Strain clamp undervoltage defect detection method based on X-ray image recognition and related device

Abstract

The invention provides a strain clamp undervoltage defect detection method based on X-ray image recognition and a related device, and the method comprises the steps: carrying out the optimization processing of an original X-ray image of a strain clamp, and obtaining an optimized ray image; inputting the optimized ray image into a deep learning model, and extracting a fine structure feature vector related to the under-voltage defect of the strain clamp in the image by using the deep learning model; inputting the extracted fine structure feature vector into a target detection deep learning model, identifying the undervoltage defect of the original X-ray image, and rating the identified defect in combination with a preset standard; and the target detection deep learning model adopts a one-stage or two-stage architecture. According to the method, the X-ray image is optimized, and the target detection deep learning model is adopted for defect detection, so that the accuracy and reliability of under-voltage defect detection of the strain clamp can be remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of strain clamp detection, and particularly relates to a method and related device for detecting underpressure defects of strain clamps based on X-ray image recognition. Background Art

[0002] The strain clamp is an important fitting in high-voltage transmission lines. Its main function is to fix the conductor and bear the tension, enabling the conductor to be firmly suspended on the strain string group or the tower, while ensuring the normal transmission of current. The crimping quality inside the strain clamp directly affects the construction quality and service life of the transmission line. If the quality does not meet the standard, in severe cases, it may trigger major accidents, causing huge economic losses and casualties. Therefore, to ensure the stability and efficient operation of the transmission line and reduce the accident risk, it is particularly important to detect the crimping quality of the strain clamp.

[0003] Compared with traditional detection methods, the X-ray non-destructive testing technology provides a more intuitive detection means, which can effectively show the internal structure of the strain clamp and reveal hidden quality problems. However, due to various factors such as detection equipment, technology, and methods, X-ray images often have problems such as insufficient brightness, unclear contrast, blurred gray-scale changes, and noise interference. These defects directly affect the evaluation accuracy of the crimping quality and reduce the reliability of the detection results.

[0004] Currently, although image recognition technology based on deep learning is widely used in defect detection, due to the quality problems of existing X-ray images, many difficulties are faced in extracting key features, resulting in more refined requirements for the image processing process. The instability of image quality may even affect the accuracy of the final detection results and cannot meet the high-precision requirements in current practical applications. Summary of the Invention

[0005] In view of this, the present invention provides a method and related device for detecting underpressure defects of strain clamps based on X-ray image recognition, which are used to solve the technical problem that the existing technology cannot overcome the quality defects of X-ray images, ensure the accuracy of image defect recognition results, and thus cannot meet the actual application requirements.

[0006] To achieve the above object, the technical solutions provided by the present invention are as follows:

[0007] In the first aspect, the present invention provides a method for detecting underpressure defects of strain clamps based on X-ray image recognition, including the following steps:

[0008] Optimize the original X-ray image of the strain clamp to obtain an optimized ray image;

[0009] Input the optimized ray image into a deep learning model, and use the deep learning model to extract the subtle structural feature vectors related to the under-voltage defect of the strain clamp in the image;

[0010] Input the extracted subtle structural feature vectors into a target detection deep learning model to identify the under-voltage defects in the original X-ray image, and rate the identified defects in combination with a preset standard; the target detection deep learning model adopts a one-stage or two-stage architecture.

[0011] Further, optimize the original X-ray image of the strain clamp, including:

[0012] Use the Gaussian filtering algorithm to denoise the original X-ray image of the strain clamp;

[0013] Use the histogram equalization method to grayscale the denoised image;

[0014] Non-linearly adjust the grayscaled image, and adjust the brightness and contrast of different gray levels in the image through a non-linear function.

[0015] Further, use the Gaussian filtering algorithm to denoise the original X-ray image of the strain clamp, including:

[0016] For the pixel values of all pixel points within the image window, calculate the corresponding weight values, and the weight calculation formula is as follows:

[0017]

[0018] In the formula, represents the weight value of pixel point ; represents the standard deviation; and represent the coordinates of the center point of the window;

[0019] Multiply the original pixel values of all pixel points within the image window by the corresponding weight values to obtain the denoised ray image.

