Strain clamp crimping defect X-ray image information preprocessing system and method

Abstract

The invention relates to the technical field of image processing, in particular to a strain clamp crimping defect X-ray image information preprocessing system and method.The strain clamp crimping defect X-ray image information preprocessing system comprises a detector hardware module which adopts a high-precision X-ray detector, is unique in design, has extremely high sensitivity and resolution ratio and can accurately capture subtle differences generated by X-ray absorption of substances with different densities; x-ray signals are efficiently converted into electric signals; and the analog-to-digital conversion module is connected with the detector hardware module, adopts an advanced analog-to-digital conversion technology and performs high-sampling-rate and high-bit quantization. According to the method, in a real-time diagnosis scene, a UDP protocol is combined with forward error correction coding, data transmission delay is effectively reduced, in an offline analysis scene, a TCP protocol is combined with data compression, data transmission integrity is guaranteed, meanwhile, occupation of data storage space is reduced, and data management efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a preprocessing system and method for X-ray image information of compression defects of strain clamps. Background Art

[0002] In the power transmission network, strain clamps are extremely important devices that shoulder multiple key tasks. At the line turning, connection and terminal parts, strain clamps, by virtue of their anchoring function, firmly fix conductors and lightning conductors on the strain insulator strings of non-straight towers. Similarly, they are also used to fix the guy wires of guyed towers. According to the operation and maintenance standards of power grid companies, in the "three-span" areas, namely the crossing sections of railways, highways and important lines, X-ray detection technology is required to perform non-destructive testing on strain clamps.

[0003] When using X-ray to detect strain clamps, X-ray fluoroscopy imaging technology is utilized. When X-rays penetrate the strain clamp, due to the differences in the thickness and density of each component inside the clamp, the ability to absorb X-rays is also different, thus forming a transmitted image with different intensity distributions. The detector converts the X-ray intensity difference into a visual image, which not only realizes the non-destructive testing of the compression process of strain clamps and obtains high-precision imaging results, but also can accurately diagnose the compression quality of the steel core of the line compression pipe, judge whether the compression is in place and firm, and whether there are potential hazards such as cracks. The application of this detection technology not only reduces the work burden of detection personnel, but also improves the detection accuracy, laying a solid foundation for the safe and stable operation of equipment.

[0004] Based on a comprehensive analysis of the X-ray detection data of strain clamps, common defect types of strain clamps are sorted out, including burrs and deformations on the surface of the steel anchor pipe, lantern-shaped deformations of the steel core, air compression of the aluminum pipe at the steel anchor pipe, non-compression of the aluminum strands at the end of the steel anchor pipe, insufficient compression of the aluminum pipe at the steel anchor groove, incorrect compression position of the aluminum pipe in the non-compression area of the steel anchor pipe, and non-compression of the aluminum pipe at the steel anchor groove, etc.

[0005] Existing technologies such as the invention with the publication number of CN118587184A disclose an X-ray detection method, system and device based on artificial intelligence, belonging to the technical field of X-ray detection. The method specifically includes: collecting X-ray image data covering different application scenarios and detection objects, preprocessing the collected X-ray image data to obtain an X-ray public data set, training an X-ray detection model, optimizing the X-ray detection model using a generative adversarial network, using the optimized X-ray detection model to segment the preprocessed X-ray image data, extract the region of interest, extract features from the extracted region of interest, identify the abnormal regions and abnormal types in the X-ray image data, label the detected abnormal regions, and generate a detailed detection report.

