Methods, systems, equipment, and media for locating regions of interest in nuclear power fuel assemblies

Abstract

This invention relates to a method, system, device, and medium for locating regions of interest (ROIs) in nuclear power fuel assemblies. The method includes: acquiring images of the fuel assembly in an underwater environment as application images; performing feature matching between the application images and preset template images using a deep learning model; analyzing the feature matching results to obtain the optimal transformation matrix between the application images and the template images; determining the ROI on the template image; and generating a corresponding target ROI on the application image based on the optimal transformation matrix. This invention can pinpoint the target ROI on the application image, thereby suppressing false identification caused by non-ROI regions and improving detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of nuclear power fuel management, and more particularly to methods, systems, equipment, and media for locating regions of interest in nuclear power fuel assemblies. Background Technology

[0002] The development and utilization of nuclear energy is of great strategic significance to my country's energy security and energy transition. Safe operation has always been the foundation of nuclear energy development. Post-refueling core assembly verification is an important task that must be carried out during nuclear power plant overhauls.

[0003] Due to the unique environment of nuclear power plants, inspections must be conducted underwater. The uneven lighting underwater, the absorption and scattering effects of light as it propagates, and the influence of suspended particles make underwater imaging extremely complex. The effects of non-uniform lighting lead to color distortion, reduced contrast, and loss of detail in underwater images, posing significant challenges to image analysis, processing, and feature extraction. Furthermore, underwater image inspection in the harsh environment of nuclear power plants differs from ordinary underwater image inspection, primarily in the following ways: low color richness is commonly observed due to factors such as the camera itself and the underwater medium; noise may appear in the images due to radiation; and the close proximity of artificial underwater light sources to the camera results in severe backscattering, leading to noticeable fogging in the images.

[0004] Underwater target recognition tasks in nuclear power plants include character recognition, foreign object recognition, and crack recognition. Directly using full-view images taken in the underwater environment of a nuclear power plant for target recognition presents two problems: First, the harsh imaging environment leads to image blurring, which may create regions similar to the target area in the full-view image, thus affecting the accuracy of target detection. Second, changes in camera pose may result in symmetrical target areas, making the full-view image similar before and after the pose change, further impacting detection accuracy. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, system, device and medium for locating the region of interest of nuclear power fuel assemblies.

[0006] The technical solution adopted by this invention to solve its technical problem is: to provide a method for locating the region of interest in a nuclear power fuel assembly, comprising the following steps:

[0007] Images of fuel assemblies in the underwater environment are acquired for application purposes.

[0008] A deep learning model is used to perform feature matching between the application image and a preset template image;

[0009] The optimal transformation matrix between the application image and the template image is obtained based on the feature matching results.

[0010] Determine the region of interest on the template image, and generate the corresponding target region of interest on the application image based on the optimal transformation matrix.

[0011] Preferably, the step of acquiring images of fuel assemblies in the underwater environment as application images further includes:

[0012] Remove noise from the image; or,

[0013] Convert the image to a standard viewpoint; or,

[0014] Image enhancement is performed on the image.

[0015] Preferably, the step of using a deep learning model to perform feature matching between the application image and a preset template image includes:

[0016] Deep learning models are used to extract features from the application image and the template image, respectively.

[0017] The extracted feature points are matched using a feature matching algorithm to obtain a matching pair between the application image and the template image.

[0018] Preferably, after the step of obtaining the matching pair between the application image and the template image, the method further includes:

[0019] The matching pairs are optimized according to a preset matching pair optimization algorithm.

[0020] Preferably, the step of analyzing the optimal transformation matrix between the application image and the template image based on the feature matching results includes:

[0021] Assign several compute pairs and corresponding check pairs based on all matching pairs, wherein each compute pair includes three matching pairs and the remaining matching pairs serve as the corresponding check pairs.

[0022] The transformation matrix for each computation pair is calculated.

[0023] Based on the preset loss function, the loss value of each calculation pair is calculated according to the transformation matrix and the corresponding check pair;

[0024] The transformation matrix of the computation pair with the minimum loss value is taken as the optimal transformation matrix.

