Real-time imaging detection system based on shell ring welding seam

Abstract

The invention discloses a real-time imaging detection system based on a shell ring welding seam, and relates to the technical field of shell ring welding seam detection, the system comprises an imaging acquisition module, and a high-resolution imaging device is used to carry out multi-dimensional and omnibearing image acquisition on the shell ring welding seam. According to the invention, clear, accurate and complete welding seam image information can be obtained through the imaging acquisition module by adopting high-resolution imaging equipment, the accuracy and comprehensiveness of detection are remarkably improved, the data transmission module can quickly transmit a large amount of image data to the image processing module for quick processing, the detection period is greatly shortened, and the detection efficiency is improved. According to the method, the detection efficiency is remarkably improved, the rapid detection requirement in large-scale production is met, through the machine learning algorithm module and through learning and training of a large amount of training data, welding seam features can be automatically and accurately extracted, defects can be automatically and accurately recognized, dependence on artificial experience is reduced, and the detection reliability and stability are improved.

Description

Technical Field

[0001] The invention relates to the technical field of cylinder joint weld detection, in particular to a real-time imaging detection system based on cylinder joint welds. Background Art

[0002] The cylinder section is one of the most important pressure-bearing components of the pressure vessel. Its main function is to provide the pressure-bearing space required by the process. During the manufacturing process, different manufacturing methods may be used according to the diameter and length of the cylinder section. When the diameter is large, it may be necessary to roll the steel plate into a cylinder on a plate rolling machine or press the steel plate into two semi-cylinders on a hydraulic press, and then weld the two together to form a whole cylinder. The quality of the cylinder section weld plays a vital role in the safety, reliability and service life of the equipment. Some defects will inevitably occur during the welding process, such as pores, slag inclusions, cracks, etc. These defects will seriously affect the strength and sealing of the weld. Through detection, these defects can be discovered and repaired in time to ensure the quality of the weld.

[0003] Traditional weld inspection methods mostly rely on manual visual inspection or simple inspection equipment. Manual visual inspection is a common method that is extremely dependent on operator experience and subjective judgment. It is not only inefficient, but also prone to missed detection and misjudgment. Although simple inspection equipment has improved the objectivity of inspection to a certain extent, they can often only detect specific types of defects and are difficult to effectively identify complex and tiny defects. As a result, traditional weld inspection methods have the problems of low efficiency, difficulty in ensuring accuracy, and inability to effectively detect complex defects. Therefore, it is necessary to propose a real-time imaging inspection system based on barrel welds to solve the problems in the prior art. Summary of the invention

[0004] The purpose of the present invention is to make up for the shortcomings of the prior art and to provide a real-time imaging detection system based on cylinder section welds, which can quickly and efficiently process and accurately analyze a large amount of cylinder section weld imaging data, significantly improve the detection efficiency, and accurately identify defects, providing a reliable and efficient solution for cylinder section weld quality inspection.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a real-time imaging detection system based on cylinder section welds, the system comprising: The imaging acquisition module uses high-resolution imaging equipment to perform multi-dimensional and all-round image acquisition of the cylinder section weld; The lighting and auxiliary unit includes an auxiliary positioning device and is equipped with specially designed lighting equipment to provide uniform, stable and suitable light conditions for imaging acquisition; The data transmission module is used to transmit the collected image data to the image processing module in real time and stably, including high-speed data lines, wireless network transmission modules and optical fiber transmission channels to adapt to different detection scenarios and environmental requirements; Image processing module, used to pre-process the transmitted imaging data to provide a high-quality data basis for subsequent analysis; The machine learning algorithm module is used to learn and train the annotated weld image data, automatically extract weld features, and classify and identify defects. It can accurately judge and locate complex and subtle weld defects. Storage module, used to store the collected original image data, intermediate data in the processing process, and the final detection results and analysis reports; The control and interaction module allows users to set detection parameters, start and stop the detection process, and provide real-time detection status feedback; The result output and display module is used to present the test results to the user in an intuitive and clear manner, including the type, location, and size information of the defects, and supports the generation of detailed test reports.

[0006] Furthermore, the specific execution steps of the imaging acquisition module include: Select appropriate high-resolution imaging equipment, including industrial cameras, X-ray imaging devices, and laser scanning equipment; Aim the imaging device at the barrel weld to cover the entire area that needs to be inspected; Start the imaging equipment to collect images of the cylinder section weld and obtain multi-dimensional and all-round image data.

