Converter valve platinum electrode scaling thickness detection method based on image gray analysis
By using an image grayscale analysis-based method, the problem of online non-destructive testing of scale thickness on platinum electrodes of converter valves was solved, achieving high-precision and non-destructive scale thickness measurement, improving testing efficiency and result reliability, and avoiding equipment damage and manual intervention.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID ZHEJIANG ELECTRIC POWER CO LTD ZHOUSHAN POWER SUPPLY CO
- Filing Date
- 2026-01-28
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies make it difficult to achieve online, non-destructive, and high-precision detection of scale thickness on platinum electrodes of converter valves. Traditional methods are cumbersome, time-consuming, and labor-intensive. Furthermore, existing non-destructive testing methods have low positioning accuracy in X-ray images with low contrast and gradual boundaries, and lack in-depth physical analysis and adaptive compensation.
A method based on image grayscale analysis is adopted to acquire X-ray transmission digital images for noise reduction and contrast enhancement, identify the boundary feature points between the scale layer and the electrode substrate, and calculate the scale thickness by combining pixel size calibration and X-ray attenuation model, so as to achieve reliable conversion from image information to physical thickness.
It enables efficient online inspection without downtime and disassembly, improves inspection accuracy and robustness, ensures the physical accuracy and engineering reliability of measurement results, and avoids equipment damage and reliance on manual intervention.
Smart Images
Figure CN122023322A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of non-destructive testing technology for converter valve equipment, and in particular to a method for detecting the scale thickness on the platinum electrode of a converter valve based on image grayscale analysis. Background Technology
[0002] The converter valve is a core component of a high-voltage direct current (HVDC) transmission system, and its stable operation is crucial. During long-term operation, the platinum electrodes in the valve's cooling system will accumulate scale due to the adsorption of metal ions from the water circuit. This scale has insulating properties; when its thickness accumulates to a certain level, it will cause the platinum electrodes to fail, triggering flashover and seriously threatening the safety of the converter valve.
[0003] Currently, the detection of platinum electrode scale thickness mainly faces the following challenges:
[0004] The inherent drawbacks of traditional offline testing: The industry-standard "shutdown, drainage, disassembly, measurement" method is cumbersome, time-consuming, labor-intensive, and has limited maintenance windows. Frequent disassembly can also damage electrodes and their sealing structures, introducing new failure risks. This has created an urgent need for online non-destructive testing technology.
[0005] Existing non-destructive testing methods have limitations in accuracy and applicability: To achieve non-disassembly inspection, methods based on vision or X-ray imaging have been explored. However, in the specific application scenario of converter valves, these methods face significant bottlenecks: The imaging conditions are harsh. The platinum electrode is located inside the pipe, and the X-rays need to penetrate the pipe wall and the medium, resulting in a low signal-to-noise ratio and poor contrast. The grayscale difference between the scale layer and the metal electrode substrate is small, and the boundary appears as a blurred, gradual transition in the image rather than a clear geometric edge.
[0006] Conventional image size measurement methods (such as edge detection based on Canny operators and simple thresholding) heavily rely on sharp, high-contrast boundaries. When faced with low-contrast X-ray images with gradient boundaries, their localization accuracy drops sharply, their noise resistance is poor, and the results have low repeatability.
[0007] Furthermore, some methods rely on conditions that are difficult to strictly guarantee on-site, such as "known precise reference dimensions" or "ideal imaging angles." Others heavily depend on operators' experience to manually select the measurement area, resulting in high subjectivity, low efficiency, and difficulty in achieving automation and standardization.
[0008] The lack of intelligent analysis methods deeply integrated with the principles of imaging physics: Existing technologies mostly treat images as purely two-dimensional information for processing, failing to fully explore the physical relationship between gray values and material thickness and density in X-ray imaging (i.e., the attenuation law). At the same time, there is a lack of systematic adaptive compensation and correction for unavoidable interference factors such as noise, background inhomogeneity, and angular tilt during the imaging process, resulting in insufficient physical accuracy and environmental robustness of the measurement results.
