Damper scale estimation method and system

Abstract

The invention relates to a damper scale estimation method and system. The method comprises the following steps: acquiring an image of a damper to be identified; processing the image of the damper to be identified to obtain a damper cylinder region, a damper coarse scale line region and a damper fine scale line region; and performing scale identification on the coarse scale line area of the damper to obtain scale result data of the damper. According to the method, the damper cylinder region, the coarse scale line region and the fine scale line region in the damper image are identified, and scale identification is performed on the coarse scale line region to obtain the scale result data, so that the efficiency and the accuracy of damper scale identification are remarkably improved, and the accuracy of damper scale identification is improved. The problems that a traditional manual visual detection or photo recognition mode is tedious, the workload is large, and personal errors are likely to be introduced are solved.

Description

Technical Field

[0001] The invention relates to the field of surface metrology, and in particular to a damper scale estimation method and system. Background Art

[0002] The damper is an important shock absorbing device. Its working principle is to consume the vibration energy of the machine equipment through the compression or relaxation of the liquid, thereby reducing the vibration amplitude and vibration time, achieving the purpose of shock absorption and buffering. It is widely used in thermal power, nuclear power, petrochemical and other energy and other industrial fields. The main uses of the damper in the energy field are: when the pipeline or equipment is subjected to sudden loads such as earthquakes, water hammers, steam hammers, etc., the damper provides resistance to movement, limits displacement, and prevents the pipeline and equipment from being subjected to violent vibration and impact, thereby protecting the pipeline and equipment from damage and ensuring the safe operation of the system; at the same time, the damper can suppress the vibration peak at the resonant frequency of the equipment, reduce the vibration transmission along the structure, reduce structural noise, and improve the service life and work efficiency of equipment and pipelines.

[0003] Dampers are widely used in the power industry. Almost all important systems are equipped with a considerable number of dampers. According to statistics, in the CPR1000 reactor type, each unit is equipped with 551 dampers, and each unit of other reactor types is also equipped with more than 300 dampers; the specific number of thermal power plants varies greatly depending on the design. The daily inspection of dampers is mainly visual inspection, checking the position of the scale line and the number of scales, ensuring that the damper scale is within the appropriate normal range and that sufficient expansion margins for pipelines or equipment are reserved. Therefore, the daily inspection of dampers is the most tedious, labor-intensive, and important task in the inspection of each power plant.

[0004] The damper is generally composed of a hydraulic cylinder and an external cylinder. There are scale lines on the hydraulic cylinder. The main function of the scale lines is to provide a reference indication of displacement, which can determine whether the working state of the damper is normal. It is also the main inspection content in the daily inspection of the damper. At present, the inspection and identification of the damper scale is mainly based on on-site visual inspection or manual recognition of on-site photos to read the damper scale. There is no research and report on the identification of damper scale lines through image recognition technology. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the damper scale inspection has low efficiency, large workload and is easy to introduce human errors, and a damper scale estimation method and system are provided.

[0006] The technical solution adopted by the present invention to solve the technical problem is: constructing a damper scale estimation method, the method comprising the following steps: Acquire an image of a damper to be identified; Processing the image of the damper to be identified to obtain a damper cylinder region, a damper coarse scale line region, and a damper fine scale line region; The rough scale line area of ​​the damper is scaled to obtain scale result data of the damper.

[0007] In one embodiment, cropping the image of the damper to be identified includes: Rotating the image of the damper to be identified so that the image of the damper appears in a horizontal state or a vertical state; The image with the adjusted angle is cropped to obtain the damper cylinder area, the damper coarse scale line area, and the damper fine scale line area.

[0008] In one embodiment, the step of performing scale identification on the thick scale line area of ​​the damper to obtain the scale result of the damper includes: Slicing the coarse scale line area of ​​the damper to obtain a plurality of image slices; Performing image quality score calculation based on the plurality of image slices to obtain an optimal image slice; Scale identification is performed on the optimal image slice to obtain damper scale data.