[0007] Furthermore, the specific execution steps of the lighting and auxiliary unit include: Evaluate the required light conditions based on the inspection environment and weld characteristics; Turn on lighting equipment and adjust lighting parameters of lighting equipment, including brightness, angle and color temperature; Use auxiliary positioning devices to accurately align the imaging device to the weld position; During the detection process, lighting parameters and auxiliary positioning are adjusted dynamically in real time according to actual imaging effects and environmental changes.

[0008] Furthermore, the specific execution steps of the data transmission module include: After the imaging acquisition module obtains the image data of the barrel weld, it converts it into a digital signal; Encapsulate the compressed image data according to a specific protocol format, and add source address, destination address and data type information; According to the detection scenario and environmental requirements, select the data transmission method, including high-speed data lines, wireless network transmission modules and optical fiber transmission channels; The encapsulated data is sent out through the selected transmission path.

[0009] Furthermore, the specific execution steps of the image processing module include: Receive image data from the data transmission module, remove encapsulation information of the data, extract original image data, decompress the compressed image data, and restore the original image data; Use the Gaussian filter algorithm to remove noise from the image. The mathematical formula of Gaussian filter is: , in, is a Gaussian function, is the standard deviation, is the original image at coordinates The pixel value at is the filtered image at coordinates The pixel value at ; The contrast and brightness of the image are enhanced by the histogram equalization algorithm. The formula of histogram equalization is expressed as: in, is the grayscale of the input image, The gray level is The probability of is the grayscale of the output image; The Sobel operator algorithm is used to detect the edge of the image. and The templates for the directions are: , , in, and They are and The gradient of the direction, is the edge strength; The image features are extracted using the oriented gradient histogram feature extraction algorithm, and the processed image data is output to the machine learning algorithm module for subsequent defect classification and identification.

[0010] Furthermore, the specific execution steps of the machine learning algorithm module include: Receive image data processed by the efficient real-time image processing algorithm module, clean the data, and remove outliers and missing values; Extract meaningful features from image data, including texture features, shape features, and grayscale features; Divide the prepared data into training set, validation set and test set; The convolutional neural network (CNN) machine learning model is used for model training. The training process includes forward propagation and back propagation. The forward propagation calculates the output: in, is the input data, is the weight matrix, is the bias term, is the activation function, and back propagation uses the gradient descent algorithm to update the parameters and , to minimize the loss function , cross entropy loss function: .

[0011] Furthermore, the specific execution steps of the machine learning algorithm module also include: Adjust the model's hyperparameters, including learning rate, number of layers, and number of nodes, by running experiments on the validation set; The trained model is evaluated using the test set. The evaluation indicators include accuracy, recall, and F1 value. The accuracy calculation formula is: in, It is a real example. It is a true counterexample. It is a false positive example. is a false negative example, and the recall rate is calculated as: , the calculation formula of F1 value is: ; Based on the evaluation results, the model is further optimized and data enhancement operations are added; The new image data to be inspected is input into the optimized model to obtain the classification and recognition results of weld defects, including defect type, location and size information.

[0012] Furthermore, the specific execution steps of the storage module include: Receive data from the image processing module and the machine learning algorithm module, including original image data, intermediate data during processing, and final detection results and analysis reports; Classify the received data and assign corresponding storage addresses to different types of data according to the classification results; Write the classified data into the allocated storage address and create an index for the stored data; When there is new data or the existing data is modified, follow the above steps to update; According to the user's request, the required data can be quickly located and read through the index.

[0013] Furthermore, the specific execution steps of the control and interaction module include: Receive detection parameters input by the user, including parameter settings of the imaging device and parameters of the image processing algorithm; Use preset rules and ranges to check the validity of received parameters; According to the user's operation, the corresponding control instructions are sent. When controlling the execution order of the detection process, the finite state machine algorithm is used. The state transition formula is expressed as: in, is the current state, is the input control instruction, is the next state; Real-time monitoring of the operating status of the detection system, including the working status of the imaging acquisition module, image processing module and machine learning algorithm module; Convert the monitored status information into feedback information that the user can understand; Communicate with other external devices and systems to achieve data sharing and collaborative work. During the communication process, data encryption algorithms are used to ensure data security; Respond to user operations and inquiries in a timely manner and provide the required information and assistance.