[0009] In summary, there is an urgent need for a non-destructive testing method specifically designed for platinum electrodes in converter valves, capable of overcoming the difficulties of low-quality X-ray imaging and achieving high-precision, automated, and physically reliable scaling thickness testing. Summary of the Invention
[0010] The technical problem to be solved and the technical task proposed by this invention is to improve and refine existing technical solutions, and to provide a method for detecting the scale thickness of platinum electrodes in converter valves based on image grayscale analysis, with the aim of achieving online, non-destructive, and high-precision detection of scale thickness on platinum electrodes in converter valves. To this end, this invention adopts the following technical solution.
[0011] A method for detecting the scale thickness on a platinum electrode of a converter valve based on image grayscale analysis includes the following steps: S1. Acquire an X-ray transmission digital image of the platinum electrode to be tested, and perform noise reduction and contrast enhancement processing on the digital image to obtain a preprocessed image; S2. On the preprocessed image, determine a measurement line segment that passes through the scaled area on the electrode surface; S3. Extract the grayscale values of all pixels on the measured line segment to form grayscale distribution data; S4. Analyze the grayscale distribution data to identify the first and second feature points that characterize the boundary between the scale layer and the electrode substrate; S5. The actual physical thickness of the scale layer is calculated based on the pixel distance between the first feature point and the second feature point on the measurement line segment, and the pixel size calibration coefficient of the system.
[0012] This technical solution performs measurements by analyzing X-ray transmission images, eliminating the need to drain the converter valve cooling system or disassemble the platinum electrodes. This avoids the operational risks, high time costs, and potential physical damage to equipment caused by frequent disassembly, enabling in-service and online inspection and significantly improving maintenance efficiency and safety. Unlike traditional image measurement methods that rely on clear geometric edges, this solution directly analyzes the grayscale distribution data on the measurement line segment and identifies feature points representing the boundary. This allows it to effectively handle images with low contrast and gradually changing boundaries caused by X-rays penetrating multiple layers of material, solving the fundamental difficulty of low positioning accuracy in such images by traditional geometric edge detection methods. By introducing a pixel size calibration coefficient, a quantitative conversion relationship is established between image pixel distance and actual physical space size. This gives the final output actual physical thickness clear physical meaning and engineering value, rather than a relative pixel value. The result can be directly used for engineering evaluation, achieving a reliable conversion from image information to key physical quantities.
[0013] As a preferred technical means: in step S1, the X-ray transmission digital image is acquired by a 16-bit depth flat panel detector; the noise reduction and contrast enhancement processing includes smoothing the image using Gaussian filtering.
[0014] Employing 16-bit depth imaging provides up to 65,536 grayscale levels, significantly preserving subtle intensity differences in X-rays as they penetrate different materials. Simultaneously, Gaussian filtering for smoothing effectively suppresses random noise in the image, improving the stability and signal-to-noise ratio of the overall detection process.
[0015] As a preferred technical means: in step S1, the contrast enhancement process further includes: performing a morphological top-hat transformation on the denoised image to correct background inhomogeneity and highlight the edge features of the scaled area.
[0016] By introducing morphological top-hat transformation, slowly changing background components in the image can be effectively removed, and uneven background grayscale caused by non-uniform X-ray sources or differences in object thickness can be corrected. This enhances the local contrast between the scaled area and the surrounding background, making the scale's contour features more prominent, which is particularly beneficial for improving the ability to identify thin-layer scale in complex backgrounds.
[0017] As a preferred technical means: In step S1, the contrast enhancement processing further includes: obtaining the gray-scale mean value of the non-scalded area in the image as a reference value, dividing the gray-scale range of the image into multiple intervals based on the reference value, applying different gray-scale transformation coefficients to different intervals, and finally normalizing the result to a preset range.