[0006] The X-ray image processing system in the prior art has three key defects: First, in terms of data transmission, a fixed protocol is adopted, which cannot be dynamically switched according to the requirements of different application scenarios, and lacks a real-time monitoring mechanism for network bandwidth, packet loss rate and other states, resulting in a significant decrease in the reliability of data transmission in high-latency environments and affecting the accuracy of subsequent analysis; Second, the image segmentation module adopts a closed automatic processing mode. When the AI model outputs an incorrect segmentation result, there is no mechanism that allows professionals to make real-time corrections interactively, resulting in a decrease in diagnostic efficiency. Summary of the Invention

[0007] The purpose of the present invention is to solve the defects existing in the prior art: First, in terms of data transmission, a fixed protocol is adopted, which cannot be dynamically switched according to the requirements of different application scenarios such as real-time diagnosis and offline analysis; Second, the image segmentation module adopts a closed automatic processing mode. When the AI model outputs an incorrect segmentation result, there is no mechanism that allows professionals to make real-time corrections interactively. A preprocessing system and method for X-ray image information of compression defects of strain clamps are proposed.

[0008] To achieve the above purpose, the present invention adopts the following technical solutions: A preprocessing system and method for X-ray image information of compression defects of strain clamps, including:

[0009] Detector hardware module: A high-precision X-ray detector is adopted, with a unique design, extremely high sensitivity and resolution, capable of accurately capturing the subtle differences generated by different density substances absorbing X-rays, and efficiently converting X-ray signals into electrical signals;

[0010] Analog-to-digital conversion module: Connected to the detector hardware module, adopting advanced analog-to-digital conversion technology, with high sampling rate and high-bit quantization, accurately converting the analog electrical signals output by the detector into digital signals;

[0011] Data cache module: Set between the data acquisition and transmission links, it can temporarily store the acquired digital image data when the data acquisition rate is higher than the transmission rate, and preliminarily organize and verify the data;

[0012] High-speed transmission module: Using cutting-edge high-speed network communication technology, by optimizing network protocols and data transmission strategies, quickly and stably transmitting the digital image data output by the data cache module to the processing center, and encrypting and verifying the data using a specific data encapsulation protocol;

[0013] Image format conversion module: After the data is transmitted to the processing center, it uniformly converts various original image formats output by different detectors into a standard format convenient for the system to process, and normalizes the image data, adjusting the image gray range and size;

[0014] Image Enhancement Module: Integrates a variety of advanced image enhancement algorithms, including the adaptive histogram equalization algorithm, which can dynamically adjust the gray-scale distribution according to the local features of the image, enhancing the image contrast and sharpness;

[0015] Noise Reduction Processing Module: Adopts advanced noise reduction algorithms based on wavelet transform, etc., accurately identifies and removes detector noise, environmental interference noise, etc. in the X-ray images of compression defects of strain clamps, while retaining the effective information of the images;

[0016] Image Segmentation Module: Utilizes semantic segmentation algorithms based on deep learning to accurately segment the X-ray images of compression defects of strain clamps after enhancement and noise reduction processing;

[0017] Result Output and Storage Module: Diversely outputs and stores the X-ray images of compression defects of strain clamps after intelligent processing, can be connected to a high-resolution display, stored in a large-capacity database and managed according to specific classification and indexing rules, and supports output in industrial general formats such as DICOM;

[0018] The output end of the detector hardware module is connected to the analog-to-digital conversion module, the output end of the analog-to-digital conversion module is sequentially connected to the data buffer module, the high-speed transmission module, and the image format conversion module, the output end of the image format conversion module is connected to the image enhancement module and the noise reduction processing module in parallel, the output ends of the image enhancement module and the noise reduction processing module are jointly connected to the image segmentation module, and the output end of the image segmentation module is connected to the result output and storage module, forming a linear and parallel hybrid data processing link,

[0019] This solution can meet the requirements of different scenarios with a single model architecture by constructing a network for extracting underlying features shared by different scenarios and combining a domain-specific adaptation layer, greatly reducing the model development and maintenance costs.

[0020] Preferably, the high-speed transmission module includes:

[0021] Dynamic Protocol Switching Unit: Automatically selects the transmission protocol according to the application scenario requirements, uses the UDP protocol combined with forward error correction coding for real-time diagnosis scenarios to reduce latency, and uses the TCP protocol combined with data compression for offline analysis scenarios to ensure integrity;

[0022] Network Status Monitoring Unit: Real-time evaluates the current network bandwidth and packet loss rate, and dynamically adjusts the data fragmentation size and retransmission strategy.

