Fuel assembly side orientation and character intelligent identification method and system

Abstract

The invention belongs to the technical field of fuel assembly orientation and character recognition, and discloses a fuel assembly side orientation and character intelligent recognition method and system. The method comprises the following steps: triggering a side orientation detection state and a side character detection state, judging whether a classification result is within a limited range according to a result predicted by a side orientation classification model, turning over the side orientation detection state and storing side orientation detection information if the classification result is within the limited range, and determining whether the side orientation detection information is within the limited range according to the number of side character positioning frames. And judging whether the number of the side character positioning frames is equal to a set threshold value or not, turning over the side character detection state and carrying out a subsequent detection process. The system comprises a fuel assembly feature positioning module, a side orientation detection module and a side character detection module. According to the invention, side orientation and side character recognition of the fuel assembly can be completed with high accuracy and high speed, and an operation program of on-site intelligent recording of appearance inspection of four sides of the fuel assembly is assisted.

Description

Technical Field

[0001] This application belongs to the field of fuel assembly orientation and character recognition technology, and particularly relates to a method and system for intelligent recognition of the side orientation and characters of a fuel assembly. Background Technology

[0002] The functions of fuel assembly side orientation identification include the following three points:

[0003] (1) During the routine unloading of a PWR (pressurized water reactor), the four-sided fuel inspection device can record video of all four sides of each fuel assembly that passes by. The side orientation identification of the fuel assembly can verify the integrity of the fuel assembly without affecting the critical path, which helps to ensure the safe operation of the nuclear reactor.

[0004] (2) When conducting extreme accident safety analysis on fuel assemblies, it is necessary to understand their side orientation in order to assess the grid strength and guide tube stress of the fuel assembly and ensure that the relevant specifications are met.

[0005] (3) Understanding the side orientation of fuel assemblies helps to apply non-destructive testing technology, acoustic emission technology and other methods to monitor and inspect the pressurized water reactor fuel assembly grid in real time, so as to detect abnormalities in advance and take corresponding repair and improvement measures.

[0006] The functions of character recognition on the side of the fuel assembly include the following four points:

[0007] (1) The side characters of the fuel assembly may contain information about its manufacturing, batch and specific identification, which is crucial for tracking and managing the fuel assembly. Identifying and recording these characters helps to quickly trace the specific manufacturing batch and history when a defect is found, so that appropriate safety measures can be taken.

[0008] (2) Understanding the side characters of the fuel assembly is crucial for accident analysis and subsequent preventative measures. These characters may contain key information about the fuel assembly design and materials, which can help analyze the causes of the accident and develop preventative measures.

[0009] (3) Nuclear fuel assemblies require maintenance or replacement after a certain period of use. The information in the side characters helps determine which components need to be replaced first, as well as their replacement history, which is crucial for ensuring the stable operation of the nuclear reactor and extending its service life.

[0010] (4) The nuclear energy industry is subject to strict regulations, and recording and tracking all relevant information about fuel assemblies is part of regulatory compliance. Side character records help ensure that the operation of nuclear facilities complies with relevant regulatory requirements and provide necessary documentation support when needed.

[0011] In summary, the side orientation and character recognition of fuel assemblies are of great significance for ensuring the safe operation, quality control, accident analysis, maintenance management, and regulatory compliance of nuclear reactors.

[0012] Currently, the conventional visual inspection method for the side orientation and characters of fuel assemblies is not only inaccurate but also easily affected by background interference. In underwater irradiation environments, thermal turbulence severely degrades video quality and interferes with visual inspection of fuel assemblies. This turbulence can cause image blurring, affecting character clarity. Furthermore, the irradiation environment itself can impact image quality, leading to image degradation, including blurring and noise. Due to the limited attention span of workers, the side orientation and characters of fuel assemblies cannot be accurately determined.

[0013] While existing character detection technologies exist, they rely on directly detecting characters on the fuel assembly. Typically, a fuel assembly has four positions, with characters present in one of them. If the position where the character is present isn't determined first, character detection would need to be performed on all four positions, impacting both efficiency and accuracy. Summary of the Invention

[0014] The purpose of this application is to overcome the shortcomings of the prior art and provide a method and system for intelligent recognition of the side orientation and characters of fuel assemblies, so as to achieve efficient detection of the side orientation and side characters of fuel assemblies and assist in the on-site intelligent recording of the operation procedure for the four-sided appearance inspection of fuel assemblies.