[0018] By utilizing the information of non-fouled areas contained in the image itself as an adaptive benchmark, global contrast enhancement is transformed into segmented optimization, which can perform differentiated enhancement on different gray-scale areas in the image (such as background, substrate, and fouling), avoiding local overexposure or loss of detail that may occur when using a single stretching coefficient on the entire image.
[0019] As a preferred technical means: the plurality of intervals includes a first interval below the reference value, a second interval around the reference value, and a third interval above the reference value; different linear transformation coefficients are used to stretch the grayscale of the first interval and the third interval respectively, while the grayscale of the second interval is kept unchanged or subjected to low intensity transformation.
[0020] A segmented enhancement strategy centered on a baseline value can significantly amplify the grayscale differences between the background and scale, and between the substrate and scale, while effectively protecting key transition details such as substrate texture from being overwhelmed by noise generated by excessive enhancement. This "enhance the ends, protect the middle" strategy improves contrast while better preserving the realism and usable details of the image.
[0021] As a preferred technical means, step S4 specifically includes: S4.1. In the grayscale distribution data, find the point with the minimum grayscale value and mark it as the point with the thickest scale. S4.2. Based on the thickest point of scale, select a subset of grayscale distribution data on one side of it; S4.3. Calculate the sequence of absolute values of the gray-level differentials of adjacent data points in the gray-level distribution data subset; S4.4. In the sequence of absolute differential values, determine the two extreme points with the largest and second largest values, and mark them as the first feature point and the second feature point, respectively.
[0022] This technical solution uses the minimum grayscale point as the analysis benchmark to ensure the correctness of the analysis direction. Calculating the absolute value of the grayscale derivative and finding its maximum and second-largest extreme points essentially locates the position with the largest grayscale change rate, which precisely corresponds to the center of the fuzzy boundary between the scale layer and the substrate. This method is insensitive to the grayscale gradient morphology of the boundary and can reliably identify the true boundary transition zone, exhibiting higher accuracy and robustness compared to the fixed threshold method.
[0023] As a preferred technical means: In step S5, the actual physical thickness T is calculated or equivalently converted based on the X-ray attenuation model using the following formula:
[0024] in, The linear attenuation coefficient of the scaling material for X-rays of a specific energy is given. This is the reference grayscale value for the scale-free electrode substrate area. This represents the grayscale value of the scaled area.
[0025] This method bases thickness calculations on the physical laws of X-ray attenuation (Lambert-Beer Law), enabling not only geometric conversions but also thickness inversion based on material physical properties. By utilizing the grayscale ratio between the fouled region and the non-fouled substrate region, the physical thickness is directly calculated, providing a solid physical basis for the measurement results. This results in greater theoretical accuracy and is less affected by changes in the overall gain or brightness of the imaging system, thus improving the physical accuracy and reliability of the measurement results.
[0026] As a preferred technical means: In step S5, the pixel size calibration coefficient of the system is used in conjunction with the imaging angle parameter. The actual physical thickness is obtained by multiplying the pixel distance by the pixel size calibration coefficient and then dividing by the cosine value of the imaging angle.
[0027] This technical solution introduces an imaging angle compensation mechanism based on basic pixel calibration. When the X-ray imaging direction is not perpendicular to the scale surface to be measured, the measured pixel distance in the image will be greater than the actual vertical thickness. By dividing by the cosine of the imaging angle, this geometric deviation caused by perspective projection can be corrected, thereby obtaining the true physical thickness of the scale layer in the normal direction, significantly improving the measurement accuracy under non-ideal imaging postures. The imaging angle is the angle between the X-ray imaging direction and the surface of the platinum electrode to be measured.
[0028] As a preferred technical means, the pixel size calibration coefficient is obtained by taking an X-ray image of a standard calibration block with known physical size and calculating the ratio of the pixel size of the calibration block in the image to its actual physical size.
