Blast furnace molten iron stream diameter online measurement method and system based on improved Hough transformation

Abstract

The embodiment of the invention provides a blast furnace molten iron stream diameter online measurement method and system based on improved Hough transformation. The method is applied to the technical field of smelting intelligent manufacturing, and comprises the following steps: acquiring a real-time molten iron stream video stream of a blast furnace tap hole site, and preprocessing the molten iron stream video stream to obtain a group of first images; setting that the boundary of the molten iron stream in each frame of the first image meets parabolic characteristics based on the non-rigidity of the molten iron stream, and determining a dynamic region of interest in the first image according to the parabolic characteristics; performing edge enhancement on the dynamic region of interest by using adaptive bilateral filtering, and performing edge detection on the dynamic region of interest after edge enhancement; utilizing an improved Hough algorithm to extract two boundary parabolas of continuous boundaries of the molten iron stream; and calculating the vertical diameter between the continuous boundaries of the molten iron stream according to the boundary parabola. The method is high in intelligent degree, the abnormal tapping state can be automatically recognized, and decision support is provided for process adjustment.

Description

Technical Field

[0001] The present disclosure relates to the technical field of smelting intelligent manufacturing, and in particular to an online measurement method and system for the diameter of blast furnace molten iron strands based on improved Hough transform. Background Art

[0002] Tapping is a crucial step in blast furnace production. Continuous, orderly, and stable slag discharge is crucial for ensuring efficient blast furnace production. The diameter of the molten iron stream is determined by the diameter of the drill bit used to create the taphole. Its variation during the tapping process is also influenced by factors such as the slag-iron ratio, slag composition, taphole mud quality, taphole depth, hearth pressure, and slag fluidity. The changes in slag stream diameter during tapping can be used to infer various parameters, including the distribution and flow of slag within the hearth, the quality of the furnace head plugging operation, and the quality of the taphole mud.

[0003] In the current blast furnace environment, the diameter of the molten iron strand is mainly estimated by manual visual inspection, which has the following problems: (1) The harsh environment of high temperature, strong light, and high dust at the blast furnace tapping site results in low accuracy and poor stability of traditional visual measurement methods; (2) There are problems such as strong subjectivity, low accuracy, and inability to continuously monitor. In addition, the high temperature and strong radiation environment poses a threat to the health of operators and poses serious safety hazards.

[0004] Patent application number CN202110437293.X discloses a method for monitoring the state of a blast furnace taphole. This method uses molten iron area and time parameters to obtain taphole size information. However, this method can only indirectly reflect taphole status and cannot directly measure strand diameter. Furthermore, it is significantly affected by strong light and dust interference. Another patent application number, CN201910388544.2, discloses an online method for detecting the state of blast furnace slag iron. This method uses an imaging device to process images of the iron trough in the tapping area. However, this method can only identify three basic tapping states and cannot provide accurate strand diameter data.

[0005] Blast furnace iron strands are characterized by high temperature, strong light radiation, rapid dynamic changes, and complex surface textures, making conventional edge detection and measurement algorithms difficult to directly apply. Furthermore, the strands appear as non-rigid parabolic boundaries in images, and their shape is constantly changing due to factors such as the taphole shape and furnace pressure fluctuations, making real-time online measurement of strand diameter impossible. Therefore, developing a machine vision method that can accurately measure strand diameter in the harsh blast furnace environment has important industrial applications.

[0006] In view of the above problems, the present disclosure provides an online measurement method for the diameter of molten iron strands in a blast furnace based on an improved Hough transform algorithm. Summary of the Invention

[0007] The present disclosure provides a solution to the problem in the prior art that the diameter of the molten iron strand in the blast furnace cannot be calculated accurately and in real time due to the influence of non-rigidity.