[0014] Furthermore, the specific execution steps of the result output and display module include: Receive inspection result data from the machine learning algorithm module, including defect type, location, and size information; Format the received test result data to make it meet the specific display format requirements; Generate corresponding charts based on the test result data, including a bar chart of defect type distribution and a line chart of defect size. When generating the charts, a statistical calculation algorithm is used to calculate the formula for the average value: in, is each sample value, is the sample size; Generate detailed test reports based on formatted data and generated charts; Output the processed result data, synthesized images, generated charts and reports to the user through a display device; Adjust the displayed content in real time according to the user's operation.

[0015] Compared with the existing technology, the real-time imaging detection system based on cylinder section weld has the following beneficial effects: The present invention uses an imaging acquisition module and a high-resolution imaging device to obtain clear, accurate and complete weld image information, which significantly improves the accuracy and comprehensiveness of detection. The data transmission module can quickly transmit a large amount of image data to the image processing module for rapid processing, which greatly shortens the detection cycle, significantly improves the detection efficiency, and meets the rapid detection needs in large-scale production. Through the machine learning algorithm module, after learning and training with a large amount of training data, it can automatically and accurately extract weld features and identify defects, reducing dependence on manual experience and improving the reliability and stability of detection.

[0016] Other advantages, objectives and features of the present invention will be set forth in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 It is a structural schematic diagram of a real-time imaging detection system based on cylinder section weld. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. 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. Example

[0020] A real-time imaging detection system based on cylinder welds, the system includes: an imaging acquisition module, using high-resolution imaging equipment, including industrial cameras, X-ray imaging devices and laser scanning equipment, etc., to perform multi-dimensional and all-round image acquisition of cylinder welds to ensure the acquisition of clear, accurate and complete weld image information. The imaging acquisition module is the source of the entire detection system and is responsible for obtaining the original image data of the cylinder welds. It uses a variety of high-resolution imaging devices to adapt to different detection scenarios and requirements. For example, when detecting large cylinder welds, industrial cameras may be used, such as CCD cameras with a resolution of 50 million pixels, which can capture the subtle structure and details of the welds. For internal welds or parts that are difficult to observe directly, X-ray imaging devices can be used, which can penetrate the cylinder material and show the internal welds. For the detection of some surface features, laser scanning equipment can provide high-precision three-dimensional images, such as measuring the convex height and concave depth of the weld.

[0021] To ensure the quality and consistency of the image, the imaging acquisition module is also equipped with automatic focus and exposure control functions. When the light conditions of the detection environment change, the system can automatically adjust the camera's focus and exposure parameters to ensure that clear and accurate images are obtained.

[0022] The lighting and auxiliary unit includes an auxiliary positioning device and is equipped with specially designed lighting equipment to provide uniform, stable and appropriate lighting conditions for imaging acquisition, reduce the impact of shadows and reflections on image quality, and ensure that the imaging equipment is accurately aligned with the weld position. The lighting and auxiliary unit provides ideal conditions for imaging acquisition to ensure the quality and analyzability of the image. For example, in a darker detection environment, it will be equipped with a high-brightness LED lighting system, and its color temperature can be adjusted between 3000K and 6000K to adapt to different materials and surface characteristics. For weld surfaces with strong reflectivity, diffuse reflection lighting will be used to reduce the impact of reflections on the image. The auxiliary positioning device may include a laser locator that can accurately indicate the alignment position of the imaging equipment, and the deviation can be controlled within ±0.1 mm.

[0023] In addition, in order to cope with complex inspection scenarios, such as the curved surface of a cylinder section or welds at different angles, the lighting system can be adjusted and combined at multiple angles to ensure that all parts of the weld can receive uniform and adequate lighting.

[0024] The data transmission module is used to transmit the collected image data to the image processing module in real time and stably to ensure the integrity and timeliness of the data. It includes high-speed data lines, wireless network transmission modules and optical fiber transmission channels to adapt to different detection scenarios and environmental requirements. The data transmission module is responsible for quickly and stably transmitting the large amount of image data obtained by the imaging acquisition module to the processing unit. For example, in the detection of barrel welds on high-speed production lines, Gigabit Ethernet is used as the data transmission channel, which can achieve a data transmission rate of several GB per second to ensure real-time performance. For some mobile detection scenarios or environments with difficult wiring, 5G wireless network transmission modules are used to utilize their high speed and low latency characteristics to ensure timely data transmission. At the same time, in order to ensure the integrity of the data, error correction coding technology such as Reed-Solomon coding is used. Even if some data is lost or erroneous during the transmission process, the original data can be restored through error correction.