[0009] In one embodiment, performing image quality score calculation based on the plurality of image slices to obtain an optimal image slice comprises: The image brightness of the plurality of image slices is calculated to obtain brightness scores of different image slices; Eliminate image slices whose brightness scores do not meet the preset brightness score, and obtain a set of image slices that meet the brightness requirement; performing contrast normalization on each image slice in the set of image slices; The image clarity is calculated for the contrast-normalized image slices to obtain the clarity scores of different image slices; Based on the clarity scores of the image slices, an optimal image slice is obtained.

[0010] In one embodiment, the image clarity calculation includes contrast calculation, Laplace calculation, Sobel calculation and normalized noise calculation; The image clarity calculation is performed on the contrast-normalized image slices to obtain clarity scores of different image slices, including: Performing contrast calculation on the contrast-normalized image slices to obtain contrast scores of the image slices; Perform Laplace calculation on the image slice after contrast normalization to obtain the Laplace value of the image slice; Perform Sobel calculation on the image slice after contrast normalization to obtain the Sobel value of the image slice; The noise of the image slice after contrast normalization is calculated to obtain the noise value of the image slice.

[0011] In one embodiment, the performing scale identification on the optimal image slice to obtain damper scale data includes: Performing image enhancement on the optimal image slice to obtain an enhanced image slice; Scale recognition is performed on the enhanced image slice to obtain damper scale data.

[0012] In one embodiment, performing image enhancement on the optimal image slice to obtain an enhanced image slice includes: The optimal image slice is sequentially enhanced by using bilateral filtering denoising, contrast-limited adaptive histogram equalization technology and nonlinear grayscale transformation method to obtain an enhanced image slice.

[0013] In one embodiment, performing scale recognition on the enhanced image slice to obtain damper scale data includes: The Sobel operator is used to perform scale recognition on the enhanced image slice to obtain scale result data of the damper.

[0014] In one embodiment, the calibration result data of the damper includes scale line length information and deflection angle information; According to the scale line length information and the deflection angle information, the damper scale data that does not meet the preset conditions in the damper scale data is eliminated to obtain the eliminated damper scale data.

[0015] The present invention also constructs a damper scale estimation system, including a computer device, wherein the computer device includes a processor and a memory storing a computer program, and the processor implements the steps of any of the above-mentioned damper scale estimation methods when executing the computer program.

[0016] The implementation of the present invention has the following beneficial effects: The present invention relates to a damper scale estimation method and system, the steps of the method include: obtaining an image of a damper to be identified; processing the image of the damper to be identified to obtain a damper barrel area, a damper coarse scale line area, and a damper fine scale line area; performing scale identification on the damper coarse scale line area to obtain scale result data of the damper. The present invention significantly improves the efficiency and accuracy of damper scale identification by identifying the damper barrel area, the coarse scale line area, and the fine scale line area in the damper image, and performing scale identification on the coarse scale line area to obtain scale result data, thereby solving the problems of cumbersomeness, large workload, and easy introduction of human errors in traditional manual visual detection or photo recognition methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 It is a schematic flow chart of the damper scale estimation method provided by the present invention; Figure 2 It is a DeelLabV3+ segmentation network structure diagram in the damper scale estimation method provided by the present invention; Figure 3A is a schematic diagram of an image that has not been angle-corrected in the damper scale estimation method provided by the present invention; Figure 3B is a schematic diagram of image angle correction in the damper scale estimation method provided by the present invention; Figure 4 is a schematic diagram of image cropping in the damper scale estimation method provided by the present invention; Figure 5 It is a schematic diagram of image segmentation in the damper scale estimation method provided by the present invention; Figure 6 This is a diagram showing the effect of bilateral filtering and denoising in the damper scale estimation method provided by the present invention; Figure 7 It is the picture effect of the slice image after contrast enhancement and Gamma correction in the damper scale estimation method provided by the present invention; Figure 8 is a scale effect diagram of a damper in the damper scale estimation method provided by the present invention; Fig. 9 It is a flow chart of algorithm design of the damper scale estimation method provided by the present invention; Fig.10It is a diagram of experimental results in the damper scale estimation method provided by the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Aiming at the technical problems existing in the background technology: the damper scale inspection has low efficiency, large workload, and is easy to introduce human errors. The present invention provides a damper scale estimation method, which significantly improves the efficiency and accuracy of damper scale identification by identifying the damper cylinder area, the coarse scale line area and the fine scale line area in the damper image, and performs scale identification on the coarse scale line area to obtain scale result data, thereby solving the problems of cumbersomeness, large workload, and easy introduction of human errors existing in traditional manual visual inspection or photo identification methods.