[0008] According to a first aspect of the present disclosure, a method for online measurement of blast furnace molten iron strand diameter based on an improved Hough transform algorithm is provided. The method comprises: Acquiring a real-time video stream of molten iron streams at a blast furnace taphole site, and preprocessing the video stream of the molten iron streams to obtain a set of first images; Setting the boundary of the molten iron stream in each frame of the first image to satisfy a parabolic characteristic based on the non-rigidity of the molten iron stream, and determining a dynamic region of interest in the first image according to the parabolic characteristic; Using adaptive bilateral filtering to perform edge enhancement on the dynamic region of interest, and using an improved Canny operator to perform edge detection on the edge-enhanced dynamic region of interest to obtain a continuous boundary; The improved Hough algorithm is used to extract two boundary parabolas of the continuous boundary of the molten iron stream; The vertical diameter between the consecutive boundaries of the molten iron stream is calculated based on the boundary parabola, and the vertical diameter is converted into an actual diameter.

[0009] According to the aspects and any possible implementation methods described above, an implementation method is further provided, which also includes: the preprocessing includes: optical filtering the collected video stream of molten iron flow, and then performing image enhancement processing; the image enhancement processing includes illumination compensation based on Retinex theory and dust processing based on a symmetric zero area algorithm.

[0010] According to the above aspect and any possible implementation, an implementation is further provided, wherein determining the dynamic region of interest in the first image according to the parabolic characteristic includes: Determining the molten iron stream region in each frame of the first image using an adaptive threshold method; The molten iron stream area is dynamically adjusted according to the parabolic characteristics to obtain a dynamic region of interest.

[0011] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the edge detection of the dynamic region of interest after edge enhancement is performed using an improved Canny operator, comprising: Calculating the gradient magnitude and direction of the dynamic region of interest; Determine an adaptive gradient high threshold and an adaptive gradient low threshold according to the mean and standard deviation of the gradient amplitude; The candidate edge points are obtained by using high threshold, low threshold and non-maximum suppression, and the candidate edge points and the points corresponding to the adaptive gradient high threshold and adaptive gradient low threshold are connected to form a continuous boundary.

[0012] According to the above aspects and any possible implementation, an implementation is further provided, wherein the method of obtaining candidate edge points by using a high threshold, a low threshold, and non-maximum suppression includes: Non-maximum suppression is used to check the gradient direction of each pixel in the dynamic region of interest, and the local maximum points in the gradient direction are retained to obtain a set of discrete candidate edge points.

[0013] According to the above aspects and any possible implementation, an implementation is further provided, wherein the improved Hough algorithm specifically includes the following steps: Discretize the parameter space of the parabola parameters; For each candidate edge point, the parameter in the parameter space whose gradient direction is consistent with the parabola tangent direction is selected for voting; Perform hierarchical clustering on the peaks formed after parameter voting in the parameter space to obtain the boundary parabolas on both sides of the molten iron stream; The boundary parabolas on both sides of the molten iron stream are fitted and optimized.

[0014] According to the above aspects and any possible implementation, a further implementation is provided, wherein the peaks formed after voting on the parameters in the parameter space are hierarchically clustered to obtain the boundary parabolas on both sides of the molten iron stream, including: Perform 3D Gaussian smoothing on the parameters in the parameter space; Finding Satisfaction V p >V thresh The local maximum point of V p Indicates the number of votes at the peak point, V thresh Indicates the preset threshold for filtering peaks; Each local maximum point is used as the initial cluster, and the closest clusters are aggregated until the distance between the initial clusters is greater than the preset minimum separation threshold; The cluster with the largest number of votes in the aggregated initial cluster is selected as the boundary parabola on both sides of the molten iron stream.

[0015] According to the above aspects and any possible implementation, there is further provided an implementation, wherein the calculating the vertical diameter between consecutive boundaries of the molten iron stream according to the boundary parabola includes: Select N horizontal coordinates at equal intervals in the dynamic region of interest, and calculate the vertical coordinates on the boundary parabolas on both sides corresponding to each horizontal coordinate, where N is a non-zero natural number; Calculate the absolute value of the difference between the two vertical coordinates to obtain the vertical diameter; Converting the vertical diameter into an actual physical diameter according to calibration parameters, wherein the calibration parameters include the pitch angle of a video acquisition device for acquiring a real-time video stream of molten iron streams at a blast furnace tapping site and the pixel size of the video stream of the molten iron streams; The actual physical diameter is dynamically weighted according to the position to obtain the actual diameter.