[0025] During the data transmission process, data compression will also be performed, such as using the JPEG2000 compression algorithm, which can reduce the data volume by more than 50% without significantly losing image quality, thereby improving transmission efficiency.

[0026] The image processing module is used to pre-process a large amount of transmitted imaging data in a very short time, including denoising, enhancement, and edge detection, to provide a high-quality data basis for subsequent analysis. It can process thousands of frames of images per second, greatly improving the detection speed. The processing and analysis unit is the core of the system, which efficiently processes and deeply analyzes the incoming image data. The efficient real-time image processing algorithm module can pre-process a large amount of image data in a very short time.

[0027] The machine learning algorithm module is used to learn and train a large amount of annotated weld image data, automatically extract weld features, and classify and identify defects. It can accurately judge and locate complex and subtle weld defects, such as tiny cracks and hidden pores. The machine learning algorithm module can automatically identify various complex weld defects by learning and training a large amount of annotated weld image data. Using the convolutional neural network (CNN) model, after being trained with tens of thousands of weld images with defect annotations, it can accurately distinguish between tiny cracks, pores, and incomplete fusion defects. For some irregularly shaped defects, such as long and thin cracks along the direction of the weld, accurate positioning and measurement can be achieved through the feature extraction and pattern recognition capabilities of the deep learning algorithm.

[0028] The storage module is used to store the collected original image data, the intermediate data in the processing process, and the final test results and analysis reports. It has large capacity, high read and write speed and reliable data storage capabilities to meet the needs of long-term test data accumulation and backtracking. The storage module is used to store various types of data in the test process to provide support for subsequent query, analysis and tracing. For example, a large-capacity enterprise-level solid-state drive (SSD) array is used to provide TB of storage space, which can store test data for months or even years. For the original image data, lossless compression format is used for storage to ensure data integrity. At the same time, a database management system is established to classify, store and index the test results, analysis reports and related parameter settings for easy and fast retrieval. During the data storage process, regular backups are also performed, such as backing up data to a tape library or cloud storage every day to prevent data loss. The control and interaction module, as the operation control center of the system, allows users to set detection parameters, start and stop the detection process, and provide real-time detection status feedback. It also supports communication and integration with other external devices and systems. The control and interaction module, as the operation center of the system, allows users to flexibly set detection parameters and monitor the detection process in real time. For example, users can set the resolution, frame rate, exposure time and other parameters of the imaging device through a graphical interface. During the detection process, the detection progress and the currently processed image are displayed in real time. At the same time, an early warning function is provided. When possible serious defects are detected, sound and light alarms are issued in time.

[0029] In order to facilitate integration with other production management systems, the control and interaction module provides standardized interfaces, such as the OPC UA protocol interface, which can seamlessly connect with the factory's automated production line to achieve sharing and collaborative processing of production data.

[0030] The result output and display module is used to present the test results to the user in an intuitive and clear manner, including the type, location, and size information of the defects, and supports the generation of detailed test reports. The result output and display module presents the test results to the user in an intuitive and clear manner. For example, weld defects are displayed in the form of color images, with red indicating serious defects, yellow indicating moderate defects, and green indicating minor defects. At the same time, a detailed text report is generated, including information such as the type, location, size, and evaluation level of the defects. For large-scale test data, data visualization technology can be used, such as drawing heat maps and trend maps of defect distribution, to help users quickly understand the overall status of the weld quality. In addition, it supports outputting the test results to a printer or saving them in the form of electronic documents, which is convenient for users to archive and share.

[0031] Through the detailed expansion and mutual correspondence of each module, a complete, efficient and accurate real-time imaging detection system based on cylinder section welds is formed. Example

[0032] Based on the first embodiment, this embodiment describes in detail the overall workflow of a real-time imaging detection system based on cylinder section welds.

[0033] First, the imaging acquisition module selects appropriate high-resolution imaging equipment, including industrial cameras, X-ray imaging devices and laser scanning equipment, and aims the imaging equipment at the cylinder section weld to cover all areas that need to be inspected. Then, the imaging equipment is started to capture images of the cylinder section weld to obtain multi-dimensional and comprehensive image data to ensure that the captured images are clear, accurate and complete, and can reflect the true condition of the weld.