[0029] This technical solution uses physical standard calibration blocks for calibration under the same imaging conditions, unifying all factors such as detector pixel size and system geometric magnification into a single calibration coefficient. This method is simple and effective, and the calibration results directly reflect the spatial scale relationship of the entire imaging system under the current configuration, ensuring the accuracy and traceability of the pixel-to-physical-size conversion.
[0030] As a preferred technical means, the method also includes step S6: comparing the actual physical thickness calculated this time with the pre-stored reference threshold or historical measurement data. If the deviation exceeds the allowable range, the noise reduction parameter or contrast enhancement parameter in step S1, or the boundary recognition parameter in step S4, or a combination thereof, are automatically adjusted, and the scale thickness detection process is re-executed based on the adjusted parameters.
[0031] This technical solution adds closed-loop feedback and adaptive optimization capabilities to the entire detection system. By comparing the output results with a reliable reference, the system can automatically sense whether the current processing parameters are suitable for the current imaging conditions (such as noise level and contrast changes). Once an abnormal deviation is detected, the system automatically adjusts the parameters and re-detects, thus effectively coping with the complex and potentially changing imaging environment on site. This improves the system's long-term stability, adaptability, and overall measurement reliability, while reducing reliance on manual intervention and parameter readjustment.
[0032] Beneficial effects: It achieves efficient non-destructive online testing: it eliminates the cumbersome process of stopping, draining, and disassembling required by traditional methods, avoids the risk of equipment damage caused by disassembly, and greatly improves testing efficiency and maintenance safety.
[0033] Improved analysis accuracy of low-quality images: In response to the characteristics of low contrast and blurred boundaries in X-ray images, a method combining gray-scale distribution analysis and gradient extremum localization is adopted, which effectively overcomes the failure problem of traditional edge detection algorithms in such scenarios and significantly improves the accuracy and robustness of boundary recognition and thickness measurement.
[0034] It ensures the physical authenticity and reliability of the measurement results: through the dual guarantee of system calibration and physical model (X-ray attenuation law), the image pixel information is accurately converted into thickness values with clear physical meaning. It can also adapt to complex working conditions through angle compensation, parameter self-optimization and other means, making the measurement results more reliable and directly applicable to engineering decisions. Attached Figure Description
[0035] Figure 1 This is a flowchart of the present invention.
[0036] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation
[0037] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.
[0038] Example 1: This invention provides a method for detecting the scale thickness on the platinum electrode of a converter valve based on image grayscale analysis, such as... Figure 1 As shown, this invention aims to solve the technical problems of requiring machine shutdown for disassembly and inspection in existing technologies, as well as the difficulty of traditional image methods in handling low-contrast blurred boundaries in X-ray images.
[0039] Step S1: Image Acquisition and Preprocessing The purpose of this step is to acquire and optimize the raw X-ray images to lay the foundation for subsequent analysis.
[0040] First, the X-ray source and flat panel detector are turned on to perform radiographic imaging on the installed platinum electrode. The raw image acquired by the flat panel detector is a 16-bit deep single-channel grayscale image with a pixel value range of 0 (pure black) to 65535 (pure white), which can accommodate an extremely wide dynamic range and accurately record the intensity attenuation differences of X-rays after passing through different materials (water pipe wall, water medium, platinum electrode substrate, scale layer).
[0041] After acquiring the original image, preprocessing is required to improve image quality. Preprocessing mainly includes two steps: noise reduction and contrast enhancement.
[0042] Adaptive noise reduction: Since X-ray imaging is susceptible to interference such as quantum noise, noise reduction is performed first. This embodiment uses adaptive Gaussian filtering. Specifically, it involves calculating the noise variance of the uniform background region without scale in the image. .according to The size of the Gaussian convolution kernel is dynamically selected based on its parameters: when When the image noise is significant, use a larger kernel size (e.g., 7×7) and a larger standard deviation (e.g., 2.0) for stronger smoothing; when For smaller images, a smaller kernel size (e.g., 3×3) and a smaller standard deviation (e.g., 0.5) are used for slight smoothing to achieve the best balance between noise reduction and preservation of edge details. To ensure computational accuracy, the 16-bit integer image data is converted to floating-point type before filtering and then restored to 16-bit format after filtering.