[0016] According to the above aspects and any possible implementation, there is further provided an implementation, further comprising: Extracting time series features from the time series data constructed from the actual diameter; the time series features include statistical features, dynamic features, and frequency domain features; Input the time series features into the constructed random forest classification model and output the abnormal status and confidence level.

[0017] According to a second aspect of the present disclosure, there is provided an online measurement system for blast furnace molten iron strand diameter based on an improved Hough transform algorithm, comprising: Video stream acquisition module: used to obtain a real-time video stream of molten iron stream at the blast furnace tapping site, and pre-process the video stream of the molten iron stream to obtain a set of first images; an interest confirmation module connected to the video stream acquisition module, configured to set, based on the non-rigidity of the molten iron stream, a parabolic characteristic of the molten iron stream boundary in each frame of the first image, and determine a dynamic region of interest in the first image according to the parabolic characteristic; a continuous boundary confirmation module, connected to the interest confirmation module, configured to perform edge enhancement on the dynamic region of interest using adaptive bilateral filtering, and perform edge detection on the dynamic region of interest after edge enhancement using an improved Canny operator to obtain a continuous boundary; a parabola confirmation module, connected to the continuous boundary confirmation module, for extracting two boundary parabolas of the continuous boundary of the molten iron stream using an improved Hough algorithm; The strand diameter calculation module is connected to the parabola confirmation module and is used to calculate the vertical diameter between the continuous boundaries of the molten iron strand according to the boundary parabola and convert the vertical diameter into an actual diameter.

[0018] The beneficial effects of the present disclosure are: This paper removes dust from the collected video stream of molten iron streams. It first uses a symmetric zero-area algorithm to perform peak analysis to determine whether each image in the video stream has areas covered by dust. For areas with bimodal characteristics, an improved OTSU algorithm is used for segmentation. In the present disclosure, the boundary of the molten iron stream is set as a parabolic model based on its non-rigid characteristics, and the dynamic region of interest of the molten iron stream is determined based on the characteristic that the molten iron stream is always in dynamic motion, focusing on the effective region and reducing the amount of subsequent calculations; The present disclosure takes into account that the overall trajectory of the molten iron stream is relatively stable, but it will be disturbed by the changes in the internal air pressure of the high-pressure donkey, resulting in a certain degree of jitter. Therefore, an improved Canny operator is used for edge detection, and the adaptive gradient threshold is dynamically calculated to enhance the overall adaptability. The present invention extracts boundary parabolas through an improved Hough algorithm. The parameter space in the Hough algorithm is discretized according to the physical properties of the molten iron stream, reducing invalid voting. Furthermore, a gradient direction consistency constraint is introduced, combined with a weighted voting strategy, to improve parabola fitting accuracy while reducing parabola extraction time. The present invention adopts an improved Hough transform algorithm, and the strand diameter measurement error is less than 1mm, which meets various control requirements derived from subsequent industries; it has strong real-time performance, and the processing time of a single frame image is less than 50ms, which can realize real-time online measurement of the molten iron strand diameter; it has strong environmental adaptability, and through optical filtering and image enhancement algorithms, it can effectively overcome the interference of high temperature, strong light and dust; the present invention is a completely non-contact measurement, avoiding the safety risks of manual approach to high-temperature molten iron; the present invention has a high degree of intelligence, can automatically identify abnormal iron tapping conditions, and provide decision support for process adjustments.

[0019] It should be understood that the contents described in the Summary of the Invention section are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are provided for a better understanding of the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, among which: Figure 1 A flow chart of an online measurement method for blast furnace molten iron strand diameter based on improved Hough transform is shown; Figure 2 Shows a schematic diagram of the on-site photo of the blast furnace molten iron flow. DETAILED DESCRIPTION

[0021] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present disclosure.