[0034] Then, the lighting and auxiliary unit evaluates the required lighting conditions according to the inspection environment and the characteristics of the weld, turns on the lighting equipment, and adjusts the lighting parameters of the lighting equipment, including brightness, angle, and color temperature, to provide uniform, stable, and appropriate light. The imaging device is accurately aligned with the weld position using the auxiliary positioning device to ensure the accuracy of the imaging. During the inspection process, the lighting parameters and auxiliary positioning are dynamically adjusted in real time according to the actual imaging effect and environmental changes to continuously optimize the imaging effect.

[0035] After the imaging acquisition module obtains the image data of the barrel weld, it converts it into a digital signal. In this process, the JPEG image compression algorithm is used, and its formula is expressed as follows: in, represents the original image data, represents the quantization parameter, Represents compressed image data, encapsulates the compressed image data according to a specific protocol format, adds source address, destination address and data type information, and selects data transmission methods according to detection scenarios and environmental requirements, including high-speed data lines, wireless network transmission modules and optical fiber transmission channels. If wireless network transmission is used, routing selection algorithms may be involved. The Dijkstra algorithm is used to calculate the shortest path, and its formula can be expressed as: in, Represents the distance from the source node to the node The shortest path length, Represents a node, Represents a collection of nodes. Representation Node To Node The edge weight is used to send the encapsulated data through the selected transmission path.

[0036] Then the image processing module receives the image data from the data transmission module, removes the encapsulation information of the data, extracts the original image data, decompresses the compressed image data, restores the original image data, and uses the Gaussian filter algorithm to remove the noise in the image. The mathematical formula of the Gaussian filter is: , in, is a Gaussian function, is the standard deviation, is the original image at coordinates The pixel value at is the filtered image at coordinates The pixel value at is used to enhance the contrast and brightness of the image through the histogram equalization algorithm. The formula of histogram equalization is expressed as: in, is the grayscale of the input image, The gray level is The probability of is the gray level of the output image. The Sobel operator algorithm is used to detect the edge of the image. The Sobel operator is and The templates for the directions are: , , in, and They are and The gradient of the direction, It is the edge strength. The features of the image are extracted using the directional gradient histogram feature extraction algorithm, and the processed image data is output to the machine learning algorithm module for subsequent defect classification and identification.

[0037] Then the machine learning algorithm module receives the image data processed by the efficient real-time image processing algorithm module, cleans the data, removes outliers and missing values, and extracts meaningful features from the image data, including texture features, shape features, and grayscale features. For the extraction of texture features, the gray level co-occurrence matrix (GLCM) algorithm is used. The formula for calculating contrast in GLCM is: in, The gray value is and The probability of occurrence of pixel pairs is calculated, and the prepared data is divided into training set, validation set and test set, usually in a certain ratio, such as 7:2:1. The machine learning model of convolutional neural network CNN is used for model training. The training process includes forward propagation and back propagation. The forward propagation calculation output is: in, is the input data, is the weight matrix, is the bias term, is the activation function, and back propagation uses the gradient descent algorithm to update the parameters and , to minimize the loss function , cross entropy loss function: .

[0038] Then, by conducting experiments on the validation set, the model's hyperparameters, including learning rate, number of layers, and number of nodes, are adjusted. The trained model is evaluated using the test set. The evaluation indicators include accuracy, recall, and F1 value. The accuracy calculation formula is: in, It is a real example. It is a true counterexample. It is a false positive example. is a false negative example, and the recall rate is calculated as: , the calculation formula of F1 value is: ,According to the evaluation results, the model is further optimized, and data enhancement operations are added, such as random rotation, scaling, and flipping of images, to expand the data set. The new image data to be detected is input into the optimized model to obtain the classification and recognition results of weld defects, including defect type (cracks, pores, etc.), location and size information.

[0039] The storage module receives data from the image processing module and the machine learning algorithm module, including the original image data, the intermediate data in the processing process, and the final test results and analysis reports. It classifies the received data, for example, by data type (image, text, value, etc.), acquisition time, weld batch, etc. According to the classification results, it allocates corresponding storage addresses to different types of data. When allocating addresses, a hash algorithm may be used to improve the efficiency of address search. The hash function can be expressed as: in, is the key value of the data, and is a constant, is the size of the hash table. The classified data is written into the allocated storage address. During the writing process, a data compression algorithm may be used to save storage space, such as the LZ77 compression algorithm, and an index is established for the stored data for fast retrieval and query. The index can be implemented based on data structures such as B-tree or B+ tree. When there is new data and the existing data is modified, the update operation is performed according to the above steps. According to the user's request, the required data is quickly located and read through the index.