[0023] Preferably, the image segmentation module includes:

[0024] Interactive Correction Interface: Receives the annotation correction instructions from the user for the segmentation results, fine-tunes the output of the deep learning model in real time, and feeds back the correction results to the model training process;

[0025] Domain Adaptation Unit: By sharing the underlying feature extraction network for different scenarios and combining with a domain-specific adaptation layer, cross-scenario model migration is achieved.

[0026] The image segmentation module uses a deep learning-based semantic segmentation algorithm. In industrial inspection, it can clearly divide different structural parts of metal components and the regions where defects are located, providing key evidence for product quality assessment. Its domain adaptation unit achieves cross-scenario model migration by sharing the underlying feature extraction network for different scenarios and combining with a domain-specific adaptation layer, improving the generality and adaptability of the model.

[0027] Preferably, the digital image data converted by the analog-to-digital conversion module maximally restores the original X-ray image information and reduces data distortion.

[0028] Preferably, the high-speed network communication technologies adopted by the high-speed transmission module include optical fiber, 5G, or Wi-Fi6. The dynamic protocol switching unit in the high-speed transmission module can automatically select the transmission protocol according to the application scenario requirements. In the real-time diagnosis scenario, the UDP protocol combined with forward error correction coding is used to effectively reduce data transmission latency, and the TCP protocol combined with data compression is used to ensure the integrity of data transmission, while reducing the occupancy of data storage space and improving data management efficiency.

[0029] Preferably, the image processed by the image enhancement module can effectively highlight the subtle defects in industrial inspection images.

[0030] Preferably, an improved U-Net model with an encoder-decoder structure is adopted. Its encoder part extracts multi-scale features through consecutive downsampling layers, and the decoder part realizes feature map upsampling through transposed convolution. Among them, the skip connection concatenates the feature map of the j-th layer of the encoder with the feature map of the corresponding layer of the decoder along the channel dimension, expressed as:

[0031]

[0032] where σ is the ReLU activation function, W j and b j are learnable parameters, and ∥ represents channel concatenation;

[0033] Loss function design: Adopt the weighted combination of cross-entropy loss L ce and Dice loss L dice :

[0034] L = αL ce +(1 - α)L dice ;

[0035] where α ∈ [0, 1] is the balancing weight, and the Dice loss, aiming at the problem of unbalanced proportion of the target area in the industrial scenario, is defined as:

[0036]

[0037] p i is the predicted probability of pixel i, g i is the ground truth label, and ε is the smoothing term.

[0038] Preferably, the image processed by the noise reduction processing module has an improved signal-to-noise ratio, providing purer image data for subsequent image analysis and diagnosis. The noise reduction processing module adopts advanced noise reduction algorithms such as wavelet transform to accurately identify and remove detector noise, environmental interference noise, etc. in the X-ray image of the compression defect of the strain clamp, while maximizing the retention of the effective information of the image, avoiding damage to the image details during the noise reduction process, improving the signal-to-noise ratio of the image, providing purer image data for subsequent image analysis and diagnosis, and enhancing the image quality.

[0039] Preferably, the image segmentation module can clearly divide different structural parts of the metal component and the area where the defect is located in industrial inspection. The result output and storage module realizes seamless integration with other industrial inspection systems. The result output and storage module supports outputting the processed X-ray image of the compression defect of the strain clamp in industrial common formats such as DICOM, facilitating seamless integration with other industrial inspection systems, realizing data sharing and collaborative work. At the same time, this module stores the images in a large-capacity database and manages them according to specific classification and indexing rules, and stores them based on data value grading, facilitating users to quickly retrieve and call important data and improving the utilization efficiency of the data.