[0020] Further, use the histogram equalization method to grayscale the denoised image, including:

[0021] Count the number of pixels at each gray level (the range of gray levels is [0, l-1]) in the image, and record it as the original gray distribution of the image;

[0022] Calculate the histogram distribution probability of each gray level, as follows:

[0023]

[0024] Wherein, represents the histogram distribution probability of the th gray level; represents the number of pixels of the th gray level; represents the total number of pixels in the image;

[0025] Calculate the cumulative distribution function as follows:

[0026]

[0027] Wherein, represents the cumulative distribution function of the th gray level;

[0028] Perform an inverse transformation on the cumulative distribution function to obtain a new pixel gray distribution and replace the original gray distribution of the image.

[0029] Furthermore, a gamma correction method is used for non-linear adjustment, and the correction formula is as follows:

[0030]

[0031] Wherein, and are the pixel values after and before correction respectively; is the gamma correction parameter.

[0032] Furthermore, the deep learning model is the YOLOv5 model. In the YOLOv5 model, a convolutional layer of the C3 module improved with the C2f structure is used to extract the features of the ray image and generate a feature vector.

[0033] Furthermore, in the optimization process, it also includes:

[0034] According to the specific features of the image, adaptively adjust the parameters of the optimization processing algorithm so that the processed image meets the preset standards in terms of gray level, contrast, and noise level.

[0035] In a second aspect, the present invention provides a detection device for under-voltage defects of strain clamps based on X-ray image recognition, including:

[0036] A preprocessing module for optimizing the original X-ray image of the strain clamp to obtain an optimized ray image;

[0037] A feature extraction module for inputting the optimized ray image into a deep learning model and using the deep learning model to extract the fine structural feature vector related to the under-voltage defect of the strain clamp in the image;

[0038] A defect detection module is used to input the extracted subtle structure feature vectors into a target detection deep learning model to identify the undervoltage defects in the original X-ray image, and rate the identified defects in combination with a preset standard; the target detection deep learning model adopts a one-stage or two-stage architecture.

[0039] In a third aspect, the present invention provides a computer device, which includes a processor and a memory:

[0040] The memory is used to store a computer program and send the instructions of the computer program to the processor;

[0041] The processor executes a method for detecting undervoltage defects of strain clamps based on X-ray image recognition according to the instructions of the computer program as described in the first aspect.

[0042] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements a method for detecting undervoltage defects of strain clamps based on X-ray image recognition as described in the first aspect.

[0043] In summary, the present invention provides a method for detecting undervoltage defects of strain clamps based on X-ray image recognition, including optimizing the original X-ray image of the strain clamp to obtain an optimized ray image; inputting the optimized ray image into a deep learning model, and using the deep learning model to extract subtle structure feature vectors related to the undervoltage defects of the strain clamp in the image; inputting the extracted subtle structure feature vectors into a target detection deep learning model to identify the undervoltage defects in the original X-ray image, and rating the identified defects in combination with a preset standard; the target detection deep learning model adopts a one-stage or two-stage architecture. By optimizing the X-ray image and using a target detection deep learning model for defect detection, the present invention can significantly improve the accuracy and reliability of detecting undervoltage defects of strain clamps.

[0044] The present invention also provides a device, equipment, and storage medium for detecting undervoltage defects of strain clamps based on X-ray image recognition, and when implemented, it has similar effects to those when the above method is implemented, which will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 A flow chart of a method for detecting undervoltage defects of a tension clamp based on X-ray image recognition provided by an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of a defect detection process provided by an embodiment of the present invention;

[0048] Figure 3 A schematic diagram of the composition of the C3 module provided in an embodiment of the present invention;

[0049] Figure 4 A schematic diagram of the composition of a C2f structure provided in an embodiment of the present invention;

[0050] Figure 5 A block diagram of a tension clamp undervoltage defect detection device based on X-ray image recognition provided by an embodiment of the present invention;

[0051] Figure 6 A block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] See also Figure 1 The embodiment of the present invention provides a method for detecting undervoltage defects of a tension clamp based on X-ray image recognition, comprising the following steps:

[0054] S1: Optimize the original X-ray image of the tension clamp to obtain an optimized X-ray image.

[0055] It should be noted that the original X-ray image of the tension clamp is an initial image of the tension clamp obtained by an X-ray imaging device, and the image may have quality problems, such as noise, poor contrast, etc.