[0025] Preferably, the step of determining the region of interest of the template image includes:

[0026] Obtain the region of interest interactively drawn by the user on the template image.

[0027] Preferably, the step of generating the corresponding target region of interest on the application image based on the optimal transformation matrix includes:

[0028] Based on the feature points of the region of interest and the optimal transformation matrix, the corresponding feature points on the application image are calculated, thereby generating the target region of interest.

[0029] The present invention also provides a region of interest location system for nuclear power fuel assemblies, comprising:

[0030] The image acquisition module is used to acquire images of the fuel assembly for application purposes;

[0031] The feature matching module is used to perform feature matching between the application image and the preset template image using a deep learning model;

[0032] The optimal transformation matrix determination module is used to analyze and obtain the optimal transformation matrix between the application image and the template image based on the feature matching results;

[0033] The target region of interest generation module is used to determine the region of interest on the template image and generate the corresponding target region of interest on the application image based on the optimal transformation matrix.

[0034] The present invention also provides a computer device, comprising: a memory and a processor, wherein the memory stores a computer program executable by the processor, and the processor executes the computer program to implement the steps of the method for locating the region of interest of a nuclear power fuel assembly as described in any of the preceding claims.

[0035] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for locating the region of interest of a nuclear power fuel assembly as described in any of the preceding claims.

[0036] The method, system, device, and medium for locating the region of interest (ROI) of nuclear power fuel assemblies according to the present invention have the following beneficial effects: by using an image of the fuel assembly in an underwater environment as the application image, performing feature matching with a template image, analyzing and obtaining the optimal transformation matrix, determining the ROI on the template image, and generating a corresponding target ROI on the application image based on the optimal transformation matrix, the target ROI on the application image can be locked, thereby suppressing false identification caused by non-ROI regions and improving the detection accuracy. Attached Figure Description

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0038] Figure 1This is a flowchart illustrating an embodiment of the method for locating the region of interest in nuclear power fuel assemblies according to the present invention.

[0039] Figure 2 This is a schematic diagram illustrating the mapping relationship between the application image and the template image matching pair of the present invention;

[0040] Figure 3 This is a structural block diagram of the region of interest localization system for nuclear power fuel assemblies according to the present invention. Detailed Implementation

[0041] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0042] like Figure 1 As shown, in one embodiment of the method for locating the region of interest of a nuclear power fuel assembly according to the present invention, the method includes the following steps:

[0043] S1. Acquire images of fuel assemblies in the underwater environment as application images.

[0044] Specifically, underwater identification tasks for fuel assemblies include character recognition, foreign object recognition, and crack recognition. Based on the target identification task, the target imaging area of ​​the fuel assembly is determined. Images of the target imaging area are captured underwater using a mobile, high-resolution, radiation-resistant camera or other image acquisition device. These images serve as the initial screening images. Then, based on selection criteria such as color richness, contrast, content detail, and sharpness, the image closest to the template image is selected as the application image for further identification and locking. This process can accelerate the identification task. Alternatively, all images can be used as application images. Furthermore, the template image is a pre-acquired image with high color richness, contrast, content detail, and sharpness.

[0045] The target image area should contain inherent structural features of the fuel assembly to facilitate subsequent matching analysis. For example, when the recognition task is to identify characters on the fuel assembly tube socket, the target image area should include some inherent rounded corner features of the tube socket. To select suitable images for underwater recognition tasks, multiple images should be acquired, preferably no fewer than 1000.

[0046] Step S1 further includes: removing noise from the image; or converting the image to a standard viewpoint; or performing image enhancement on the image.

[0047] Specifically, preprocessing is required before selecting images. Optionally, denoising algorithms such as Total Variation (TV) and spatial filtering can be used to remove noise from the images to be selected. Since the camera pose changes during image capture, the viewing angles of each image differ. Therefore, the viewing angle of the template image is used as the standard viewing angle. Based on the inherent features of the image and geometric relationships, spatial transformation is performed on the images to be selected, uniformly converting them to the standard viewing angle. In other embodiments, a fixed angle can also be used as the standard viewing angle. Depending on the needs, methods such as DCP (Dark Channel Prior Dehazing), MSRCR, and CLAHE can be used to enhance the images to be selected, improving the visual effect of underwater images and increasing image clarity to facilitate target recognition by computers.