[0015] To achieve the above objectives, this application provides the following technical solution:

[0016] In a first aspect, this application provides a method for intelligent recognition of the side orientation and characters of a fuel assembly, including:

[0017] S10: Receive video of the appearance of the fuel assembly from all four sides and extract images at set intervals;

[0018] S20: Preprocess the extracted image;

[0019] S30: Input the preprocessed image and predict the location of features in the image;

[0020] S40: When the predicted feature is the upper tube seat, determine whether the y coordinate of the lower right corner of the upper tube seat's positioning frame reaches the threshold. If it does, trigger the side orientation detection state.

[0021] S50: Based on the upper tube seat positioning box, extract the region of interest for side orientation, and use the side orientation classification model to predict the side orientation;

[0022] S60: Based on the prediction results of the side orientation classification model, determine whether the classification result is within the specified range. If it is not within the specified range, wait for the next detection image. If it is within the specified range, flip the side orientation detection state and save the side orientation detection information.

[0023] S70: When the predicted feature is the upper tube seat, determine whether the y coordinate of the lower right corner of the positioning box of the upper tube seat reaches the threshold. If it does, trigger the side character detection state.

[0024] S80: Based on the positioning frame of the upper tube seat, extract the region of interest for the side characters and predict the position of the side characters;

[0025] S90: Based on the number of side character positioning boxes, determine whether the number of side character positioning boxes is equal to the set threshold. If it is not equal to the threshold, wait for the next detection image. If it is equal to a certain threshold, flip the side character detection state and proceed with the subsequent detection process.

[0026] S100: Sort according to the position of the side character positioning box and extract the target area of ​​the side character;

[0027] S110: Based on the target region of the lateral character, predict the classification result of the lateral character;

[0028] S120: Combine side characters based on the predicted classification of the target region of the side characters;

[0029] S130: Display and save information about side character detection.

[0030] According to one embodiment of this application, one frame of image is extracted every N frames.

[0031] According to one embodiment of this application, the preprocessing includes:

[0032] Given an extracted image, delineate the regions that do not need to be detected.

[0033] Automatically extract the coordinates of areas that do not need to be detected using existing algorithms;

[0034] In subsequent video images, image regions that do not need to be detected are automatically cropped using existing algorithms.

[0035] According to one embodiment of this application, a fuel assembly feature localization model is used to predict feature locations in an image.

[0036] According to one embodiment of this application, training a fuel assembly feature localization model includes:

[0037] Using videos from multiple nuclear power plant overhauls, a number of images were extracted and labeled with the upper tube seat, the outer strip of the grid, and the lower tube seat to construct a fuel assembly feature localization dataset.

[0038] A fuel component feature localization model was trained and tested using a deep learning framework and a fuel component feature dataset.

[0039] According to one embodiment of this application, the fuel assembly features include an upper tube seat, an outer grid strip, and a lower tube seat.

[0040] According to one embodiment of this application, the threshold in S40 is 0.95 times the height of the preprocessed image.

[0041] According to one embodiment of this application, a side orientation classification model is used to predict the side orientation.

[0042] According to one embodiment of this application, training a side orientation classification model includes:

[0043] Using videos from multiple nuclear power plant overhauls, a number of side azimuth images covering the upper tube seat were extracted and classified to construct a fuel assembly side azimuth classification dataset.

[0044] A deep learning framework and a fuel assembly side orientation classification dataset were used to train and test a fuel assembly side orientation classification model.

[0045] According to one embodiment of this application, the threshold in S70 is 0.8 times the height of the preprocessed image.

[0046] According to one embodiment of this application, a side character positioning model is used to predict the position of side characters.

[0047] According to one embodiment of this application, training a side character localization model includes:

[0048] Extract a number of side images of the upper tube seat with characters, and manually label the character positions to construct a fuel assembly side character localization dataset.

[0049] A fuel component side character localization model was trained and tested using a deep learning framework and a fuel component side character localization dataset.

[0050] According to one embodiment of this application, the sorting method is from left to right.

[0051] According to one embodiment of this application, a side character classification model is used to predict the classification result of side characters.

[0052] According to one embodiment of this application, training a side character classification model includes:

[0053] Extract a number of side images of the upper tube seat with characters, label the character positions, extract the character images, classify the character images, and construct a fuel assembly side character classification dataset.

