until the discrimination probability is less than or equal to the set probability threshold, or the loop count reaches the set number of rounds.

13. The method according to claim 12, characterized in that, The step of using an image reconstruction model to reconstruct the image to be detected, in order to obtain a reconstructed image of the image to be detected, includes: The image to be detected is input into the image reconstruction model; In the image reconstruction model, feature extraction is performed on the image to be detected to obtain the image features of the image to be detected; Image reconstruction is performed using the image features of the image to be detected to obtain the reconstructed image of the image to be detected.

14. The method according to claim 13, characterized in that, The calculation of the similarity between the image to be detected and the reconstructed image of the image to be detected includes at least one of the following calculation methods: The pixel difference between the image to be detected and its reconstructed image is calculated as the similarity between the image to be detected and its reconstructed image. Obtain the image features of the image to be detected and the image features of the reconstructed image of the image to be detected; calculate the distance between the image features of the image to be detected and the image features of the reconstructed image, which is used as the similarity between the image to be detected and the reconstructed image of the image to be detected.

15. The method according to claim 14, characterized in that, Determining the quality of the object to be detected based on the similarity between the image to be detected and its reconstructed image includes: The similarity calculated by at least one of the above calculation methods is weighted to obtain a weighted calculation result; The quality of the object to be tested is determined based on the weighted calculation results.

16. The method according to claim 15, characterized in that, Determining the quality of the object to be detected based on the weighted calculation result includes: The weighted calculation results are normalized to obtain the anomaly score of the image to be detected; If the anomaly score of the image to be detected is greater than or equal to the set anomaly threshold, then the object to be detected is determined to have a defect.

17. The method according to any one of claims 12-16, characterized in that, The image to be detected is an X-ray image of the object to be detected; the historical image is a historical X-ray image of a normal object.

18. A tire inspection system, characterized in that, include: X-ray machines and computer equipment; The X-ray machine is used to emit X-rays onto the tire to be inspected; And receive the radiation signal generated by the X-rays penetrating the tire to be tested; The radiation signal is converted into an X-ray image, and the X-ray image is provided to the computer device. The computer device is used to reconstruct the X-ray image using an image reconstruction model to obtain a reconstructed image of the X-ray image; and to calculate the similarity between the X-ray image and the reconstructed image. The mass of the tire to be inspected is determined based on the similarity between the X-ray image and the reconstructed image. The image reconstruction model is trained using historical X-ray images of normal tires as samples; the image reconstruction model is a generator used when the generative adversarial network stops training; the generator used when training stops is trained using the following method: The current generator is trained with the goal of minimizing the first loss function and the historical X-ray images as training samples to obtain the target generator; the first loss function is determined based on the historical X-ray images and the reconstructed images generated by the current generator. The historical X-ray image is input into the target generator for image reconstruction to obtain the reconstructed image of the historical X-ray image; The current discriminator in the generative adversarial network is used to distinguish between the historical X-ray image and the reconstructed image of the historical X-ray image; Calculate the discrimination probability of the reconstructed image of the historical X-ray image by the current discriminator; If the discrimination probability is greater than the set probability threshold, then the current discriminator is trained with the goal of minimizing the second loss function and with the historical X-ray images and their reconstructed images as training samples to obtain the target discriminator; the second loss function is determined based on the actual discrimination performance of the current discriminator on the historical X-ray images and their reconstructed images. If the current loop count has not reached the set number of rounds, then the target generator is used as the current generator; and the target discriminator is used as the current discriminator, and the operation of training the current generator is returned until the discrimination probability is less than or equal to the set probability threshold, or the loop count reaches the set number of rounds.

19. The system according to claim 18, characterized in that, The computer equipment is deployed in the cloud or locally on the X-ray machine.

20. An X-ray machine, characterized in that, include: A mechanical body; the mechanical body is provided with a receiving cavity; an X-ray emitter and an X-ray detector are disposed in the receiving cavity; The mechanical body is also equipped with a signal acquisition unit and a processing unit; the signal acquisition unit is electrically connected between the X-ray detector and the processing unit. The cavity is used to hold the tire to be inspected; the X-ray emitter is used to emit X-rays outward; the X-rays can penetrate the tire to be inspected. The X-ray detector is used to receive the radiation signal generated by the X-rays penetrating the tire to be tested; The signal acquisition unit is used to acquire the radiation signal, convert the radiation signal into an X-ray image, and provide the X-ray image to the processing unit. The processing unit is configured to reconstruct the X-ray image using an image reconstruction model to obtain a reconstructed image of the X-ray image; calculate the similarity between the X-ray image and the reconstructed image; and determine the quality of the tire to be inspected based on the similarity between the X-ray image and the reconstructed image. The image reconstruction model is trained using historical X-ray images of normal tires as samples; the image reconstruction model is a generator used when the generative adversarial network stops training; the generator used when training stops is trained using the following method: The current generator is trained with the goal of minimizing the first loss function and the historical X-ray images as training samples to obtain the target generator; the first loss function is determined based on the historical X-ray images and the reconstructed images generated by the current generator. The historical X-ray image is input into the target generator for image reconstruction to obtain the reconstructed image of the historical X-ray image; The current discriminator in the generative adversarial network is used to distinguish between the historical X-ray image and the reconstructed image of the historical X-ray image; Calculate the discrimination probability of the reconstructed image of the historical X-ray image by the current discriminator; If the discrimination probability is greater than the set probability threshold, then the current discriminator is trained with the goal of minimizing the second loss function and with the historical X-ray images and their reconstructed images as training samples to obtain the target discriminator; the second loss function is determined based on the actual discrimination performance of the current discriminator on the historical X-ray images and their reconstructed images. If the current loop count has not reached the set number of rounds, then the target generator is used as the current generator; and the target discriminator is used as the current discriminator, and the operation of training the current generator is returned until the discrimination probability is less than or equal to the set probability threshold, or the loop count reaches the set number of rounds.

21. A computer device, characterized in that, include: A memory and a processor; wherein the memory is used to store computer programs; The processor is coupled to the memory for executing the computer program for the steps in the method of any one of claims 1-17.

22. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed by one or more processors, the one or more processors are caused to perform the steps of the method according to any one of claims 1-17.

23. A computer program product, characterized in that, include: A computer program; said computer program is executed by a processor to implement the method of any one of claims 1-17.

Images