[0048] S2. Use a deep learning model to perform feature matching between the application image and the preset template image.

[0049] Specifically, a deep learning model is used to extract features from both the application image and the template image. A feature matching algorithm is then used to match the extracted feature points, resulting in matching pairs between the application image and the template image. Compared to conventional SIFT and ORB feature matching methods, deep learning offers faster matching speeds and higher accuracy compared to conventional template matching methods. A pre-trained convolutional neural network or other deep learning model is used to extract feature points from both the application image and the template image. Matching pairs are then obtained by establishing a correspondence between the template image and the application image using MatchNet or UCN algorithms. A matching pair refers to corresponding feature points in the two images. Figure 2 This diagram illustrates the mapping relationship between the application image and the template image matching pair. The left side represents the template image, and the right side represents the application image.

[0050] After obtaining a matching pair between an application image and a template image, the matching pair is optimized according to a preset matching pair optimization algorithm. In this embodiment, an adaptive method is used to optimize the matching pairs. The algorithm works as follows: first, the average distance of all matching pairs is calculated; then, the product of the average distance and a specific coefficient is calculated; the result is used as a threshold, and matching pairs with a distance less than this threshold are retained, thus achieving optimization. The specific coefficient is set according to the actual detection needs.

[0051] S3. Analyze the feature matching results to obtain the optimal transformation matrix between the application image and the template image. Step S3 includes: assigning several computation pairs and corresponding check pairs based on all matching pairs, wherein each computation pair includes three matching pairs, and the remaining matching pairs are used as corresponding check pairs; calculating the transformation matrix of each computation pair; calculating the corresponding loss value of each computation pair based on a preset loss function, according to the transformation matrix and the corresponding check pair; and taking the transformation matrix of the computation pair with the smallest loss value as the optimal transformation matrix.

[0052] Specifically, based on all matching pairs, the coordinate data of feature points in the image are transformed to generate a template matrix for the template image and an application matrix for the application image. Three pairs of coordinate points corresponding to the template and application matrices are selected as a calculation pair, and the remaining coordinate points are used as a verification pair. Based on this calculation pair, the transformation matrix between the application and template images is solved through affine transformation. Then, a preset loss function, such as the L2 loss function, is used to calculate the loss value of the verification pair based on the transformation matrix. In this embodiment, multiple sets of calculation and verification pairs are allocated at equal intervals. All calculation pairs are traversed to obtain multiple transformation matrices and corresponding loss values. Finally, the transformation matrix with the smallest loss value is selected as the optimal transformation matrix between the application and template images.

[0053] S4. Determine the region of interest (ROI) on the template image and generate the corresponding target ROI on the application image based on the optimal transformation matrix. This involves obtaining the ROI interactively drawn by the user on the template image, calculating the corresponding feature points on the application image based on the feature points of the ROI and the optimal transformation matrix, thereby generating the target ROI.

[0054] Specifically, the user interactively defines the region of interest (ROI) on the template image using a mouse. Based on the feature points of the ROI and the optimal transformation matrix, the corresponding feature points on the application image are calculated. Then, the target ROI on the application image is generated based on these feature points. In an optional embodiment, the user only needs to define the ROI once. When multiple application images exist, after obtaining the optimal transformation matrix for each application image, the corresponding target ROI is generated based on the ROI on the template image.

[0055] When photographing reactor cores, factors such as core water flow, thermal disturbance, radiation, camera oscillation, and lighting conditions result in poor photographic quality and low efficiency. To address the issues of using full-view images taken directly from the underwater environment for identification tasks, it is necessary to add a region of interest (ROI) locking mechanism. The ROI localization method for nuclear power fuel assemblies implemented in this invention can suppress false identifications caused by non-ROI regions when using application images for identification tasks, thereby improving detection accuracy, accelerating identification speed, and shortening the critical path duration of major overhauls.