[0040] Preferably, it includes the following steps:

[0041] S1. Signal acquisition step: Use a high-precision X-ray detector to capture the absorption difference of X-rays by different density substances and convert the X-ray signal into an analog electrical signal;

[0042] S2. Analog-to-digital conversion step: Convert the analog electrical signal into a digital signal through high sampling rate and high-bit quantization;

[0043] S3. Data caching step: Temporarily store digital image data when the acquisition rate is higher than the transmission rate and perform data verification;

[0044] S4. High-speed transmission step: Adopt a dynamic protocol switching strategy to transmit data. Use the UDP protocol and forward error correction coding in the real-time diagnosis scenario, and use the TCP protocol and data compression in the offline analysis scenario;

[0045] S5. Format conversion step: Uniformly convert image data from different sources into a standard format, and perform grayscale normalization and size adjustment;

[0046] S6. Image enhancement step: Dynamically adjust the local grayscale distribution of the image through the adaptive histogram equalization algorithm;

[0047] S7. Noise reduction processing step: Remove detector noise and environmental noise based on wavelet transform, and retain effective information;

[0048] S8. Image segmentation step: Use a deep learning model for semantic segmentation, and receive user interaction instructions to correct the segmentation results in real time;

[0049] S9. Output storage step: Output the processing results according to the DICOM standard, and store them hierarchically based on data value.

[0050] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0051] In the present invention, the dynamic protocol switching unit in the high-speed transmission module can automatically select a transmission protocol according to the application scenario requirements. In the real-time diagnosis scenario, the UDP protocol is combined with forward error correction coding to effectively reduce data transmission latency. In the offline analysis scenario, the TCP protocol is combined with data compression to ensure the integrity of data transmission, while reducing the occupation of data storage space and improving data management efficiency.

[0052] In the present invention, the network status monitoring unit evaluates the current network bandwidth and packet loss rate in real time, and dynamically adjusts the data shard size and retransmission strategy to ensure that digital image data can still be quickly and stably transmitted from the data cache module to the processing center in a complex network environment. An adaptive network optimization mechanism is adopted to improve the reliability of data transmission and avoid data loss or transmission interruption caused by network fluctuations.

[0053] In the present invention, the image segmentation module uses a semantic segmentation algorithm based on deep learning. In industrial inspection, it can clearly divide different structural parts of metal components and the areas where defects are located, providing a key basis for product quality assessment. Its domain adaptation unit realizes cross-scene model migration by sharing the underlying feature extraction network of different scenes and combining with a domain-specific adaptation layer, improving the generality and adaptability of the model.

[0054] In the present invention, the interactive correction interface of the image segmentation module receives the annotation correction instructions of the user for the segmentation results, fine-tunes the output of the deep learning model in real time, and feeds back the correction results to the model training process. This human-computer interaction mechanism not only improves the accuracy of the segmentation results but also continuously optimizes the deep learning model to better adapt to different application scenarios and user requirements.

[0055] In the present invention, the noise reduction processing module adopts advanced noise reduction algorithms such as wavelet transform to accurately identify and remove detector noise, environmental interference noise, etc. in the X-ray images of compression defects of strain clamps, while maximizing the retention of the effective information of the images, avoiding damage to the image details during the noise reduction process, improving the signal-to-noise ratio of the images, providing purer image data for subsequent image analysis and diagnosis, and enhancing the image quality.

[0056] In the present invention, the result output and storage module supports outputting the processed X-ray images of compression defects of strain clamps in industrial general formats such as DICOM, which is convenient for seamless integration with other industrial detection systems to achieve data sharing and collaborative work. At the same time, this module stores the images in a large-capacity database, manages them according to specific classification and indexing rules, and stores them based on data value grading, facilitating users to quickly retrieve and call important data and improving the utilization efficiency of the data.

[0057] In the present invention, the detector hardware module adopts a uniquely designed high-precision X-ray detector with extremely high sensitivity and resolution, which can accurately capture the subtle differences in the absorption of X-rays by substances with different densities. In industrial detection, it can keenly detect extremely small defects inside metal components, such as cracks as thin as hair, greatly improving the accuracy of product quality inspection and reducing defective products.