[0054] A fuel component side character classification model was trained and tested using a deep learning framework and a fuel component side character classification dataset.

[0055] Secondly, this application provides a fuel assembly side orientation and character intelligent recognition system, comprising:

[0056] The fuel assembly feature localization module is used to extract images from the video of the appearance of the four sides of the fuel assembly, preprocess the images, and predict the feature locations in the images.

[0057] The side orientation detection module is used to detect the side orientation of the fuel assembly, predict the side orientation, judge the classification result predicted by the side orientation classification model, and flip the side orientation detection state.

[0058] The side character detection module is used to detect side characters on the fuel assembly, predict the position of the side characters, determine the number of side character positioning boxes, extract the target area of ​​the side characters, and combine the side characters based on the classification results.

[0059] According to one embodiment of this application, the fuel assembly feature localization module includes:

[0060] The video input unit is used to input videos of the appearance of the fuel assembly from all four sides.

[0061] The image extraction unit is used to extract the video of the appearance of the fuel assembly on all four sides input by the video input unit. One frame of image is extracted every N frames.

[0062] The image preprocessing unit is used to preprocess the image extracted by the image extraction unit and crop out the undetected areas in the image;

[0063] The fuel assembly feature localization unit is used to predict the feature locations in the image after the image preprocessing unit has processed it.

[0064] According to one embodiment of this application, the fuel assembly feature localization unit predicts the feature positions in the image processed by the prediction image preprocessing unit using a trained fuel assembly feature localization model.

[0065] According to one embodiment of this application, the side orientation detection module includes:

[0066] The side orientation detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. Once the threshold is reached, the side orientation detection state is triggered.

[0067] The side orientation region extraction and prediction unit is used to extract the side orientation region of interest based on the upper tube seat positioning frame and predict the side orientation through the side orientation classification model.

[0068] The side orientation classification judgment unit is used to determine whether the classification result is within the specified range based on the prediction result of the side orientation classification model.

[0069] The side orientation storage unit is used to store side orientation detection information.

[0070] According to one embodiment of this application, when the side orientation classification judgment unit determines that the classification result is within a limited range, the side orientation detection state is flipped and the side orientation detection information is saved; otherwise, the system waits for the next detection image.

[0071] According to one embodiment of this application, the side character detection module includes:

[0072] The side character detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. If it does, the side character detection state is triggered.

[0073] The side character region extraction and prediction unit is used to extract the region of interest of the side characters based on the positioning box of the upper tube seat, and predict the position of the side characters through the side character positioning model.

[0074] The side character sorting unit is used to sort characters according to their positions and extract the target area of ​​the side characters.

[0075] The side character classification judgment unit is used to determine whether the number of side character positioning boxes is equal to a set threshold based on the number of side character positioning boxes.

[0076] The side character classification prediction unit is used to predict the classification result of the side character based on the target region of the side character using the side character classification model;

[0077] The lateral character combination unit is used to combine lateral characters based on the predicted classification of the target region of the lateral characters.

[0078] The side character storage unit is used to display and store information about side character detection.

[0079] According to one embodiment of this application, in the side character classification judgment unit, when the number of side character positioning boxes is equal to a threshold, the side character detection state is flipped and subsequent detection processes are performed; otherwise, the system waits for the next detection image.

[0080] Thirdly, this application provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the intelligent recognition method for the side orientation and characters of the fuel assembly.

[0081] Fourthly, this application provides a computer-readable storage medium storing computer-readable instructions, which, when executed, implement the aforementioned method for intelligent recognition of the side orientation and characters of the fuel assembly.

[0082] Compared with the prior art, the intelligent recognition method and system for the side orientation and characters of fuel assemblies provided in this application have the following advantages:

[0083] This application can simultaneously support four video feeds for side orientation detection of fuel assembly, significantly improving detection efficiency. The application first determines the side where the character is located through side orientation detection, then further narrows down the character detection area through fuel assembly feature localization, followed by locating the position of individual characters, and finally classifying the characters. Compared to directly inspecting the side orientation and characters of the fuel assembly with the naked eye, this method is more intelligent and improves detection efficiency.

[0084] This application can complete the side orientation and side character recognition of fuel assemblies with high accuracy and high speed, and assist in the on-site intelligent recording of the operation procedure for inspecting the appearance of fuel assemblies on all four sides.