[0021] Figure 1 It is a flow chart of an embodiment of a damper scale estimation method of the present invention.

[0022] like Figure 1 As shown, the steps of the damper scale estimation method include: S1: Acquire an image of the damper to be identified; In this step, it is first explained that the image of the damper to be identified can be a high-definition photo or video frame taken by a digital camera, a mobile phone camera or a professional image acquisition device.

[0023] S2: Processing the image of the damper to be identified to obtain a damper cylinder area, a damper coarse scale line area, and a damper fine scale line area; In this step, it is first explained that the image processing process includes but is not limited to steps such as image cropping, enhancement, filtering and feature extraction. By applying these technologies, the key areas in the damper image, namely the damper barrel area, the coarse scale line area and the fine scale line area, can be automatically or manually identified and separated.

[0024] S3: Performing scale recognition on the coarse scale line area of ​​the damper to obtain scale result data of the damper.

[0025] In this step, it should be noted that by identifying the damper cylinder area, the coarse scale line area and the fine scale line area in the damper image and performing special scale recognition processing on the coarse scale line area, the scale result data of the damper is finally obtained.

[0026] In the technical solution of this embodiment, by identifying the damper cylinder area, coarse scale line area and fine scale line area in the damper image, and performing scale recognition on the coarse scale line area, the scale result data is obtained, which significantly improves the efficiency and accuracy of damper scale recognition and solves the problems of traditional manual visual inspection or photo recognition methods that are cumbersome, labor-intensive, and easy to introduce human errors.

[0027] Furthermore, the processing of the image of the damper to be identified in step S2 includes: S21: rotating the image of the damper to be identified so that the image of the damper to be identified is in a horizontal state or a vertical state; and cropping based on the image with the adjusted angle.

[0028] like Figure 2 Specifically, in order to overcome the challenges brought by the randomness of the operator's shooting angle to the subsequent image processing, Figure 2 As shown in , firstly, an image of the damper to be identified is obtained, and a deep learning convolutional network is used to preprocess the obtained image of the damper to be identified. This step aims to segment the complex damper image into three main areas: the damper cylinder area (ZNQT), the damper coarse scale line area (ZNQC), and the damper fine scale line area (ZNQX). Figure 2As shown in the figure, in the field of image processing and recognition, DeepLabV3+ is a segmentation network. Through its unique architecture design, including 1*1 convolution (for feature dimension reduction or nonlinear enhancement), 3*3 deep separable convolution (DSDConv, designed to reduce the amount of calculation while maintaining performance), pooling layer (used to reduce the spatial dimension of feature map and extract key information), upsampling layer (used to restore the spatial resolution of feature map) and concatenation operation (Concat, fusing feature information at different levels), it can effectively perform fine segmentation on complex images. In the recognition task of damper image, the DeepLabV3+ network first receives the damper image to be recognized as input. Subsequently, through its multi-layer convolution and pooling operations, the network gradually extracts multi-level features in the image, which contain both local details of the image and broader contextual information. In the process of feature extraction, the application of deep separable convolution (DSDConv) significantly reduces the computational complexity while maintaining a high segmentation accuracy. As the network goes deeper, the spatial resolution of the feature map gradually decreases, but the semantic information gradually becomes richer. In order to balance accuracy and resolution, DeepLabV3+ introduces an upsampling and splicing mechanism. The spatial resolution of the feature map is restored through upsampling operations, and the feature map with low resolution but high semantic information is combined with the feature map with high resolution but low semantic information through splicing operations, thus achieving effective fusion of feature information. Finally, the network outputs a set of precisely segmented image regions (such as orange for the damper barrel area, green for the damper coarse scale line area, and blue for the damper fine scale line area). Each region corresponds to a specific part in the damper image, such as the damper barrel, coarse scale line, fine scale line, etc. These regions are marked with binary masks of different colors, so that the different components of the damper can be clearly distinguished visually.