[0058] In the present invention, by constructing a network for extracting underlying features shared by different scenarios and combining it with a domain-specific adaptation layer, a set of model architectures can meet the requirements of different scenarios, greatly reducing the model development and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is the system architecture diagram of the processing system of the present invention;

[0060] Figure 2 is the system architecture diagram of the high-speed transmission module in the processing system of the present invention;

[0061] Figure 3 is the system architecture diagram of the image segmentation module in the processing system of the present invention;

[0062] Figure 4 is the method flow chart of the processing method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0063] As shown in the figure: An intelligent preprocessing system for X-ray images of compression defects of strain clamps includes:

[0064] Detector hardware module: Adopts a high-precision X-ray detector with a unique design, extremely high sensitivity and resolution, which can accurately capture the subtle differences in the absorption of X-rays by substances with different densities and efficiently convert the X-ray signals into electrical signals;

[0065] Analog-to-Digital Conversion Module: Connected to the detector hardware module, it uses advanced analog-to-digital conversion technology to accurately convert the analog electrical signals output by the detector into digital signals with a high sampling rate and high-bit quantization.

[0066] Data Buffer Module: Set between the data acquisition and transmission links, it can temporarily store the acquired digital image data when the data acquisition rate is higher than the transmission rate, and preliminarily organize and verify the data.

[0067] High-Speed Transmission Module: Using cutting-edge high-speed network communication technology, by optimizing network protocols and data transmission strategies, it quickly and stably transmits the digital image data output by the data buffer module to the processing center, and encrypts and verifies the data using a specific data encapsulation protocol.

[0068] Image Format Conversion Module: After the data is transmitted to the processing center, it unifies the conversion of various original image formats output by different detectors into a standard format convenient for the system to process, and normalizes the image data, adjusting the image gray range and size.

[0069] Image Enhancement Module: Integrating a variety of advanced image enhancement algorithms, including the adaptive histogram equalization algorithm, it can dynamically adjust the gray distribution according to the local features of the image, enhancing the image contrast and sharpness.

[0070] Noise Reduction Processing Module: Using advanced noise reduction algorithms based on wavelet transform, etc., it accurately identifies and removes detector noise, environmental interference noise, etc. in the X-ray images of the compression defects of strain clamp, while retaining the effective information of the images.

[0071] Image Segmentation Module: Using the semantic segmentation algorithm based on deep learning, it accurately segments the X-ray images of the compression defects of strain clamp after enhancement and noise reduction processing.

[0072] Result Output and Storage Module: Diversely outputs and stores the X-ray images of the compression defects of strain clamp after intelligent processing, can be connected to a high-resolution display, stored in a large-capacity database and managed according to specific classification and indexing rules, and supports output in industrial common formats such as DICOM.

[0073] The output end of the detector hardware module is connected to the analog-to-digital conversion module, the output end of the analog-to-digital conversion module is sequentially connected to the data buffer module, the high-speed transmission module, and the image format conversion module. The output end of the image format conversion module is connected to the image enhancement module and the noise reduction processing module in parallel. The output ends of the image enhancement module and the noise reduction processing module are jointly connected to the image segmentation module, and the output end of the image segmentation module is connected to the result output and storage module, forming a linear and parallel hybrid data processing link.

[0074] This solution constructs a network for extracting underlying features shared across different scenarios and combines it with a domain-specific adaptation layer. With this one model architecture, different scenario requirements can be met, significantly reducing the costs of model development and maintenance.

[0075] Specifically, the high-speed transmission module includes:

[0076] A dynamic protocol switching unit: Automatically selects the transmission protocol according to the application scenario requirements. For real-time diagnosis scenarios, the UDP protocol combined with forward error correction coding is used to reduce latency. For offline analysis scenarios, the TCP protocol combined with data compression is used to ensure integrity.

[0077] A network status monitoring unit: Real-time evaluates the current network bandwidth and packet loss rate, and dynamically adjusts the data fragmentation size and retransmission strategy.