[0056] This step uses image processing algorithms, such as filtering algorithms to remove noise, histogram equalization and other methods to enhance image contrast, and makes targeted adjustments to X-ray images based on their characteristics, thereby solving the problem of unstable X-ray image quality in the existing technology, improving image quality, and making subsequent feature extraction easier and more accurate. The optimized image provides better data input for the deep learning model, which helps to improve the accuracy of feature extraction and thus improve the accuracy of the final detection results.

[0057] S2: Input the optimized ray image into the deep learning model, and use the deep learning model to extract the subtle structural feature vectors related to the under-voltage defect of the strain clamp from the image.

[0058] It should be noted that the deep learning model is a complex model based on artificial neural networks, with powerful feature learning capabilities, and can automatically learn complex patterns and features from data. The deep learning model in this step is used to extract the features related to the under-voltage defect of the strain clamp from the optimized image. The subtle structural feature vector is a quantitative representation of the subtle structural information related to the under-voltage defect of the strain clamp in the image, forming a vector. This vector contains the key feature information that can characterize the under-voltage defect and is used for subsequent defect identification.

[0059] The deep learning model in this step is trained with a large amount of labeled data to learn the feature patterns of the images in the normal state and under-voltage defect state of the strain clamp. When the optimized image is input, the model can extract the subtle structural features related to the under-voltage defect according to the learned patterns and convert them into feature vectors.

[0060] S3: Input the extracted subtle structural feature vectors into the object detection deep learning model to identify the under-voltage defects in the original X-ray image, and rate the identified defects in combination with the preset criteria; the object detection deep learning model adopts a one-stage or two-stage architecture.

[0061] It should be noted that the object detection deep learning model is a deep learning model specifically used to detect specific objects (i.e., the under-voltage defect of the strain clamp) in the image. The one-stage architecture is an object detection architecture that directly predicts the category and location of the object in the network, with a relatively fast detection speed, but may have slightly lower detection accuracy in complex scenarios. The two-stage architecture performs object detection in two stages. In the first stage, candidate regions that may contain the object are generated, and in the second stage, these candidate regions are classified and refined in terms of location, with relatively high detection accuracy but relatively slow speed.

[0062] In this step, the object detection deep learning model uses the previously extracted subtle structural feature vectors and its trained classification and localization capabilities to determine whether there is an under-voltage defect of the strain clamp in the original X-ray image and determine its location. Then, according to the preset criteria, the identified defects are rated.

[0063] This embodiment provides a method for detecting undervoltage defects in tension clamps based on X-ray image recognition. The method first optimizes the original image to improve the image quality in view of the problem of unstable quality of existing X-ray images. Then, the powerful feature learning ability of the deep learning model is used to extract the fine structure feature vectors related to the undervoltage defects of the tension clamps from the optimized image. Finally, these feature vectors are input into the target detection deep learning model to realize the recognition and rating of undervoltage defects.

[0064] The detection method provided in this embodiment is specifically optimized for existing X-ray image quality issues, laying a good foundation for subsequent feature extraction and defect identification, and solving the problem of unstable image quality affecting the accuracy of detection results. At the same time, the deep learning model is used to automatically learn the subtle structural features of the undervoltage defect of the tension clamp, avoiding the limitations of manual feature extraction, and being able to mine more accurate feature information and improve detection accuracy. Different target detection deep learning models with different architectures are also used, which can be flexibly selected according to actual needs to balance detection speed and accuracy.

[0065] The defect detection process is as follows: Figure 2 As shown, Figure 2 The first to fourth small pictures correspond to the image optimization process, feature extraction process, defect recognition process and preprocessing process respectively. The above process is introduced below in combination with some other embodiments of the present invention.

[0066] In one embodiment, the original X-ray image of the tension clamp is optimized, including:

[0067] S11: The original X-ray image of the tension clamp is denoised using a Gaussian filtering algorithm.

[0068] The Gaussian filter algorithm is an image filtering algorithm based on a two-dimensional normal distribution (also called a Gaussian distribution). De-noising is achieved by taking a weighted average of the pixels in the image window, where the weights follow a two-dimensional normal distribution.

[0069] It uses the characteristics of the Gaussian function to update each pixel in the image according to the weighted average of its surrounding pixels, and its weight is determined by a two-dimensional normal distribution. For noise points in the image, since they are quite different from the surrounding pixels, the weighted average method can make their values ​​closer to the surrounding pixels, thus achieving the effect of denoising.