[0022] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0023] In the present disclosure, a method for online measurement of blast furnace molten iron strand diameter based on improved Hough transform is provided. Figure 1 ,include: S1. Acquire a real-time video stream of molten iron streams at a blast furnace taphole, and preprocess each frame of the video stream to obtain a set of first images. See also Figure 2 The iron mouth is surrounded by a high-temperature, high-brightness, and high-smoke environment. The naked eye cannot directly observe the state of the iron flow on site, so a high-definition industrial camera is required for video acquisition.

[0024] It should be noted that high-definition industrial cameras need to be placed on both sides of the molten iron stream, and the angle between their optical axis and the flow direction of the stream should be as close to 90° as possible, and generally at least greater than 45°.

[0025] To observe the molten iron stream, a high-definition industrial camera needs to be installed close to the iron outlet. The high-definition industrial camera must be cooled and subjected to conventional dust-proof treatment. Therefore, a protective cover made of high-temperature resistant material is used to surround the high-definition industrial camera, and a unique cooling structure is configured to ensure that the high-definition industrial camera can work continuously and stably. The temperature range of the molten iron stream is 1450℃~1550℃. The molten iron stream is a dazzling white light with high-temperature infrared radiation. Conventional optical systems cannot directly collect and process it. The present invention is configured with a filter to suppress strong light and infrared rays.

[0026] The equipment must be installed at a location that allows for real-time viewing of the molten iron stream. After the HD industrial camera is installed, it begins capturing on-site images, i.e., the molten iron stream.

[0027] The preprocessing of the molten iron stream video stream specifically includes the following: The collected video stream of the molten iron stream is optically filtered to remove most of the visible light and near-infrared interference; and then image enhancement processing is performed, which includes illumination compensation and dust processing.

[0028] (1) Light compensation An illumination estimation and compensation algorithm based on Retinex theory is used to eliminate the effect of uneven illumination. The formula is as follows: R(x,y)=logI(x,y)-log[F(x,y)*I(x,y)], Where R(x, y) represents the final extracted reflection component, that is, the target information after removing uneven illumination); I(x, y) represents the pixel value of the original image at the coordinate (x, y), such as the brightness or color channel value; F(x, y) represents the weight value of the Gaussian filter kernel at (x, y), which is used to simulate local illumination (low-frequency component); * represents the convolution operation of the image I and the Gaussian kernel F to obtain an estimate of the illumination component; x and y represent the spatial coordinates of the pixel in the image, that is, the row and column positions of the pixel in the image. x usually represents the horizontal direction, that is, the column index, and y usually represents the vertical direction, that is, the row index.

[0029] (2) Dust removal The symmetric zero area algorithm is used to perform peak analysis to determine whether each image in the video stream has an area covered by dust. The traditional OTSU algorithm maximizes the inter-class variance σ 2 Determine the optimal segmentation threshold T, the formula is: , in, is the current threshold Next, the variance between foreground and background; Indicates the foreground, that is, the percentage of pixels in the molten iron stream area or without dust, Represents the background, that is, the pixel ratio of dust or non-iron flow area; is the candidate grayscale threshold traversed by the algorithm during the calculation process; is the ratio of foreground pixels, that is, the gray value is less than or equal to The ratio of the number of pixels to the total number of pixels; is the percentage of pixels with background value, i.e. the gray value is greater than The ratio of the number of pixels to the total number of pixels; is the average gray value of the foreground, is the average gray value of the background.

[0030] For regions with bimodal characteristics, low grayscale peaks correspond to dust areas, while high grayscale peaks correspond to dust-free areas. The bimodal distribution of dust images may cause the optimal threshold of the traditional OTSU algorithm to be biased towards the low grayscale peak, i.e., the dust area, thereby misclassifying dust-free areas as dust areas.