[0085] The deep learning-based side orientation and side character detection method provided in this application can assist on-site operations more accurately and efficiently. Attached Figure Description

[0086] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the technical description will be briefly introduced below.

[0087] Figure 1 A flowchart of the intelligent recognition method for the side orientation and characters of the fuel assembly provided in this application;

[0088] Figure 2 A structural diagram of the intelligent recognition method for the side orientation and characters of the fuel assembly provided in this application;

[0089] Figure 3 The side orientation recognition effect diagram provided in this application;

[0090] Figure 4 The side orientation recognition effect diagram provided in this application;

[0091] Figure 5 The side-view three-dimensional recognition effect diagram provided in this application;

[0092] Figure 6 The side orientation four-way recognition effect diagram provided for this application;

[0093] Figure 7 The side character recognition effect diagram provided for this application;

[0094] Figure 8 A schematic diagram of the structure of the fuel assembly side orientation and character intelligent recognition system provided in this application. Detailed Implementation

[0095] The following detailed description provides further details on specific implementation methods.

[0096] like Figures 1 to 7 As shown, this application provides a method for intelligent recognition of the side orientation and characters of a fuel assembly, including steps S10 to S130, the specific steps of which are as follows:

[0097] S10: Receive video footage of the fuel assembly's exterior from all four sides. Extract one frame from every N frames. The value of N can be determined based on the actual situation; N is a positive integer.

[0098] S20: Preprocess the extracted image and crop out the undetected areas in the image.

[0099] In one embodiment, preprocessing includes the following steps:

[0100] S21: Input an extracted image and manually draw out the areas that do not need to be detected using specialized software or tools;

[0101] S22: Automatically extract the coordinates of areas that do not need to be detected using existing algorithms;

[0102] S23: In subsequent video images, image regions that do not need to be detected are automatically cropped using existing algorithms.

[0103] S30: Using the fuel assembly feature localization model, input the preprocessed image and predict the location of features in the image. Fuel assembly features include the upper tube seat, the outer strip of the grid, and the lower tube seat.

[0104] In one embodiment, training the fuel component feature localization model includes:

[0105] S31: Using videos from multiple nuclear power plant overhauls, extract an appropriate number of images (e.g., 5000 images), label the upper tube seat, the outer strip of the grid, and the lower tube seat, and construct a fuel assembly feature localization dataset;

[0106] S32: Train and test a fuel component feature localization model using a deep learning framework and a fuel component feature dataset.

[0107] S40: When the predicted feature is the upper tube seat, determine whether the y-coordinate of the lower right corner of the upper tube seat's bounding box reaches a certain threshold. If it does, trigger the side orientation detection state. It should be noted that this threshold can be set according to the actual situation. For example, the threshold can be 0.95 times the height of the preprocessed image.

[0108] S50: Based on the upper tube seat positioning frame, extract the region of interest for side orientation, and use the side orientation classification model to predict the side orientation.

[0109] In one embodiment, training the side orientation classification model includes:

[0110] S51: Using videos from multiple nuclear power plant overhauls, extract an appropriate number (e.g., 5000) of side orientation images covering the upper tube seat, and manually classify them to construct a fuel assembly side orientation classification dataset.

[0111] S52: Train and test a fuel assembly side orientation classification model using a deep learning framework and a fuel assembly side orientation classification dataset.

[0112] S60: Based on the prediction results of the lateral orientation classification model, determine whether the classification result falls within a certain limited range. For example, if the classification result of the lateral orientation classification model falls within [1,2,3,4], and there are four sides, classify accordingly. If it is not within the limited range, wait for the next detection image; if it is within the limited range, flip the lateral orientation detection state and save the lateral orientation detection information.

[0113] S70: When the predicted feature is the upper tube seat, determine the lower right corner of the upper tube seat's positioning frame (e.g., Figure 7 The system checks whether the y-coordinate of the image (as shown) reaches a certain threshold. If it does, the side character detection state is triggered. It should be noted that this threshold can be set according to the actual situation; for example, the threshold can be 0.8 times the height of the preprocessed image.

[0114] S80: Based on the upper tube seat positioning box, extract the region of interest for the side characters, and use the side character positioning model to predict the position of the side characters.