[0078] Specifically, the image segmentation module includes:

[0079] An interactive correction interface: Receives the annotation correction instructions from the user for the segmentation result, fine-tunes the output of the deep learning model in real time, and feeds back the correction result to the model training process.

[0080] A domain adaptation unit: Achieves cross-scenario model migration by sharing the underlying feature extraction network across different scenarios and combining with a domain-specific adaptation layer.

[0081] The image segmentation module uses a semantic segmentation algorithm based on deep learning. In industrial inspection, it can clearly divide different structural parts of metal components and the areas where defects are located, providing a key basis for product quality assessment. Its domain adaptation unit achieves cross-scenario model migration through the underlying feature extraction network across different scenarios and combines with a domain-specific adaptation layer, improving the generality and adaptability of the model.

[0082] Specifically, the digital image data converted by the analog-to-digital conversion module maximally restores the original X-ray image information, reducing data distortion.

[0083] Specifically, the high-speed network communication technologies adopted by the high-speed transmission module include optical fiber, 5G, or WIFI6. The dynamic protocol switching unit in the high-speed transmission module can automatically select the transmission protocol according to the application scenario requirements. In real-time diagnosis scenarios, the UDP protocol combined with forward error correction coding is used to effectively reduce data transmission latency. In offline analysis scenarios, the TCP protocol combined with data compression is used to ensure the integrity of data transmission, while reducing the occupancy of data storage space and improving data management efficiency.

[0084] Specifically, the image processed by the image enhancement module can effectively highlight the subtle defects in industrial inspection images.

[0085] In this embodiment: An improved U-Net model with an encoder-decoder structure is adopted. The encoder part extracts multi-scale features through consecutive downsampling layers, and the decoder part realizes upsampling of the feature map through transposed convolution. Among them, the skip connection splices the feature map of the j-th layer of the encoder with the feature map of the corresponding layer of the decoder along the channel dimension, which is expressed as:

[0086]

[0087] where σ is the ReLU activation function, W j and b j are learnable parameters, and ∥ represents channel splicing;

[0088] Loss function design: The cross-entropy loss L ce and the Dice loss L dice are used in a weighted combination:

[0089] L = αL ce +(1 - α)L dice ;

[0090] where α ∈ [0, 1] is the balancing weight. The Dice loss is defined for the problem of unbalanced proportion of the target area in the industrial scenario as:

[0091]

[0092] p i is the predicted probability of pixel i, g i is the ground truth label, and ε is a smoothing term.

[0093] Specifically, the image processed by the noise reduction processing module has an improved signal-to-noise ratio, providing purer image data for subsequent image analysis and diagnosis. The noise reduction processing module adopts advanced noise reduction algorithms such as wavelet transform to accurately identify and remove detector noise, environmental interference noise, etc. in the X-ray images of compression defects of strain clamps, while maximizing the retention of the effective information of the image, avoiding damage to the image details during the noise reduction process, improving the signal-to-noise ratio of the image, providing purer image data for subsequent image analysis and diagnosis, and enhancing the image quality.

[0094] In this embodiment, the image segmentation module can clearly divide different structural parts of metal components and the regions where defects are located in industrial inspection. The result output and storage module realizes seamless integration with other industrial inspection systems. The result output and storage module supports outputting the processed X-ray images of compression defects of strain clamps in industrial common formats such as DICOM, which facilitates seamless integration with other industrial inspection systems, realizes data sharing and collaborative work. At the same time, this module stores the images in a large-capacity database, manages them according to specific classification and indexing rules, and stores them hierarchically based on data value, which is convenient for users to quickly retrieve and call important data and improves the utilization efficiency of data.