[0029] After obtaining the segmentation results, these binary masks are further processed to extract the border information of each segmented area. This step is crucial for subsequent image analysis, measurement or recognition tasks because it provides the specific location and shape information of each key area. By combining the border information and segmentation mask, the damper image can be understood and processed more accurately, achieving more efficient automated recognition and measurement functions.

[0030] like Figure 3A and Figure 3B As shown, the centroid position of each mask is calculated as the geometric center of the target area, and the image is rotated to ensure that the damper is presented in a horizontal or vertical state, where Figure 3A 3B is the image before rotation, and 3B is the image after rotation. This simplifies the subsequent image processing and line detection tasks. The rotated image is not only easy to understand visually, but also provides a standardized input for subsequent processing. Figure 4As shown in the figure, after completing the angle correction, the original image is accurately cropped according to the previously identified mask border information, and three independent and clear areas of the damper cylinder, coarse scale line and fine scale line are extracted respectively. This process not only removes the redundant parts in the image, but also enhances the feature expression of the target area, laying a solid foundation for subsequent more efficient image analysis and recognition.

[0031] Further, step S3 includes S31, which includes dividing the coarse scale line area of ​​the damper to obtain a number of image slices; calculating image quality scores based on the several image slices to obtain an optimal image slice; and performing scale recognition on the optimal image slice to obtain damper scale result data.

[0032] It should be noted that the image quality score calculation includes the image brightness calculation and the image clarity calculation.

[0033] Furthermore, the step S31 in which image quality scores are calculated based on a plurality of image slices to obtain the optimal image slice includes S311, S311: performing image brightness calculation on a plurality of image slice images to obtain brightness scores of different image slices, eliminating image slices whose brightness scores do not meet preset brightness scores, and obtaining an image slice set that meets the brightness requirements; performing contrast normalization on each image slice in the image slice set.

[0034] Perform image clarity calculation on contrast-normalized image slices to obtain clarity scores of different image slices; obtain the optimal image slice based on the clarity scores of the image slices. The image clarity calculation includes contrast calculation, Laplace calculation, Sobel calculation, and standardized noise calculation.

[0035] Perform contrast calculation on the contrast-normalized image slice to obtain the contrast score of the image slice; perform Laplace calculation on the contrast-normalized image slice to obtain the Laplace value of the image slice; perform Sobel calculation on the contrast-normalized image slice to obtain the Sobel value of the image slice; perform noise calculation on the contrast-normalized image slice to obtain the noise value of the image slice.

[0036] like Figure 5 As shown, specifically, in order to accurately obtain the scale result data of the damper, the scale recognition is first performed on the coarse scale line (ZNQC) area of ​​the damper. This process first involves dividing the coarse scale line area into equal parts to generate several image slices. In order to select the optimal slice most suitable for scale recognition from many image slices, the image quality is scored.

[0037] First, image brightness detection is performed. By calculating the brightness of each image slice, their brightness scores are obtained, and slices whose brightness scores do not meet the preset standards are eliminated to ensure that the images processed subsequently have appropriate lighting conditions. Then, the remaining image slices are subjected to contrast normalization to eliminate the impact of different lighting conditions on image quality.

[0038] Next, image clarity was assessed, and the remaining image slices were contrast normalized to eliminate the effects of different lighting conditions on image quality.

[0039] Subsequently, a variety of clarity evaluation methods, including contrast calculation, Laplace operator second-order derivative normalization, Sobel edge detection normalization, and noise calculation, are used to comprehensively evaluate the clarity of each image slice. The contrast score measures the degree of contrast between light and dark in the image; the Laplace value reflects the change in the second-order derivative of the image, that is, the sharpness of the edges and details; the Sobel value strengthens the edge information of the image in the horizontal and vertical directions; and the noise calculation evaluates the amount of noise in the image to ensure that the selected image slice is not disturbed by too much noise.

[0040] After comprehensively considering the contrast score, Laplace score, Sobel score and normalized noise score, the final clarity score of each image slice is determined by calculating their weighted average. Finally, based on this scoring system, the image slice with the highest score is selected as the optimal slice for subsequent scale recognition.