[0056] Implementing this invention can shorten the core photography time for fuel assemblies and reduce the risks associated with core photography operations. It can save approximately one hour of critical path time during a single major overhaul and approximately one hour during a single spent fuel pool fuel assembly inventory, improving the economic efficiency of the power plant. During core photography, due to the negative pressure created by the cooling water flow in the pressure vessel, there is a risk that the core photography device may be sucked into the pressure vessel outlet pipe during fuel assembly numbering photography near the pressure vessel outlet pipe, potentially affecting the safety of the primary circuit equipment. The longer the device remains in this position, the greater the risk. Implementing this invention shortens the dwell time, significantly reducing this risk.

[0057] like Figure 3 As shown, in one embodiment of the nuclear power fuel assembly region of interest localization system of the present invention, it includes:

[0058] Image acquisition module 11 is used to acquire images of the fuel assembly as application images. Specifically, based on the target recognition task, the target shooting area of ​​the fuel assembly is determined. Images of the target shooting area are captured in the underwater environment using a portable, high-resolution, radiation-resistant camera device or other image acquisition device. These images are then used as images to be screened. Application images that are close to the template images are selected based on screening requirements such as color richness, contrast, content detail, and sharpness.

[0059] Furthermore, the image acquisition module 11 is also used to remove noise from the image; or to convert the image to a standard viewpoint; or to enhance the image. Specifically, it employs denoising algorithms such as Total Variation (TV) and spatial filtering to remove noise from the image to be screened; it uses the viewpoint of the template image as the standard viewpoint, and based on the inherent features of the image and geometric relationships, performs spatial transformation on the image to be screened, uniformly converting the image to be screened to the standard viewpoint; and it uses methods such as DCP (Dark Channel Prior Dehazing), MSRCR, and CLAHE to enhance the image to be screened.

[0060] The feature matching module 12 is used to perform feature matching between the application image and the preset template image using a deep learning model. Specifically, the feature matching module 12 uses a deep learning model to extract features from the application image and the template image respectively, matches the extracted feature points using a feature matching algorithm to obtain matching pairs between the application image and the template image, and optimizes the matching pairs according to a preset matching pair optimization algorithm.

[0061] The optimal transformation matrix determination module 13 is used to analyze and obtain the optimal transformation matrix between the application image and the template image based on the feature matching results. Specifically, the optimal transformation matrix determination module 13 assigns several computation pairs and corresponding check pairs based on all matching pairs, wherein each computation pair includes three matching pairs, and the remaining matching pairs are used as corresponding check pairs; calculates the transformation matrix of each computation pair; calculates the corresponding loss value of each computation pair based on a preset loss function, according to the transformation matrix and the corresponding check pair; and selects the transformation matrix of the computation pair with the smallest loss value as the optimal transformation matrix.

[0062] The target region of interest (ROI) generation module 14 is used to determine the ROI on the template image and generate the corresponding ROI on the application image based on the optimal transformation matrix. Specifically, the user interactively defines the ROI on the template image. Based on the feature points of the ROI and the optimal transformation matrix, the corresponding feature points on the application image are calculated, and then the ROI on the application image is generated based on these feature points. In an optional embodiment, the user only needs to define the ROI once. When multiple application images exist, after obtaining the optimal transformation matrix for each application image, the corresponding ROI is generated based on the ROI on the template image.

[0063] In one embodiment of an electronic device provided by the present invention, the device includes a memory and a processor. The memory stores a computer program executable by the processor. When the processor executes the computer program, it implements the steps of the method for locating the region of interest of any nuclear power fuel assembly as described above. Specifically, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed via a computer device and, when executed, performs the functions defined in the methods of the embodiments of the present invention. The computer device in the present invention can be a laptop, desktop computer, workstation, industrial control computer, server, etc.

[0064] In one embodiment of a computer storage medium provided by the present invention, a computer program is stored thereon. When the computer program is executed by a processor, it implements the steps of the method for locating the region of interest of any nuclear power fuel assembly as described above. Specifically, it should be noted that the computer-readable medium described above in the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, capable of transmitting, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0065] The aforementioned computer-readable storage medium may be included in the aforementioned computer device; or it may exist independently and not assembled into the computer device.