[0043] Segmented contrast enhancement processing: The image after noise reduction often lacks contrast, and the scaled areas are not obvious. This embodiment adopts a combined enhancement strategy.
[0044] First, a morphological top-hat transformation is performed. A circular structuring element slightly larger than the width of the scale is selected, and a morphological opening operation is performed on the image. Then, the opening result is subtracted from the original image. This operation effectively suppresses background unevenness and significantly highlights the bright scale areas in a dark background.
[0045] Secondly, segmented grayscale stretching based on a baseline value is performed. A known platinum electrode substrate area without scale is selected in the image, and its average grayscale value M is calculated as the baseline value. Centered on M, the entire grayscale range is divided into three intervals: interval one (below M-50), interval two (M-50 to M+50), and interval three (above M+50). Different linear gain coefficients are applied to intervals one and three to amplify the grayscale difference between scale and background, and between scale and substrate; the linear gain coefficient for interval two is kept at 1, essentially unchanged, to protect the texture details of the electrode substrate itself from being over-enhanced and distorted. Finally, the stretched image data is re-normalized to the range of 0-65535. After the above processing, a pre-processed image with significantly improved quality is obtained.
[0046] Step S2: Interactive measurement area determination The purpose of this step is to precisely specify the location of the scale cross-section to be analyzed on the image.
[0047] The pre-processed image is displayed on the software's interactive interface. The user operates on the scaled area using the mouse: pressing the left mouse button sets the starting point of the measurement line segment, and releasing the left mouse button sets the ending point; a measurement line segment L is defined between these two points. During mouse movement, the interface previews the line segment in real time. Simultaneously, the software automatically outlines the approximate contour of the scale using the result of the top-hat transformation in step S1, and prompts the user during the preview to ensure that line segment L passes perpendicularly through the thickest area of scale and completely covers the width of the scale.
[0048] Step S3: Grayscale Data Analysis The purpose of this step is to extract the original grayscale information on the measured line segment.
[0049] The Bresenham line scanning algorithm is used to quickly and accurately calculate the coordinates of all pixels covered by the measurement line segment L. Then, the 16-bit grayscale values at these coordinate positions are sequentially read from the preprocessed image and arranged in order to form a grayscale distribution data sequence. This sequence visually reflects the continuous change of grayscale from the background to the scale layer and then to the substrate along the direction of line segment L.
[0050] Step S4: Boundary Feature Point Identification This step is the core of the method of the present invention, and aims to accurately locate the boundary of the scale from the grayscale data.
[0051] Locate the thickest point of scale: Traverse the grayscale distribution data obtained in step S3, find the point with the minimum grayscale value, which corresponds to the position where X-rays penetrate the thickest layer, i.e., the thickest point of scale layer, and mark it as point A.
[0052] Extracting and analyzing a subset: Using point A as the boundary, select the grayscale data to its left (or right) to form a subset, which is used to analyze the boundary on that side. This is because the scale layer is usually symmetrical or approximately symmetrical about the thickest point, so analyzing only one side is sufficient.
[0053] Calculate the gray-level gradient: Calculate the absolute value of the gray-level derivative (difference) of adjacent points in the gray-level data subset to form a gray-level gradient sequence. The magnitude of the gradient value represents the degree of gray-level change, and local maxima should appear at the boundaries of the gradient.