[0041] Specific image brightness detection: Brightness mean What is calculated is the average value of all brightness values ​​of the image, which can be used to measure the overall brightness level of the image. The formula is as follows: in, is the image at position Gray value, is the total number of pixels in the image. The average deviation from the median (128) is used to evaluate the dispersion of image brightness values ​​relative to the median. The formula is: The brightness coefficient uses the brightness mean and average deviation to evaluate whether the image is brighter or darker, and takes into account the uniformity of brightness distribution. The calculation formula is: If it is greater than a certain threshold , it means that the brightness distribution of the image is uneven.

[0042] Furthermore, the specific method of image clarity evaluation is as follows: By contrast normalization, Laplace operator is used to calculate the second-order derivative normalization of the image and the image , The Sobel edge enhancement standardization in the direction reflects the edge information of the image. The clearer the image, the larger the corresponding variance after the wave. Then the image is evaluated for noise. The larger the difference with the mean absolute deviation (MAD) after Gaussian smoothing, the higher the noise. The specific calculation formula is as follows: in, is the standard deviation of the image pixel values.

[0043] in, is the variance of the Laplace transform result.

[0044] in, , are the Sobel edge detection results of the image in the x and y directions, respectively, and is the total number of image pixels.

[0045] in: , The original image pixel value, are the weights of the Gaussian filter.

[0046] Finally, (Contrast Rating), (Laplace score) and (Sobel score) and add the average, and combine the standardized (Noise Rating): After the column part of the damper is cut, the cut sub-images can be evaluated using the above four evaluation indicators to obtain the optimal image.

[0047] S312: Perform image enhancement on the optimal image slice to obtain an enhanced image slice; perform scale recognition on the enhanced image slice to obtain damper scale result data. The optimal image slice is sequentially enhanced using bilateral filtering denoising, contrast-limited adaptive histogram equalization technology, and nonlinear grayscale transformation method to obtain an enhanced image slice. The Sobel operator is used to perform scale recognition on the enhanced image slice to obtain the damper scale result data.

[0048] The damper scale result data includes scale line length information and deflection angle information; according to the scale line length information and the deflection angle information, the damper scale result data that does not meet the preset conditions is eliminated from the damper scale result data to obtain the eliminated damper scale result data.

[0049] like Figure 6 Specifically, the selected optimal image slices are further subjected to fine image enhancement processing to improve the accuracy and reliability of scale recognition. First, the noise in the image is removed by bilateral filtering technology, while retaining edge details to avoid edge blurring problems that may be caused by traditional filtering methods. Figure 6 The effect of bilateral filtering denoising is shown in Figure 1. Bilateral filtering comprehensively considers the spatial proximity and color similarity between pixels, ensuring effective noise suppression and clear edge retention. Subsequently, the contrast-limited adaptive histogram equalization (CLAHE) technology is used to perform local contrast enhancement on the image. CLAHE effectively avoids the noise amplification problem that may be introduced by global histogram equalization by limiting the contrast amplification degree of each small block. At the same time, combined with Gamma correction, the grayscale value of the image is nonlinearly transformed to further adjust the brightness and contrast of the image, making the image details clearer and the contrast more vivid. After the above image enhancement processing, an image slice with significantly improved quality is obtained. Next, the Sobel operator is used for scale recognition, especially for accurate extraction of the scale lines of the damper. The Sobel operator effectively detects the edge information in the image, especially those significant features corresponding to the scale lines, by calculating the gradient of the image in the horizontal and vertical directions. In the process of scale line extraction, the length information and deflection angle information of each scale line are calculated. This information not only helps to identify the scale more accurately, but also can be used for subsequent data screening and correction. Specifically, according to preset conditions (such as scale line length threshold and deflection angle range), those scale result data that do not meet the conditions are eliminated, thereby ensuring that the final damper scale result data is both accurate and reliable.

[0050] The specific bilateral filtering processing method is: in order to filter out image noise while trying not to lose its edge information, bilateral filtering based on spatial proximity and pixel similarity is used, which can effectively avoid the problem of traditional filtering methods blurring edges while smoothing noise.

[0051] Output pixel value of bilateral filtering In unknown location It can be calculated by the following formula: in: is the pixel value at that location in the original image.