[0066] It is understood that the above embodiments only illustrate preferred embodiments of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can freely combine the above technical features without departing from the concept of the present invention, and can also make several modifications and improvements, all of which fall within the protection scope of the present invention. Therefore, all equivalent transformations and modifications made with respect to the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A method for locating the region of interest in a nuclear power fuel assembly, characterized in that, Includes the following steps: Images of fuel assemblies in the underwater environment are acquired for application purposes. A deep learning model is used to perform feature matching between the application image and a preset template image, wherein the template image is an image with high color richness, contrast and clarity obtained under clear conditions; The optimal transformation matrix between the application image and the template image is obtained by analyzing the feature matching results, including: based on all matching pairs, converting the coordinate data of feature points in the image to generate a template matrix of the template image and an application matrix of the application image; allocating several computation pairs and corresponding check pairs from the coordinate points corresponding to the template matrix and the application matrix at equal intervals, wherein each computation pair includes three matching pairs, and the remaining matching pairs are used as corresponding check pairs; for each computation pair, solving the transformation matrix between the application image and the template image through affine transformation; calculating the corresponding loss value based on a preset loss function, wherein the loss function is the L2 loss function; traversing all computation pairs, and determining the transformation matrix corresponding to the computation pair with the smallest loss value as the optimal transformation matrix between the application image and the template image; Determine the region of interest on the template image, and generate the corresponding target region of interest on the application image based on the optimal transformation matrix.

2. The method of claim 1, wherein, The step of acquiring images of fuel assemblies in the underwater environment as application images further includes: Remove noise from the image; or, Convert the image to a standard viewpoint; or, Image enhancement is performed on the image.

3. The method of claim 1, wherein, The step of using a deep learning model to perform feature matching between the application image and the preset template image includes: Deep learning models are used to extract features from the application image and the template image, respectively. The extracted feature points are matched using a feature matching algorithm to obtain a matching pair between the application image and the template image.

4. The method for locating the region of interest in a nuclear power fuel assembly according to claim 3, characterized in that, After the step of obtaining the matching pair between the application image and the template image, the method further includes: The matching pairs are optimized according to a preset matching pair optimization algorithm.

5. The method for locating the region of interest in a nuclear power fuel assembly according to claim 1, characterized in that, The step of determining the region of interest of the template image includes: Obtain the region of interest interactively drawn by the user on the template image.

6. The method for locating the region of interest in a nuclear power fuel assembly according to claim 1, characterized in that, The step of generating the corresponding target region of interest on the application image based on the optimal transformation matrix includes: Based on the feature points of the region of interest and the optimal transformation matrix, the corresponding feature points on the application image are calculated, thereby generating the target region of interest.

7. A region of interest localization system for nuclear power fuel assemblies, characterized in that, include: The image acquisition module is used to acquire images of the fuel assembly for application purposes; The feature matching module is used to perform feature matching between the application image and the preset template image using a deep learning model, wherein the template image is an image with high color richness, contrast and clarity obtained under clear conditions; The optimal transformation matrix determination module is used to analyze and obtain the optimal transformation matrix between the application image and the template image based on the feature matching results. This includes: based on all matching pairs, converting the coordinate data of feature points in the image to generate a template matrix for the template image and an application matrix for the application image; allocating several computation pairs and corresponding verification pairs from the coordinate points corresponding to the template matrix and the application matrix at equal intervals, wherein each computation pair includes three matching pairs, and the remaining matching pairs serve as corresponding verification pairs; for each computation pair, solving for the transformation matrix between the application image and the template image through affine transformation; calculating the corresponding loss value based on a preset loss function, wherein the loss function is an L2 loss function; and traversing all computation pairs to determine the transformation matrix corresponding to the computation pair with the smallest loss value as the optimal transformation matrix between the application image and the template image. The target region of interest generation module is used to determine the region of interest on the template image and generate the corresponding target region of interest on the application image based on the optimal transformation matrix.

8. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executable by the processor, and the processor executes the computer program to implement the steps of the method for locating the region of interest of a nuclear power fuel assembly as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for locating the region of interest of a nuclear power fuel assembly as described in any one of claims 1 to 6.

Images