[0054] Dynamic feature point identification: First, a dynamic gradient threshold is set based on the grayscale standard deviation of the image background region. Then, all peak points exceeding this threshold are searched in the gradient sequence. Next, the continuity of these candidate peak points is checked, and isolated noise points that are too far from adjacent candidate points (e.g., more than 2 pixels) are eliminated. Finally, among the remaining spatially continuous candidate points, the two points with the largest and second largest gradient values are selected. The point B with the largest gradient corresponds to the boundary between the scale layer and the external medium (or background) (the second feature point), and the point C with the second largest gradient corresponds to the boundary between the scale layer and the platinum electrode substrate (the first feature point).
[0055] Step S5: Thickness Calculation The purpose of this step is to convert pixel distance into physical thickness.
[0056] Calculate pixel distance: In the image coordinate system, calculate the pixel distance between the first feature point C and the second feature point B. .
[0057] Obtaining calibration coefficients: The system needs to be spatially calibrated beforehand. This involves using a known precise physical dimension. A standard calibration block (e.g., 10.00 mm) is placed at the testing station, its X-ray image is acquired, and the pixel distance of the corresponding size in the image is measured. Then the pixel size calibration coefficient (Unit: mm / pixel).
[0058] Angle compensation: If the X-ray beam is not perpendicular to the platinum electrode surface, there exists an imaging tilt angle. This angle can be determined by the equipment's mechanical structure or visual calibration. The actual thickness perpendicular to the surface is then... It needs to be corrected to: .
[0059] Physical model-based solution: As a more accurate or verification method, calculations can be performed based on an X-ray attenuation model. The grayscale of the clean substrate region is measured separately on the image. grayscale of scaled areas And, taking into account the mass attenuation coefficient of the scaling material (such as aluminum hydroxide as the main component) on the X-ray energy used, it is converted into a linear attenuation coefficient. Final thickness You can use the formula: Calculation. This result can be corroborated with the result of the geometric calibration method, improving its reliability. It can be obtained by looking up tables or experimental calibration based on material composition and X-ray energy. I0 is obtained by calculating the mean value of pixel grayscale in the corresponding area of the image, and I is obtained by calculating the mean value of pixel grayscale in the central area of the scale layer in the image.
[0060] Output: Finally, the software interface displays the calculated actual physical thickness of the scale layer (unit: μm or mm).
[0061] The implementation of this method relies on an X-ray nondestructive testing system, such as... Figure 2 As shown, the system mainly includes: an X-ray source 1, a 16-bit digital flat panel detector 4, and a computer. A platinum electrode 3 is installed in a water pipe 2.
[0062] During the test, X-ray source 1 is turned on, and the emitted X-rays penetrate water pipe 2 containing platinum electrodes. Digital flat panel detector 4 receives the penetrated X-rays and converts them into digital image signals, which are transmitted to a computer (such as a laptop) via wired or wireless means.
[0063] The computer runs the software program of this detection method and receives digital images from the detector. The software then automatically executes all steps: image preprocessing (S1), interactive or automatic determination of the measurement area (S2), extraction and analysis of grayscale data (S3, S4), and finally calculation of the physical thickness of the scale (S5). The calculation results can be displayed in real time, and status prompts such as "qualified" or "unqualified" can be added according to preset safety thresholds.
[0064] Example 2 The similarities to Example 1 will not be repeated here; the difference lies in the addition of the following steps: S6: Result Validation and Parameter Self-Optimization This step is used to improve the system's adaptability and long-term stability.
[0065] The system can store historical qualified measurement data or preset experience thresholds for reference. When the deviation between the current measurement result and the reference value exceeds a preset range (e.g., 5%), the system can determine that the current imaging conditions or parameters may be unsatisfactory. At this time, the system automatically initiates an optimization process: fine-tuning parameters such as the Gaussian filter kernel size, contrast enhancement gain coefficient, or dynamic gradient threshold in step S1, and then re-executing steps S1 to S5 for measurement until the results stabilize within a reasonable range. This process reduces manual intervention and ensures the consistency of detection results under different operating conditions.