[0052] So The neighborhood window centered at .

[0053] is the position in the neighborhood window The pixel value of .

[0054] It is a color space kernel function that measures the similarity in color intensity between the center pixel and the neighboring pixels. It is usually a Gaussian function with the difference in pixel values ​​as the variable.

[0055] It is a spatial kernel function that measures the proximity of the central pixel to the neighboring pixels in spatial position. It is usually a Gaussian function with spatial distance as a variable.

[0056] Represents intensity differences in color space.

[0057] Represents the Euclidean distance of spatial locations.

[0058] is the normalization factor, which ensures that the pixel value after filtering is within a reasonable range and is calculated as follows: The color space kernel function is usually defined as: Where is the standard deviation of the color space kernel, controlling the range of color similarity.

[0059] Spatial kernel function. Usually defined as: where is the standard deviation of the spatial kernel, controlling the range of spatial proximity.

[0060] In this way, bilateral filtering is able to take into account both spatial and intensity similarities between pixels, thereby preserving image edges while removing noise.

[0061] CLAHE (Contrast Limited Adaptive Histogram Equalization) Technology: CLAHE avoids the noise amplification problem that may be caused by global histogram equalization by limiting contrast enhancement, and effectively enhances the image contrast. Through Gamma correction, the brightness and contrast of the image are further adjusted to make the visual effect of the image clearer. And Gamma correction enhances the local details of the image by performing nonlinear transformation on the grayscale value of the image. The combination of CLAHE and Gamma correction can significantly improve the visual effect of the image, making the image details clearer and the contrast higher.

[0062] CLAHE first divides the image into several small blocks, each of which is usually 8x8, 16x16, etc. In order to apply histogram equalization in the local area, thereby improving the local contrast.

[0063] In order to prevent excessive amplification of noise during histogram equalization, CLAHE applies contrast limitation to the histogram of each small block. A threshold (clipLimit) is set. If the value of a bin in the histogram exceeds this threshold, the excess part is evenly distributed to all bins. For each histogram : in, is the value of the i-th bin in the original histogram, and is the value after contrast limiting is applied. The total number exceeding the threshold will be evenly distributed to all bins, and each restricted histogram will be equalized. The new pixel value of the processed image can be calculated by the following formula: in is the cumulative distribution function of the histogram values, yes The minimum non-zero value of is the total number of pixels in the image, and are the maximum and minimum possible values ​​of the image pixel values, respectively. After limiting and equalizing the contrast by block, in order to eliminate the visible boundaries between small blocks, bilinear interpolation is used to smooth the small blocks.

[0064] Gamma correction is a nonlinear grayscale transformation method that can enhance the local details of the image by performing power operations on the grayscale values ​​of the image, making the image visually clearer. By adjusting the brightness and darkness of the image, the dark details are made more obvious while avoiding over-enhancement of the bright areas. The formula is as follows: (0.5) in: is the pixel value of the input image, usually converted to the range , for 8-bit images, the pixel values ​​are in the range.

[0065] is the output image pixel value after gamma correction.

[0066] is the gamma value, which controls the degree of correction.

[0067] when , the contrast of highlights will increase, making the image look brighter.

[0068] when When the brightness is increased, the contrast of dark areas will increase, making the image look darker.

[0069] like Figure 7 As shown, Figure 7 This is the picture effect after CLAHE and Gamma correction.

[0070] In order to identify the scale lines in the image, edge detection can effectively extract the scale information in the image and is easy to implement. By calculating the gradient of the image grayscale change, the collection of points where the image pixels change sharply is found.

[0071] The specific Sobel operator is the core of the scale detection method: the Sobel operator convolves the image in the vertical and horizontal directions. The process is a sequential convolution of vertical first and then horizontal. The weighted convolution results of the two finally obtain a clear scale detection effect. The areas with larger gradient values ​​are scale lines, and these areas usually correspond to significant features in the image.

[0072] The algorithm idea is: (1) Calculate the vertical gradient With horizontal gradient .

[0073] (2) Calculation and The square root of .

[0074] (3) Carry out the instantiation process and judge The relationship between the size of the threshold and If the value is greater than the threshold, the output is 1; if the value is less than the threshold, the output is 0.