[0070] S12: grayscale the denoised image using a histogram equalization method.

[0071] Histogram equalization is a method to enhance the contrast of an image by changing the histogram distribution of the image. It calculates the histogram distribution probability of the image, and then performs an inverse transformation using the histogram cumulative distribution function to obtain the processed pixel gray-level distribution, thereby redistributing the gray levels of the image, making the gray-level distribution of the image more uniform and enhancing the overall contrast of the image.

[0072] S13: Perform non-linear adjustment on the grayscale image, and adjust the brightness and contrast of different gray-level regions in the image through a non-linear function.

[0073] Non-linear adjustment refers to performing non-linear tone editing on the gamma curve of the image using non-linear functions such as gamma correction. By adjusting the ratio of the dark and light parts in the image, the enhancement or weakening of the image contrast can be achieved.

[0074] In this embodiment, first, the Gaussian filtering algorithm is used to accurately calculate the weight value according to the distance between the pixel point and the center point of the window and the standard deviation, multiply the original pixel value of the pixel point within the image window by the corresponding weight value, effectively removing the noise caused by equipment and environmental factors, and making the image smoother. This not only provides a clearer and more stable image basis for subsequent operations such as histogram equalization and gamma correction, avoiding the interference of noise on subsequent processing, but also provides high-quality input data for the deep learning model, greatly improving the accuracy and reliability of the entire detection of under-voltage defects in strain clamps. At the same time, by introducing an adaptive parameter adjustment mechanism, the parameters can be flexibly adjusted according to the noise characteristics of the image to achieve a better denoising effect. This series of optimized processing methods can better highlight the feature performance of under-voltage defects in the strain clamp in the image, such as more clearly presenting the deformation of the crimping part, the change of internal density, and the change of texture features during under-voltage, overcoming the problem that the poor quality of existing X-ray images affects feature extraction and the final detection result, making subsequent feature extraction, defect recognition, etc. more accurate and efficient, and providing a stronger guarantee for the safety and reliability of power equipment.

[0075] In one embodiment, the Gaussian filtering algorithm is used to perform denoising processing on the original X-ray image of the strain clamp, including:

[0076] For the pixel values of all pixel points within the image window, calculate the corresponding weight value, and the weight calculation formula is as follows:

[0077]

[0078] In the formula, represents the weight value of pixel point ; represents the standard deviation; and represent the coordinates of the center point of the window;

[0079] Multiply the original pixel values of all pixel points within the image window by the corresponding weight values to obtain the denoised ray image.

[0080] In one embodiment, the denoised image is grayed out using the histogram equalization method, including:

[0081] Count the number of pixels at each gray level in the image and record it as the original gray distribution of the image;

[0082] Calculate the histogram distribution probability for each gray level as follows:

[0083]

[0084] In the formula, represents the histogram distribution probability of the th gray level; represents the th gray level of the number of pixels; represents the total number of pixels in the image;

[0085] Calculate the cumulative distribution function as follows:

[0086]

[0087] In the formula, represents the cumulative distribution function of the

[0088] Inverse-transform the cumulative distribution function to obtain a new pixel gray distribution and replace the original gray distribution of the image.

[0089] In one embodiment, the gamma correction method is used for non-linear adjustment, and the correction formula is as follows:

[0090]

[0091] In the formula, and are the pixel values after and before correction, respectively; is the gamma correction parameter.

[0092] In one embodiment, the deep learning model is the YOLOv5 model. In the YOLOv5 model, the convolutional layer of the C3 module improved by the C2f structure is used to extract the features of the ray image and generate feature vectors.

[0093] The YOLOv5 model is a popular deep learning model for object detection, which is widely used in the field of object detection for its fast speed and relatively high accuracy. It divides the object detection task into different scales, processes feature maps of different scales through a Feature Pyramid Network (PAN), generates multi-scale predictions, and uses a prediction head to perform object classification and localization predictions on features of different scales.

[0094] The convolutional layer of the C3 module is an important part of the YOLOv5Backbone structure. The number of stacked C2f structures is controlled by the parameter "n", which can vary with models of different scales. Its structure is as Figure 3 shown.