[0066] The above are specific embodiments of the present invention, which demonstrate the substantial features and progress of the present invention. Equivalent modifications can be made to them according to actual usage needs, under the guidance of the present invention, and all such modifications are within the scope of protection of this solution.
Claims
1. A method for detecting the scale thickness on a platinum electrode of a converter valve based on image grayscale analysis, characterized in that, Includes the following steps: S1. Acquire an X-ray transmission digital image of the platinum electrode to be tested, and perform noise reduction and contrast enhancement processing on the digital image to obtain a preprocessed image; S2. On the preprocessed image, determine a measurement line segment that passes through the scaled area on the electrode surface; S3. Extract the grayscale values of all pixels on the measured line segment to form grayscale distribution data; S4. Analyze the grayscale distribution data to identify the first and second feature points that characterize the boundary between the scale layer and the electrode substrate; S5. The actual physical thickness of the scale layer is calculated based on the pixel distance between the first feature point and the second feature point on the measurement line segment, and the pixel size calibration coefficient of the system.
2. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 1, characterized in that, In step S1, the X-ray transmission digital image is acquired by a 16-bit depth flat panel detector; the noise reduction and contrast enhancement processing includes smoothing the image using Gaussian filtering.
3. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 2, characterized in that, In step S1, the contrast enhancement process further includes performing a morphological top-hat transformation on the denoised image to correct background inhomogeneity and highlight the edge features of the scaled area.
4. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 2, characterized in that, In step S1, the contrast enhancement process further includes: obtaining the gray-scale mean value of the non-scalded area in the image as a reference value, dividing the gray-scale range of the image into multiple intervals based on the reference value, applying different gray-scale transformation coefficients to different intervals, and finally normalizing the result to a preset range.
5. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 4, characterized in that, In step S1, the multiple intervals include a first interval below the reference value, a second interval around the reference value, and a third interval above the reference value; different linear transformation coefficients are used to stretch the grayscale of the first interval and the third interval respectively, while the grayscale of the second interval is kept unchanged or subjected to low-intensity transformation.
6. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 1, characterized in that, Step S4 specifically includes: S4.
1. In the grayscale distribution data, find the point with the minimum grayscale value and mark it as the point with the thickest scale. S4.
2. Based on the thickest point of scale, select a subset of grayscale distribution data on one side of it; S4.
3. Calculate the sequence of absolute values of the gray-level differentials of adjacent data points in the gray-level distribution data subset; S4.
4. In the sequence of absolute differential values, determine the two extreme points with the largest and second largest values, and mark them as the first feature point and the second feature point, respectively.
7. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 1, characterized in that, In step S5, the actual physical thickness T is calculated or equivalently converted based on the X-ray attenuation model using the following formula: in, The linear attenuation coefficient of the scaling material for X-rays of a specific energy is given. This is the reference grayscale value for the scale-free electrode substrate area. This represents the grayscale value of the scaled area.
8. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 1, characterized in that, In step S5, the pixel size calibration coefficient of the system is used in conjunction with the imaging angle parameter. The actual physical thickness is obtained by multiplying the pixel distance by the pixel size calibration coefficient and then dividing by the cosine value of the imaging angle.
9. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 8, characterized in that, The pixel size calibration coefficient is obtained by taking an X-ray image of a standard calibration block with known physical size and calculating the ratio of the pixel size of the calibration block in the image to its actual physical size.
10. The method for detecting the scale thickness of a platinum electrode in a converter valve based on image grayscale analysis according to claim 1, characterized in that, It also includes step S6: comparing the calculated actual physical thickness with the pre-stored reference threshold or historical measurement data. If the deviation exceeds the allowable range, the noise reduction parameter or contrast enhancement parameter in step S1, or the boundary recognition parameter in step S4, or a combination thereof, are automatically adjusted, and the scale thickness detection process is re-executed based on the adjusted parameters.