[0031] Therefore, the present disclosure uses the improved OTSU algorithm for segmentation, and the specific method is: First, a symmetric zero-area algorithm is used to analyze the histogram to identify dust-covered areas. This helps determine the location of the double peaks, improves robustness to dust interference, and ensures that the segmentation results are more accurate for the actual molten iron stream area. By using histogram analysis to determine the location of the double peaks, the threshold search is restricted to the valley between the double peaks and the high-grayscale peak to avoid interference from low-grayscale peaks. A weighting factor is added to the threshold calculation to prioritize high-grayscale areas, ensuring that the molten iron stream area is not incorrectly segmented by dust noise.

[0032] Through the improved OTSU algorithm, the threshold search range is limited, the weight of the high grayscale area is optimized, and the dust area is pre-identified, which effectively solves the bimodal distribution missegmentation problem caused by dust in the traditional OTSU algorithm.

[0033] In addition, each frame of the preprocessed first image may be subjected to angle correction and scale calibration. The angle correction is used to adjust the direction of the image through geometric transformation, and the scale calibration is used to determine the proportional relationship between the pixels in the image and the actual physical size.

[0034] S2, setting the boundary of the molten iron stream in each frame of the first image to be a parabola based on the non-rigidity of the molten iron stream, and determining a dynamic region of interest in the first image according to the parabola characteristic; The shape of a molten iron stream changes over time, and traditional rigid boundary models cannot accurately describe its boundaries. Therefore, setting the stream boundary to a parabola effectively captures its gravity-dominated flow pattern. Furthermore, the parameters in the parabola equation can change over time to reflect the dynamic adjustment of the stream boundary. Only three parabola parameters are required, reducing computational complexity and improving real-time performance.

[0035] Through experiments, the parabola model, cubic curve model and straight line model were compared and verified, and the parabola model can achieve a balance between accuracy and efficiency.

[0036] Affected by many factors such as the gas pressure in the blast furnace, the blast volume, the molten iron composition, and the iron trough design, the pressure in the iron mouth area will change instantaneously, changing the flow rate of the molten iron stream, causing the molten iron stream to be in a jumping state, that is, an unstable state. After the molten iron stream falls into the main trough, it is still in a high-temperature state. Accurately obtaining the dynamic area of ​​interest of the molten iron stream faces great interference.

[0037] The present disclosure specifically realizes the extraction of the dynamic region of interest of the molten iron stream through the following method.

[0038] (1) Determine the molten iron flow area Due to the high temperature characteristics of the molten iron stream, the thermal radiation intensity is high, which will appear as a high-brightness area in the image, while the background brightness is low. Therefore, the molten iron stream area can be divided. The present disclosure uses an adaptive threshold method to determine the molten iron stream area in each frame of the first image, and dynamically calculates the threshold based on the brightness characteristics of the local area of ​​the image to divide the image into foreground and background, where the foreground is the molten iron stream area, specifically: ; Where T is the segmentation threshold, I is the input grayscale image matrix, representing the high-temperature area image of the molten iron stream acquired by the camera; max(I) is the maximum grayscale value in image I, reflecting the intensity of the high-brightness area in the image; OTSU(I) is the optimal segmentation threshold T calculated by the previous OTSU algorithm, which is used to distinguish between the background and the molten iron area; mean(I) is the average grayscale value of image I, reflecting the overall brightness level; α, β, and γ are all dynamic weight coefficients used to balance the contributions of different threshold components to meet the normalization constraint, and α+β+γ=1. Specifically, α is the weight controlling the maximum grayscale value, β is the weight controlling the OTSU threshold, and γ is the weight controlling the average grayscale value.

[0039] (2) Dynamically adjust the molten iron flow area The molten iron stream has two boundaries, so there are two corresponding boundary parabolas, and the two parabolas correspond to different points as the starting point coordinates.