[0095] The C2f structure is an improved structure of the C3 module, which adopts the shuffling idea of CSPNet and the concept of residual structure. Its structure is as Figure 4 shown.

[0096] When the convolutional layer of the C3 module improved by the C2f structure in this embodiment performs forward propagation, it divides the input image features. One part directly performs forward propagation through the convolutional layer, and the other part passes through a more complex structure and then adds it to the former. This structure utilizes residual connections, enabling information to be transmitted more smoothly in the network and avoiding the loss or attenuation of information in the deep network. Compared with the traditional convolutional layer structure, the convolutional layer of the C3 module improved by the C2f structure can better utilize upstream information. While increasing the network depth and receptive field, it avoids the problem of gradient disappearance caused by deepening the network, enabling the model to learn more complex feature representations, improving the feature extraction ability for under-voltage defects of strain clamps, and thus improving the performance and accuracy of the entire detection system.

[0097] In one embodiment, during the optimization process, it further includes:

[0098] According to specific features of the image, adaptively adjust the parameters of the optimization processing algorithm to make the processed image meet the preset standards in terms of grayscale, contrast, and noise level.

[0099] During the image processing process, different original X-ray images of strain clamps may have different noise levels, contrasts, and grayscale distributions because they are affected by various factors, such as imaging devices, imaging environments, and the states of strain clamps. By analyzing specific features of the image, such as evaluating the grayscale distribution and contrast of the image by calculating the histogram of the image and measuring the noise level using statistical methods, it can provide a basis for adjusting the parameters of the optimization processing algorithm.

[0100] For the Gaussian filtering algorithm, the kernel size and standard deviation of Gaussian filtering are adaptively adjusted according to the noise level and distribution of the image. For areas or images with high noise, a larger kernel and standard deviation can be used to enhance the denoising effect; for areas or images with low noise, a smaller kernel and standard deviation are used to avoid loss of details caused by over-smoothing.

[0101] For histogram equalization and gamma correction, the parameters of gamma correction are adjusted according to the initial gray distribution of the image. If the image is overall dark, a smaller gamma value is required to enhance the details in the dark areas; if the contrast of the image is insufficient, the parameters of histogram equalization need to be adjusted to expand the gray range and enhance the contrast.

[0102] Based on the same inventive concept, the embodiment of the present application also provides a device for detecting under-voltage defects of strain clamps based on X-ray image recognition for implementing the method for detecting under-voltage defects of strain clamps based on X-ray image recognition involved above. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in the embodiment of the device for detecting under-voltage defects of strain clamps based on X-ray image recognition provided below can refer to the limitations on the method for detecting under-voltage defects of strain clamps based on X-ray image recognition in the above text, and will not be repeated here.

[0103] Please refer to Figure 5 , the embodiment of the present invention provides a device for detecting under-voltage defects of strain clamps based on X-ray image recognition, including:

[0104] A preprocessing module, configured to optimize the original X-ray image of the strain clamp to obtain an optimized ray image;

[0105] A feature extraction module, configured to input the optimized ray image into a deep learning model, and use the deep learning model to extract the fine structure feature vectors related to the under-voltage defects of the strain clamp in the image;

[0106] A defect detection module, configured to input the extracted fine structure feature vectors into a target detection deep learning model to identify the under-voltage defects in the original X-ray image, and rate the identified defects in combination with a preset standard; the target detection deep learning model adopts a one-stage or two-stage architecture.

[0107] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of the functional units and modules are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working process of the units and modules in the above system can refer to the corresponding process in the foregoing method embodiment and will not be elaborated here.

[0108] Referring to Figure 6 , an embodiment of the present invention further provides a computer device, including: a memory, a processor, and a computer program stored on the memory. When the computer program is executed on the processor, it implements a method for detecting under-voltage defects of strain clamps based on X-ray image recognition as described in any one of the above methods.

[0109] The computer device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that Figure 6 merely an example of a computer device, which does not constitute a limitation on the computer device, and may include more or fewer components than shown in the figure, or combine some components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0110] The so-called processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0111] In some embodiments, the memory may be an internal storage unit of the computer device, such as the hard disk or memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device. Further, the memory may also include both the internal storage unit and the external storage device of the computer device. The memory is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program. The memory may also be used to temporarily store data that has been output or is to be output.