[0075] (4) Finally, the detected image is obtained. Figure 8 As shown, Figure 8 This is the scale effect diagram of the damper.

[0076] Compared with other methods, the convolution process in this paper will not produce the convolution kernel inversion effect. The weighted convolution in two directions is used, and the result will be more reasonable.

[0077] like Fig. 9As shown in the figure, in the image processing and analysis process, the damper image to be processed is first received as input, and then the pre-trained convolutional neural network model is used to finely segment the image to distinguish key areas such as the damper tube (ZNQT), coarse scale line (ZNQC) and fine scale line (ZNQX), and generate corresponding binary masks. Then, in view of the possible image tilt problem, the angle correction technology is used to ensure that the image is in the correct direction. Subsequently, the border information of each area is extracted from the corrected image, and the original image is cut into sub-images containing separate areas based on this information to construct a data set. By evaluating the data set, the sub-image with the best quality is selected for denoising to remove image noise. After that, the image enhancement technology such as contrast enhancement and sharpening is used to improve the quality of the sub-image, and then the image processing algorithm or machine learning model is used to accurately extract the coarse scale line and fine scale line of the damper from the enhanced sub-image. Finally, the number of extracted scale lines is output, which provides crucial data support for subsequent accurate measurement, in-depth analysis and report preparation.

[0078] Verify the experiment: like Fig.10 As shown in the figure, the computer configuration used in the experiment is as follows: the processor is Intel(R) Core(TM) i7-8750HCPU@2.20GHz, the memory is 16GB, the operating system is 64-bit Windows10, the development environment is Pycharm2024.2, the programming language is Python3.10, and OpenCV and NumPy and other libraries are used for image processing and mathematical calculations. In order to verify the effectiveness of the algorithm in this paper, the actual damper pictures with a size of 3672×4896 are collected as experimental test data. First, the damper image is segmented by a convolutional neural network, and the scale estimation is performed after morphological processing of the segmentation result. The experimental results are shown in Fig.10 shown.

[0079] The present invention also constructs a damper scale estimation system, including a computer device, the computer device including a processor and a memory storing a computer program, and the processor implements the steps of any of the above damper scale estimation methods when executing the computer program.

[0080] The present invention proposes a scale line recognition and processing method based on image processing and edge detection. By applying adaptive binarization, CLAHE adaptive histogram equalization and edge detection techniques, the scale lines in the image can be effectively recognized and denoised to improve the image quality. The experimental results verify the effectiveness and practicality of this method.

[0081] The beneficial effects are: 1. The present invention is based on the method of image processing and scale detection, which can realize the recognition of the damper image on site, analyze and read the damper scale, and judge the operating status of the damper.

[0082] 2. The present invention can be combined with intelligent inspection robots, intelligent inspection and other technologies in power plants to realize inspection and analytical diagnosis of numerous dampers in the field, which can save a lot of on-site inspection workload for operating personnel and reduce the problems of misreading and missing damper inspection work.

[0083] 3. The present invention uses a convolutional neural network deep learning method to segment and mask the damper cylinder area cylinder (ZNQT), the damper coarse scale line (ZNQC), and the damper fine scale line (ZNQX) area to obtain the area to be identified. Subsequently, only the coarse scale line (ZNQC) area image needs to be preprocessed and scale recognition work is performed, which can greatly improve image processing efficiency and recognition accuracy; 4. The present invention combines the deep learning convolutional network method to equally divide the coarse scale line (ZNQC) area to obtain equal sub-graphs of the area, which can facilitate the subsequent selection of the optimal sub-graph for analysis and identification; 5. The present invention selects the best image in the molecular image by monitoring the image brightness and evaluating the clarity, which can remove the interference of factors such as light and environment on the image, extract the most suitable area for recognition, and ensure the reliability of the recognition result; 6. The bilateral filtering based on spatial proximity and pixel similarity of the present invention can protect the image edge while removing noise and highlight the scale line feature information of the damper.

[0084] 7. The present invention adopts the combined use of CLAHE technology and Gamma correction to significantly improve the visual effect of the image, effectively enhance the image contrast, enhance the light and dark relationship of the image, and make the dark details of the image more obvious.