[0040] The control and interaction module receives the detection parameters input by the user, including the parameter settings of the imaging device and the parameters of the image processing algorithm, and uses the preset rules and ranges to perform a legality check on the received parameters. For example, for the exposure time parameter of the imaging device, check whether it is within the legal range. According to the user's operation, such as starting and stopping the detection process, the corresponding control instructions are sent. When controlling the execution order of the detection process, the finite state machine algorithm is used. The state transition formula is expressed as: in, is the current state, is the input control instruction, It is the next state, which monitors the operating status of the detection system in real time, including the working status of the imaging acquisition module, image processing module and machine learning algorithm module, and converts the monitored status information into feedback information that the user can understand, such as progress percentage, whether there is an abnormality, etc. It communicates with other external devices and systems to achieve data sharing and collaborative work. During the communication process, data encryption algorithms are used to ensure data security. For example, the AES encryption algorithm is used. The encryption formula can be expressed as: in, is the key, is plain text data, It is encrypted data, responding to user operations and inquiries in a timely manner, and providing the required information and assistance.

[0041] Finally, the result output and display module receives the test result data from the machine learning algorithm module, including the type, location and size information of the defect, and formats the received test result data to meet the specific display format requirements. For example, the position information is converted from coordinate form to an easy-to-understand text description. If the test result needs to be superimposed with the original image, an image synthesis algorithm is used, for example, an alpha blending algorithm is used, and the formula is: in, is the detection result image to be superimposed, is the original image, is the transparency parameter.

[0042] Generate corresponding charts based on the test result data, including a bar chart of defect type distribution and a line chart of defect size. When generating the charts, a statistical calculation algorithm is used to calculate the formula for the average value: in, is each sample value, It is the number of samples. A detailed test report is generated based on the formatted data and generated charts. Report generation may involve text typesetting and template filling algorithms. The processed result data, synthesized images, generated charts and reports are output to users through display devices (such as display screens, printers, etc.). The display content is adjusted in real time based on user operations such as zooming, scrolling, filtering, etc. When responding to interactive operations, a view transformation algorithm may be used to achieve effects such as image zooming and translation.

[0043] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the same elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A real-time imaging detection system based on cylinder section welds, characterized in that: The system includes: The imaging acquisition module uses high-resolution imaging equipment to perform multi-dimensional and all-round image acquisition of the cylinder section weld; The lighting and auxiliary unit includes an auxiliary positioning device and is equipped with specially designed lighting equipment to provide uniform, stable and suitable light conditions for imaging acquisition; The data transmission module is used to transmit the collected image data to the image processing module in real time and stably, including high-speed data lines, wireless network transmission modules and optical fiber transmission channels to adapt to different detection scenarios and environmental requirements; Image processing module, used to pre-process the transmitted imaging data to provide a high-quality data basis for subsequent analysis; The machine learning algorithm module is used to learn and train the annotated weld image data, automatically extract weld features, and classify and identify defects. It can accurately judge and locate complex and subtle weld defects. Storage module, used to store the collected original image data, intermediate data in the processing process, and the final detection results and analysis reports; The control and interaction module allows users to set detection parameters, start and stop the detection process, and provide real-time detection status feedback; The result output and display module is used to present the test results to the user in an intuitive and clear manner, including the type, location, and size information of the defects, and supports the generation of detailed test reports.

2. A real-time imaging detection system based on cylinder section weld according to claim 1, characterized in that: The specific execution steps of the imaging acquisition module include: Select appropriate high-resolution imaging equipment, including industrial cameras, X-ray imaging devices, and laser scanning equipment; Aim the imaging device at the barrel weld to cover the entire area that needs to be inspected; Start the imaging equipment to collect images of the cylinder section weld and obtain multi-dimensional and all-round image data.

3. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the lighting and auxiliary unit include: Evaluate the required light conditions based on the inspection environment and weld characteristics; Turn on lighting equipment and adjust lighting parameters of lighting equipment, including brightness, angle and color temperature; Use auxiliary positioning devices to accurately align the imaging device to the weld position; During the detection process, lighting parameters and auxiliary positioning are adjusted dynamically in real time according to actual imaging effects and environmental changes.

4. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the data transmission module include: After the imaging acquisition module obtains the image data of the barrel weld, it converts it into a digital signal; Encapsulate the compressed image data according to a specific protocol format, and add source address, destination address and data type information; According to the detection scenario and environmental requirements, select the data transmission method, including high-speed data lines, wireless network transmission modules and optical fiber transmission channels; The encapsulated data is sent out through the selected transmission path.

5. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the image processing module include: Receive image data from the data transmission module, remove encapsulation information of the data, extract original image data, decompress the compressed image data, and restore the original image data; Use the Gaussian filter algorithm to remove noise from the image. The mathematical formula of Gaussian filter is: , in, is a Gaussian function, is the standard deviation, is the original image at coordinates The pixel value at is the filtered image at coordinates The pixel value at ; The contrast and brightness of the image are enhanced by the histogram equalization algorithm. The formula of histogram equalization is expressed as: in, is the grayscale of the input image, The gray level is The probability of is the grayscale of the output image; The Sobel operator algorithm is used to detect the edge of the image. and The templates for the directions are: , , in, and They are and The gradient of the direction, is the edge strength; The image features are extracted using the oriented gradient histogram feature extraction algorithm, and the processed image data is output to the machine learning algorithm module for subsequent defect classification and identification.

6. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the machine learning algorithm module include: Receive image data processed by the efficient real-time image processing algorithm module, clean the data, and remove outliers and missing values; Extract meaningful features from image data, including texture features, shape features, and grayscale features; Divide the prepared data into training set, validation set and test set; The convolutional neural network (CNN) machine learning model is used for model training. The training process includes forward propagation and back propagation. The forward propagation calculates the output: in, is the input data, is the weight matrix, is the bias term, is the activation function, and back propagation uses the gradient descent algorithm to update the parameters and , to minimize the loss function , cross entropy loss function: .

7. A real-time imaging detection system based on cylinder section weld according to claim 6, characterized in that: The specific execution steps of the machine learning algorithm module also include: Adjust the model's hyperparameters, including learning rate, number of layers, and number of nodes, by running experiments on the validation set; Use the test set to evaluate the trained model. The evaluation indicators include accuracy, recall, and F1 value. The calculation formula for accuracy is: in, It is a real example. It is a true counterexample. It is a false positive example. is a false negative example, and the recall rate is calculated as: , the calculation formula of F1 value is: ; Based on the evaluation results, the model is further optimized and data enhancement operations are added; The new image data to be inspected is input into the optimized model to obtain the classification and recognition results of weld defects, including defect type, location and size information.

8. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the storage module include: Receive data from the image processing module and the machine learning algorithm module, including original image data, intermediate data during processing, and final detection results and analysis reports; Classify the received data and assign corresponding storage addresses to different types of data according to the classification results; Write the classified data into the allocated storage address and create an index for the stored data; When there is new data or the existing data is modified, follow the above steps to update; According to the user's request, the required data can be quickly located and read through the index.

9. The real-time imaging detection system based on cylinder section weld according to claim 1 is characterized in that: The specific execution steps of the control and interaction module include: Receive detection parameters input by the user, including parameter settings of the imaging device and parameters of the image processing algorithm; Use preset rules and ranges to check the validity of received parameters; According to the user's operation, the corresponding control instructions are sent. When controlling the execution order of the detection process, the finite state machine algorithm is used. The state transition formula is expressed as: in, is the current state, is the input control instruction, is the next state; Real-time monitoring of the operating status of the detection system, including the working status of the imaging acquisition module, image processing module and machine learning algorithm module; Convert the monitored status information into feedback information that users can understand; Communicate with other external devices and systems to achieve data sharing and collaborative work. During the communication process, data encryption algorithms are used to ensure data security; Respond to user operations and inquiries in a timely manner and provide the required information and assistance.

10. The real-time imaging detection system based on cylinder section weld according to claim 1, characterized in that: The specific execution steps of the result output and display module include: Receive inspection result data from the machine learning algorithm module, including defect type, location, and size information; Format the received test result data to make it conform to the specific display format requirements; Generate corresponding charts based on the test result data, including a bar chart of defect type distribution and a line chart of defect size. When generating the charts, a statistical calculation algorithm is used to calculate the formula for the average value: in, is each sample value, is the sample size; Generate detailed test reports based on formatted data and generated charts; Output the processed result data, synthesized images, generated charts and reports to the user through a display device; Adjust the displayed content in real time according to the user's operation.

Images