[0112] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it implements a method for detecting under-voltage defects of strain clamps based on X-ray image recognition as described in any one of the above methods.

[0113] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above embodiment methods of the present application, a computer program can be used to instruct relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium may not be an electrical carrier signal and a telecommunication signal.

[0114] In the above embodiments, the descriptions of the various embodiments have their own focuses. For parts not detailed or recorded in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0115] Those of ordinary skill in the art will realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0116] In the embodiments disclosed in this application, it should be understood that the disclosed devices / terminal devices and methods can be implemented in other ways. For example, the device / terminal device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the devices or units can be in electrical, mechanical or other forms.

[0117] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting undervoltage defects of tension clamps based on X-ray image recognition, characterized in that: The steps include: Optimizing the original X-ray image of the tension clamp to obtain an optimized X-ray image; Inputting the optimized radiographic image into a deep learning model, and using the deep learning model to extract fine structural feature vectors related to undervoltage defects in the tension clamp in the image; Inputting the extracted fine structure feature vector into a target detection deep learning model to identify undervoltage defects in the original X-ray image, and rating the identified defects in combination with preset standards; The target detection deep learning model adopts a one-stage or two-stage architecture.

2. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 1, characterized in that: Optimize the original X-ray image of the tension clamp, including: The Gaussian filtering algorithm is used to denoise the original X-ray image of the tension clamp; The denoised image is grayed out using the histogram equalization method; The grayscale image is nonlinearly adjusted, and the brightness and contrast of different grayscale areas in the image are adjusted through nonlinear functions.

3. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 2 is characterized in that: The Gaussian filtering algorithm is used to denoise the original X-ray image of the tension clamp, including: For the pixel values ​​of all pixels in the image window, calculate the corresponding weight value. The weight calculation formula is as follows: In the formula, Represents pixel The weight value of represents standard deviation; and Indicates the coordinates of the center point of the window; The original pixel values ​​of all pixels in the image window are multiplied by the corresponding weight values ​​to obtain a denoised ray image.

4. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 2, characterized in that: The denoised image is grayed out using a histogram equalization method, including: Count the number of pixels at each gray level in the image and record it as the original gray distribution of the image; Calculate the histogram distribution probability of each gray level as follows: In the formula, Indicates Histogram distribution probability of gray levels; Indicates The number of pixels with gray levels; Represents the total number of pixels in the image; Calculate the cumulative distribution function as follows: In the formula, express The cumulative distribution function of gray levels; The cumulative distribution function is inversely transformed to obtain a new pixel grayscale distribution, and replace the original grayscale distribution of the image.

5. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 2, characterized in that: The gamma correction method is used for nonlinear adjustment, and the correction formula is as follows: In the formula, and are the pixel values ​​before and after correction, respectively; is the gamma correction parameter.

6. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 1, characterized in that: The deep learning model is a YOLOv5 model. In the YOLOv5 model, a C3 module convolution layer with an improved C2f structure is used to extract features of the radiographic image and generate a feature vector.

7. The method for detecting undervoltage defects of tension clamps based on X-ray image recognition according to claim 1, characterized in that: The optimization process also includes: According to the specific characteristics of the image, the parameters of the processing algorithm are adaptively adjusted and optimized so that the processed image meets the preset standards in terms of grayscale, contrast and noise level.

8. A tension clamp undervoltage defect detection device based on X-ray image recognition, characterized in that: include: A preprocessing module is used to optimize the original X-ray image of the tension clamp to obtain an optimized X-ray image; A feature extraction module, used for inputting the optimized radiographic image into a deep learning model, and using the deep learning model to extract fine structural feature vectors related to undervoltage defects of the tension clamp in the image; A defect detection module, used for inputting the extracted fine structure feature vector into a target detection deep learning model, identifying undervoltage defects in the original X-ray image, and rating the identified defects in combination with a preset standard; The target detection deep learning model adopts a one-stage or two-stage architecture.

9. A computer device, characterized in that: The device comprises a processor and a memory: The memory is used to store a computer program and send instructions of the computer program to the processor; The processor executes a method for detecting undervoltage defects of a tension clamp based on X-ray image recognition according to any one of claims 1 to 7 according to the instructions of the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for detecting undervoltage defects of a tension clamp based on X-ray image recognition according to any one of claims 1 to 7 is implemented.

Images