[0115] In one embodiment, training the side character localization model includes:

[0116] S81: Extract an appropriate number (e.g., 5000) of side images of characters on the upper tube seat, manually label the character positions, and construct a fuel assembly side character localization dataset.

[0117] S82: Train and test the fuel component side character localization model using a deep learning framework and the fuel component side character localization dataset.

[0118] S90: Based on the number of side character bounding boxes, determine whether the number of side character bounding boxes is equal to a certain threshold. If it is not equal to the threshold, wait for the next detection image; if it is equal to the threshold, flip the side character detection state and proceed with the subsequent detection process.

[0119] S100: Sort the characters from left to right according to their positions and extract the target area of ​​the characters.

[0120] S110: Based on the target region of the side character, use the side character classification model to predict the classification result of the side character.

[0121] In one embodiment, training the lateral character classification model includes:

[0122] S111: Extract a suitable amount (e.g., 5000 images) of side images with characters on the upper tube seat, manually label the character positions, extract the character images, and manually classify the character images to construct a fuel assembly side character classification dataset.

[0123] S112: Train and test the fuel component side character classification model using a deep learning framework and a fuel component side character classification dataset.

[0124] S120: Based on the predicted classification of the target area of ​​the side characters, combine the side characters. A complete side number consists of 6 characters, and it is necessary to combine the individual predicted characters.

[0125] S130: Display and save information about side character detection.

[0126] like Figure 8 As shown, this application provides a fuel assembly side orientation and character intelligent recognition system based on the above-mentioned fuel assembly side orientation and character intelligent recognition method, comprising:

[0127] The fuel assembly feature localization module is used to extract images of the four sides of the fuel assembly from video, preprocess the images, and predict the feature locations in the images using the fuel assembly feature localization model.

[0128] The side orientation detection module is used to detect the side orientation of the fuel assembly, predict the side orientation through a side orientation classification model, judge the classification result predicted by the side orientation classification model, flip the side orientation detection state, and save the side orientation detection information.

[0129] The side character detection module is used to detect side characters of the fuel assembly. It predicts the position of the side characters using a side character localization model, determines the number of side character localization boxes, extracts the target area of ​​the side characters, predicts the classification result of the side characters using a side character classification model, combines the side characters, and saves the detection results.

[0130] In one embodiment, the fuel assembly feature localization module includes:

[0131] The video input unit is used to input videos of the appearance of the fuel assembly from all four sides.

[0132] The image extraction unit is used to extract the video of the appearance of the fuel assembly on all four sides input by the video input unit. One frame of image is extracted every N frames.

[0133] The image preprocessing unit is used to preprocess the image extracted by the image extraction unit and crop out the undetected areas in the image;

[0134] The fuel assembly feature localization unit is used to predict the location of features in the image after processing by the image preprocessing unit. The fuel assembly feature localization unit predicts the location of features in the image after processing by the image preprocessing unit using a trained fuel assembly feature localization model. Fuel assembly features include the upper tube seat, the outer strip of the grid, and the lower tube seat.

[0135] In one embodiment, the side orientation detection module includes:

[0136] The side orientation detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. Once the threshold is reached, the side orientation detection state is triggered.

[0137] The side orientation region extraction and prediction unit is used to extract the side orientation region of interest based on the upper tube seat positioning frame and predict the side orientation through the side orientation classification model.

[0138] The side orientation classification judgment unit is used to determine whether the classification result is within a certain limited range based on the prediction result of the side orientation classification model. If it is not within the limited range, it waits for the next detection image. If it is within the limited range, it flips the side orientation detection state and saves the side orientation detection information.

[0139] The side orientation storage unit is used to store side orientation detection information.

[0140] In one embodiment, the side character detection module includes:

[0141] The side character detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. If it does, the side character detection state is triggered.

[0142] The side character region extraction and prediction unit is used to extract the region of interest of the side characters based on the positioning box of the upper tube seat, and predict the position of the side characters through the side character positioning model.

[0143] The side character sorting unit is used to sort the side characters from left to right according to their positions and extract the target area of ​​the side characters.

[0144] The side character classification and judgment unit is used to determine whether the number of side character positioning boxes is equal to a certain threshold. If it is not equal to the threshold, it waits for the next detection image. If it is equal to the threshold, it flips the side character detection state and proceeds with the subsequent detection process.