[0040] Depending on the volume of the blast furnace, the taphole diameter is between 40mm and 65mm. A coordinate system is constructed in the first image, with the origin of the coordinate system located in the upper left corner of the first image. The starting point is not the origin, and the coordinate system of each frame of the first image is the same.

[0041] Determine the starting point coordinates P of the dynamic region of interest respectively s : P s =(i, 0.5(y t(i) +y b(i) )), Among them, i represents the horizontal coordinate, which is, y t(i) is the ordinate value of the upper boundary parabola at point i, y b(i) is the vertical coordinate value of the lower boundary parabola at point i, with P s As the starting point, because the molten iron stream is always in dynamic motion, the dynamic region of interest is dynamically expanded along the parabola direction, that is, the parabola is based on the starting point coordinates and the monitoring area is adjusted in real time according to the adaptive threshold method to obtain the dynamic region of interest.

[0042] In the previous section, the adaptive threshold method is used to detect the position of the molten iron strand area in the first image of each frame in real time. After segmentation, the horizontal range of the molten iron strand area is obtained. i is the starting horizontal coordinate of the area in the image coordinate system. Because it is recalculated in each frame and dynamically adjusted as the position of the molten iron strand changes, y t(i) and y b(i) are the ordinate values ​​of the upper and lower boundary parabolas of the current frame fit at x = i. Each fit is independent and corresponds to a new starting point, so the starting point coordinates are not fixed. In this disclosure, the starting point coordinates are defined as the dynamic midpoint of the left boundary of the stream region, that is, the midpoint of the upper and lower boundary coordinates at x = i.

[0043] In summary, the above operations are performed on the first image of each frame. Since the molten iron stream is constantly moving, a dynamic region of interest can be obtained. The dynamic region of interest is determined in real time, the effective area is focused, the subsequent calculation amount is reduced, and the processing speed is improved.

[0044] S3, using adaptive bilateral filtering to perform edge enhancement on the dynamic region of interest, and using an improved Canny operator to perform edge detection on the dynamic region of interest after edge enhancement; (1) Adaptive bilateral filtering is used to smooth the noise while retaining the edge information, that is, edge enhancement: , in, Represents the adaptive output pixel value; u and v both represent pixels, Indicated in pixels u The neighborhood window centered at v is any pixel within the window; represents the adaptive normalization factor, Represents a spatial Gaussian kernel, where the weight decays with distance; Indicates the degree of edge preservation. Represents pixels u and v The spatial distance, Represents the spatial standard deviation, which is used to control the degree of spatial smoothing; represents the intensity (range) Gaussian kernel, where the weight decays with the difference in pixel values; Indicates the standard deviation of the range, which controls the decay rate of the intensity weight; Represents pixels u and v The grayscale difference, represents the grayscale of pixel v, Represents the grayscale of pixel u.

[0045] The above steps preserve the edges while suppressing high-temperature dust noise. Due to the pulsation of the molten iron stream, that is, the unstable fluctuations that occur during the flow process, it is necessary to perform multi-scale edge enhancement on the parabola formed by the stream to facilitate subsequent edge calculations.

[0046] (2) Using the improved Canny operator for edge detection, the specific steps include: 1. Based on the coordinate system constructed when determining the coordinates of the parabola starting point, calculate the gradient magnitude and direction of the dynamic region of interest: , , Among them, G x and G y Represent the gradient components in the horizontal and vertical directions respectively, G represents the gradient amplitude, Indicates the gradient direction.

[0047] 2. Determine the adaptive gradient high threshold and adaptive gradient low threshold based on the mean and standard deviation of the gradient amplitude: , , in, It represents the mean value of the gradient amplitude, reflecting the overall edge strength; represents the standard deviation of the gradient amplitude, which is used to measure the discreteness of the gradient distribution; k represents the adjustment coefficient, which can be adjusted according to the noise level; T high It represents the adaptive gradient high threshold, which is the threshold for edge points. It dynamically adapts to strong light and dust interference, improving the accuracy. low Represents the adaptive gradient low threshold, which is the threshold of suspected edge points.