[0095] Specifically, it includes the following steps:

[0096] S1. Signal acquisition step: Use a high-precision X-ray detector to capture the absorption differences of X-rays by substances with different densities, and convert the X-ray signal into an analog electrical signal;

[0097] S2. Analog-to-digital conversion step: Convert the analog electrical signal into a digital signal through high sampling rate and high-bit quantization;

[0098] S3. Data caching step: Temporarily store digital image data when the acquisition rate is higher than the transmission rate, and perform data verification;

[0099] S4. High-speed transmission step: Use a dynamic protocol switching strategy to transmit data. The UDP protocol and forward error correction coding are used in the real-time diagnosis scenario, and the TCP protocol and data compression are used in the offline analysis scenario;

[0100] S5. Format conversion step: Uniformly convert image data from different sources into a standard format, and perform gray-scale normalization and size adjustment;

[0101] S6. Image enhancement step: Dynamically adjust the local gray-scale distribution of the image through the adaptive histogram equalization algorithm;

[0102] S7. Noise reduction processing step: Remove detector noise and environmental noise based on wavelet transform, and retain effective information;

[0103] S8. Image segmentation step: Use a deep learning model for semantic segmentation, and receive user interaction instructions to correct the segmentation result in real time;

[0104] S9. Output storage step: Output the processing result according to the DICOM standard, and store it hierarchically based on data value.

Claims

1. A preprocessing system for X-ray image information of compression defects of strain clamps, characterized in that, Including: Detector hardware module: Adopting a high-precision X-ray detector with a unique design, it has extremely high sensitivity and resolution, can accurately capture the subtle differences caused by the absorption of X-rays by substances with different densities, and efficiently convert X-ray signals into electrical signals; Analog-to-digital conversion module: Connected to the detector hardware module, it uses advanced analog-to-digital conversion technology, with a high sampling rate and high-bit quantization, to accurately convert the analog electrical signals output by the detector into digital signals; Data cache module: Set between the data acquisition and transmission links, it can temporarily store the acquired digital image data when the data acquisition rate is higher than the transmission rate, and preliminarily organize and verify the data; High-speed transmission module: Using cutting-edge high-speed network communication technology, by optimizing the network protocol and data transmission strategy, it quickly and stably transmits the digital image data output by the data cache module to the processing center, and encrypts and verifies the data using a specific data encapsulation protocol; Image format conversion module: After the data is transmitted to the processing center, it uniformly converts the various original image formats output by different detectors into a standard format convenient for the system to process, and normalizes the image data, adjusting the image gray range and size; Image enhancement module: Integrating a variety of advanced image enhancement algorithms, including the adaptive histogram equalization algorithm, it can dynamically adjust the gray distribution according to the local characteristics of the image, enhancing the image contrast and sharpness; Noise reduction processing module: Using advanced noise reduction algorithms based on wavelet transform, etc., it accurately identifies and removes detector noise, environmental interference noise, etc. in the X-ray images of the compression defects of strain clamps, while retaining the effective information of the images; Image segmentation module: Using semantic segmentation algorithms based on deep learning to accurately segment the X-ray images of the compression defects of strain clamps after enhancement and noise reduction processing; Result output and storage module: Diversely outputs and stores the X-ray images of the compression defects of strain clamps after intelligent processing, can be connected to a high-resolution display, stored in a large-capacity database and managed according to specific classification and indexing rules, and supports output in industrial common formats; The output end of the detector hardware module is connected to the analog-to-digital conversion module, the output end of the analog-to-digital conversion module is sequentially connected to the data cache module, the high-speed transmission module, and the image format conversion module, the output end of the image format conversion module is connected to the image enhancement module and the noise reduction processing module in parallel, the output ends of the image enhancement module and the noise reduction processing module are jointly connected to the image segmentation module, and the output end of the image segmentation module is connected to the result output and storage module, forming a linear and parallel hybrid data processing link.

2. The preprocessing system for X-ray image information of the compression defect of the strain clamp according to claim 1, characterized in that: The high-speed transmission module includes: Dynamic protocol switching unit: Automatically selects the transmission protocol according to the application scenario requirements, and uses the UDP protocol combined with forward error correction coding for real-time diagnosis scenarios to reduce latency, and uses the TCP protocol combined with data compression for offline analysis scenarios to ensure integrity; Network status monitoring unit: Real-time evaluates the current network bandwidth and packet loss rate, and dynamically adjusts the data shard size and retransmission strategy.