[0145] The side character classification prediction unit is used to predict the classification result of the side character based on the target region of the side character using the side character classification model;

[0146] The lateral character combination unit is used to combine lateral characters based on the predicted classification of the target region of the lateral characters.

[0147] The side character storage unit is used to display and store information about side character detection.

[0148] Therefore, this application achieves efficient detection of the side orientation and side characters of the fuel assembly by inputting a video of the appearance inspection of the four sides of the fuel assembly and using techniques such as image preprocessing, image enhancement, deep learning, side orientation judgment, and side character detection.

[0149] This application can complete the side orientation and side character recognition of fuel assemblies with high accuracy and high speed, and assist in the on-site intelligent recording of the operation procedure for inspecting the appearance of fuel assemblies on all four sides.

[0150] In addition, this application also provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the above-mentioned intelligent recognition method for the side orientation and characters of the fuel assembly.

[0151] In addition, this application also provides a computer-readable storage medium storing computer-readable instructions, which, when executed, implement the above-described method for intelligent recognition of the side orientation and characters of the fuel assembly.

[0152] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0153] The above description is only a specific embodiment of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application.

Claims

1. A method for intelligent recognition of the side orientation and characters of a fuel assembly, characterized in that, include: S10: Receive video of the appearance of the fuel assembly from all four sides and extract images at set intervals; S20: Preprocess the extracted image; S30: Input the preprocessed image and predict the location of features in the image; S40: When the predicted feature is the upper tube seat, determine whether the y coordinate of the lower right corner of the upper tube seat's positioning frame reaches the threshold. If it does, trigger the side orientation detection state. S50: Based on the upper tube seat positioning box, extract the region of interest for side orientation, and use the side orientation classification model to predict the side orientation; S60: Based on the prediction results of the side orientation classification model, determine whether the classification result is within the specified range. If it is not within the specified range, wait for the next detection image. If it is within the specified range, flip the side orientation detection state and save the side orientation detection information. S70: When the predicted feature is the upper tube seat, determine whether the y coordinate of the lower right corner of the positioning box of the upper tube seat reaches the threshold. If it does, trigger the side character detection state. S80: Based on the positioning frame of the upper tube seat, extract the region of interest for the side characters and predict the position of the side characters; S90: Based on the number of side character positioning boxes, determine whether the number of side character positioning boxes is equal to the set threshold. If it is not equal to the threshold, wait for the next detection image. If it is equal to a certain threshold, flip the side character detection state and proceed with the subsequent detection process. S100: Sort according to the position of the side character positioning box and extract the target area of ​​the side character; S110: Based on the target region of the lateral character, predict the classification result of the lateral character; S120: Combine side characters based on the predicted classification of the target region of the side characters; S130: Display and save information about side character detection.

2. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S10, one frame is extracted every N frames.

3. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S20, preprocessing includes: S21: Input an extracted image and delineate the regions that do not need to be detected; S22: Automatically extract the coordinates of areas that do not need to be detected using an algorithm; S23: In subsequent video images, the algorithm automatically crops out image regions that do not need to be detected.

4. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S30, a fuel assembly feature localization model is used to predict feature locations in the image.

5. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 4, characterized in that, The training of the fuel component feature localization model includes: S31: Using videos from multiple nuclear power plant overhauls, extract a number of images, label the upper tube seat, the outer strip of the grid, and the lower tube seat, and construct a fuel assembly feature localization dataset; S32: Train and test a fuel component feature localization model using a deep learning framework and a fuel component feature dataset.

6. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 4 or 5, characterized in that, The fuel assembly features include an upper tube seat, outer grid strips, and a lower tube seat.

7. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S40, the threshold is 0.95 times the height of the preprocessed image.

8. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S50, a side orientation classification model is used to predict the side orientation.

9. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 8, characterized in that, Training the side orientation classification model includes: S51: Using videos from multiple nuclear power plant overhauls, extract a number of side orientation images covering the upper tube seat, classify them, and construct a fuel assembly side orientation classification dataset. S52: Train and test a fuel assembly side orientation classification model using a deep learning framework and a fuel assembly side orientation classification dataset.

10. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S70, the threshold is 0.8 times the height of the preprocessed image.

11. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S80, the side character positioning model is used to predict the position of side characters.

12. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 11, characterized in that, Training the side character localization model includes: S81: Extract a number of side images of the upper tube seat with characters, manually label the character positions, and construct a fuel assembly side character localization dataset. S82: Train and test the fuel component side character localization model using a deep learning framework and the fuel component side character localization dataset.

13. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S100, the sorting method is from left to right.

14. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 1, characterized in that, In S110, a lateral character classification model is used to predict the classification results of lateral characters.

15. The intelligent recognition method for the side orientation and characters of fuel assembly according to claim 14, characterized in that, Training the lateral character classification model includes: S111: Extract a number of side images of the upper tube seat with characters, label the character positions, extract the character images, classify the character images, and construct a fuel assembly side character classification dataset. S112: Train and test the fuel component side character classification model using a deep learning framework and a fuel component side character classification dataset.

16. A fuel assembly side orientation and character intelligent recognition system, characterized in that, include: The fuel assembly feature localization module is used to extract images from the video of the appearance of the four sides of the fuel assembly, preprocess the images, and predict the feature locations in the images. The side orientation detection module is used to detect the side orientation of the fuel assembly, predict the side orientation, judge the classification result predicted by the side orientation classification model, and flip the side orientation detection state. The side character detection module is used to detect side characters on the fuel assembly, predict the position of the side characters, determine the number of side character positioning boxes, extract the target area of ​​the side characters, and combine the side characters based on the classification results.

17. The intelligent recognition system for the side orientation and characters of fuel assembly according to claim 16, characterized in that, The fuel assembly feature localization module includes: The video input unit is used to input videos of the appearance of the fuel assembly from all four sides. The image extraction unit is used to extract the video of the appearance of the fuel assembly on all four sides input by the video input unit. One frame of image is extracted every N frames. The image preprocessing unit is used to preprocess the image extracted by the image extraction unit and crop out the undetected areas in the image; The fuel assembly feature localization unit is used to predict the feature locations in the image after the image preprocessing unit has processed it.

18. The fuel assembly side orientation and character intelligent recognition system according to claim 17, characterized in that, The fuel assembly feature localization unit predicts the feature locations in the image processed by the prediction image preprocessing unit using a trained fuel assembly feature localization model.

19. The intelligent recognition system for the side orientation and characters of the fuel assembly according to claim 16, characterized in that, The side orientation detection module includes: The side orientation detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. Once the threshold is reached, the side orientation detection state is triggered. The side orientation region extraction and prediction unit is used to extract the side orientation region of interest based on the upper tube seat positioning frame and predict the side orientation through the side orientation classification model. The side orientation classification judgment unit is used to determine whether the classification result is within the specified range based on the prediction result of the side orientation classification model. The side orientation storage unit is used to store side orientation detection information.

20. The intelligent recognition system for the side orientation and characters of the fuel assembly according to claim 19, characterized in that, If the side orientation classification judgment unit determines that the classification result is within the specified range, it flips the side orientation detection state and saves the side orientation detection information; otherwise, it waits for the next detection image.

21. The intelligent recognition system for the side orientation and characters of fuel assembly according to claim 16, characterized in that, The side character detection module includes: The side character detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the set threshold. If it does, the side character detection state is triggered. The side character region extraction and prediction unit is used to extract the region of interest of the side characters based on the positioning box of the upper tube seat, and predict the position of the side characters through the side character positioning model. The side character sorting unit is used to sort characters according to their positions and extract the target area of ​​the side characters. The side character classification judgment unit is used to determine whether the number of side character positioning boxes is equal to a set threshold based on the number of side character positioning boxes. The side character classification prediction unit is used to predict the classification result of the side character based on the target region of the side character using the side character classification model; The lateral character combination unit is used to combine lateral characters based on the predicted classification of the target region of the lateral characters. The side character storage unit is used to display and store information about side character detection.

22. The intelligent recognition system for the side orientation and characters of the fuel assembly according to claim 21, characterized in that, In the side character classification and judgment unit, if the number of side character positioning boxes is equal to the threshold, the side character detection state is flipped and the subsequent detection process is carried out; otherwise, the system waits for the next detection image.

23. An electronic device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor, when executing the computer-readable instructions, implements the fuel assembly side orientation and character intelligent recognition method as described in any one of claims 1-15.

24. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed, implement the fuel assembly side orientation and character intelligent recognition method as described in any one of claims 1-15.

Images