[0085] 8. The present invention uses the Sobel operator as the core of the edge detection method, and performs sequential convolution on the image in the vertical and horizontal directions to obtain a clear scale detection result of the damper scale line.

[0086] It can be understood that the above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the patent scope of the present invention. It should be pointed out that, for ordinary technicians in this field, the above technical features can be freely combined without departing from the concept of the present invention, and several deformations and improvements can be made, which all belong to the protection scope of the present invention. Therefore, all equivalent changes and modifications made to the scope of the claims of the present invention should belong to the coverage of the claims of the present invention.

Claims

1. A damper scale estimation method, characterized in that: include: Acquire an image of a damper to be identified; Processing the image of the damper to be identified to obtain a damper cylinder region, a damper coarse scale line region, and a damper fine scale line region; The rough scale line area of ​​the damper is scaled to obtain scale result data of the damper.

2. The damper scale estimation method according to claim 1, characterized in that: The processing of the image of the damper to be identified comprises: Rotating the image of the damper to be identified so that the image of the damper appears in a horizontal state or a vertical state; Crop based on the image at an adjusted angle.

3. The damper scale estimation method according to claim 1, characterized in that: The step of performing scale identification on the thick scale line area of ​​the damper to obtain the scale result of the damper includes: Slicing the coarse scale line area of ​​the damper to obtain a plurality of image slices; Performing image quality score calculation based on the plurality of image slices to obtain an optimal image slice; Scale identification is performed on the optimal image slice to obtain damper scale data.

4. The damper scale estimation method according to claim 3, characterized in that: The performing image quality score calculation based on the plurality of image slices to obtain the optimal image slice comprises: The image brightness of the plurality of image slices is calculated to obtain brightness scores of different image slices; Eliminate image slices whose brightness scores do not meet the preset brightness score, and obtain a set of image slices that meet the brightness requirement; performing contrast normalization on each image slice in the set of image slices; The image clarity is calculated for the contrast-normalized image slices to obtain the clarity scores of different image slices; Based on the clarity scores of the image slices, an optimal image slice is obtained.

5. The damper scale estimation method according to claim 4, characterized in that: The image clarity calculation includes contrast calculation, Laplace calculation, Sobel calculation and standardized noise calculation; The image clarity calculation is performed on the contrast-normalized image slices to obtain clarity scores of different image slices, including: Performing contrast calculation on the contrast-normalized image slices to obtain contrast scores of the image slices; Perform Laplace calculation on the image slice after contrast normalization to obtain the Laplace value of the image slice; Perform Sobel calculation on the image slice after contrast normalization to obtain the Sobel value of the image slice; The noise of the image slice after contrast normalization is calculated to obtain the noise value of the image slice.

6. The damper scale estimation method according to claim 3, characterized in that: The step of performing scale identification on the optimal image slice to obtain damper scale data includes: Performing image enhancement on the optimal image slice to obtain an enhanced image slice; Scale recognition is performed on the enhanced image slice to obtain damper scale data.

7. The damper scale estimation method according to claim 6, characterized in that: The performing image enhancement on the optimal image slice to obtain the enhanced image slice comprises: The optimal image slice is sequentially enhanced by using bilateral filtering denoising, contrast-limited adaptive histogram equalization technology and nonlinear grayscale transformation method to obtain an enhanced image slice.

8. The damper scale estimation method according to claim 7, characterized in that: The step of performing scale recognition on the enhanced image slice to obtain damper scale data includes: The Sobel operator is used to perform scale recognition on the enhanced image slice to obtain scale result data of the damper.

9. The damper scale estimation method according to claim 8, characterized in that: The scale result data of the damper includes scale line length information and deflection angle information; According to the scale line length information and the deflection angle information, the damper scale data that does not meet the preset conditions in the damper scale data is eliminated to obtain the eliminated damper scale data.

10. A damper scale estimation system, characterized in that: It comprises a computer device, characterized in that the computer device comprises a processor and a memory storing a computer program, and the processor implements the steps of the damper scale estimation method according to any one of claims 1 to 9 when executing the computer program.

Images