3. The preprocessing system for X-ray image information of the compression defect of a strain clamp according to claim 1, characterized in that: The image segmentation module includes: Interactive correction interface: Receives the user's annotation correction instructions for the segmentation results, fine-tunes the output of the deep learning model in real-time, and feeds the correction results back to the model training process; Domain adaptation unit: Achieves cross-scenario model migration by sharing the underlying feature extraction network of different scenarios and combining with a domain-specific adaptation layer.

4. A preprocessing system for X-ray image information of compression defects of strain clamps according to claim 1, characterized in that: The digital image data converted by the analog-to-digital conversion module maximally restores the original X-ray image information and reduces data distortion.

5. The preprocessing system for X-ray image information of the compression defect of a strain clamp according to claim 1, characterized in that: The high-speed network communication technologies adopted by the high-speed transmission module include optical fiber, 5G, or Wi-Fi6.

6. The preprocessing system for X-ray image information of the compression defect of a strain clamp according to claim 1, characterized in that: The image processed by the image enhancement module can effectively highlight the subtle defects in the detection image.

7. A preprocessing system for X-ray image information of compression defects of strain clamps according to claim 1, characterized in that: An improved U-Net model with an encoder-decoder structure, where the encoder part extracts multi-scale features through successive downsampling layers, and the decoder part implements feature map upsampling through transposed convolution. Among them, skip connections concatenate the feature maps of the j-th layer of the encoder with the feature maps of the corresponding layer of the decoder along the channel dimension, which is expressed as: where σ is the ReLU activation function, W j and b j are learnable parameters, and ∥ represents channel concatenation; Loss function design: Use cross-entropy loss L ce and Dice loss L dice for weighted combination: L = αL ce +(1 - α)L dice ; Where α∈[0,1] is the balance weight, and the Dice loss is defined for the problem of unbalanced target area ratio in industrial scenarios as: p i is the predicted probability of pixel i, g i is the true label, and ε is the smoothing term.

8. The preprocessing system for X-ray image information of the compression defect of the strain clamp according to claim 1, characterized in that: The image processed by the noise reduction processing module improves the signal-to-noise ratio and provides cleaner image data for subsequent image analysis and diagnosis.

9. The preprocessing system for X-ray image information of the compression defect of a strain clamp according to claim 1, wherein: The image segmentation module can clearly divide different structural parts of metal components and the areas where defects are located in industrial inspections, and the result output and storage module realizes seamless integration with other industrial inspection systems.

10. A preprocessing system for X-ray image information of compression defects of strain clamps, according to any one of claims 1-9, characterized in that: Including the following steps: S1. Signal acquisition step: Uses a high-precision X-ray detector to capture the absorption differences of X-rays by substances with different densities, and converts the X-ray signal into an analog electrical signal; S2. Analog-to-digital conversion step: Converts the analog electrical signal into a digital signal through high sampling rate and high bit quantization; S3. Data caching step: Temporarily stores digital image data when the acquisition rate is higher than the transmission rate, and performs data verification; S4. High-speed transmission step: Transmits data using a dynamic protocol switching strategy, using the UDP protocol and forward error correction coding for real-time diagnosis scenarios, and using the TCP protocol and data compression for offline analysis scenarios; S5. Format conversion step: Uniformly converts image data from different sources into a standard format, and performs gray normalization and size adjustment; S6. Image enhancement step: Dynamically adjusts the local gray distribution of the image through an adaptive histogram equalization algorithm; S7. Noise reduction processing step: Removes detector noise and environmental noise based on wavelet transform and retains valid information; S8. Image segmentation step: Performs semantic segmentation using a deep learning model and receives user interaction instructions to correct the segmentation results in real-time; S9. Output storage step: Outputs the processing results according to the DICOM standard and stores them based on data value classification.

Images