[0048] The molten iron stream is a high-temperature fluid. It will fluctuate due to the pressure change inside the blast furnace, but the overall trajectory is relatively stable. high and T high Dynamic calculation can more easily adapt to the drastic fluctuations of gradients in high temperature scenarios, and setting the adjustment coefficient enhances adaptability.

[0049] 3. Use high threshold, low threshold and non-maximum suppression to obtain candidate edge points, and connect the candidate edge points and the points corresponding to the high threshold and low threshold to form a continuous boundary.

[0050] Non-maximum suppression is used to examine the gradient direction of each pixel in the dynamic region of interest, retain the local maximum points in the gradient direction, and select the local maximum points between the high and low thresholds as a set of discrete candidate edge points. Non-maximum suppression can eliminate the rough line effect of the edge, such as turning a 3-pixel wide bright band into a 1-pixel wide point chain, thereby obtaining a precise single-pixel edge.

[0051] The obtained candidate edge points are connected with the double threshold, and the discrete points are strung together into a complete curve to obtain a continuous boundary.

[0052] S4. Use the improved Hough algorithm to extract two boundary parabolas of the continuous boundary of the molten iron stream.

[0053] The improved Hough algorithm specifically includes the following steps: (1) Discretize the parameter space of the parabola parameters The general equation of the parabola y=ax 2 +bx+c parameter space (a, b, c) is discretized.

[0054] The velocity of the molten iron stream when it is ejected from the high-pressure furnace through the molten iron channel to the external open space at a certain speed is limited, the curvature of the trajectory is small, and the initial velocity direction when ejected is close to the horizontal direction. Considering the above physical characteristics, the parameter range is constrained: a∈[-0.01,0.01], b∈[-1,1], c∈[0,H].

[0055] Wherein, H is the height of the first image, and the discretization step size is set according to the accuracy requirement, which is not specifically limited in this disclosure.

[0056] (2) For each candidate edge point, the parameters in the parameter space whose gradient direction is consistent with the parabola tangent direction are selected for voting; The present disclosure adopts a weighted voting strategy based on local consistency, which has less computational complexity and is less susceptible to noise interference than the global voting of edge points in the traditional Hough transform.

[0057] Based on the coordinate system constructed previously, calculate each edge point e i =(x i ,y i )’s local gradient direction ; In parameter space, only the parabola parameters that satisfy directional consistency are voted: , in, Indicates that the parabola j is at x i The tangent angle at Indicates the angle tolerance.

[0058] The voting weight for each edge point is set as follows: w i =G i ·C i , Among them, G i is the gradient amplitude of the edge point, C i is the (local) curvature consistency factor within the edge point neighborhood. The voting weight adjusts the contribution of each edge point in the parameter space voting, making the edge points that are more reliable and more consistent with the local geometric characteristics have a greater impact on the final result, improving the accuracy of detection and noise resistance.

[0059] (3) Perform hierarchical clustering on the peaks formed after parameter voting in the parameter space to obtain the boundary parabolas on both sides of the molten iron stream.

[0060] 1. Perform 2D Gaussian smoothing on the parameters in the parameter space to suppress false peaks caused by noise and make the true peaks more prominent.

[0061] 2. Find the one that satisfies V p >V thresh The local maximum point of V p Indicates the number of votes for the peak point, which reflects the possibility that the point may be the true boundary. thresh Indicates the preset threshold for filtering peaks. Only those with a vote count greater than this threshold will be considered as potential boundaries. The distance between the local maximum and the detected peak point must be greater than the preset minimum separation threshold to avoid repeated detection of the same parabola. 3. Take each local maximum point as the initial cluster and aggregate the other clusters closest to it until the distance between the initial clusters is greater than the preset minimum separation threshold.

[0062] 4. Select the cluster with the most votes in the initial cluster after aggregation as the boundary parabola on both sides of the molten iron stream.

[0063] (4) Fit and optimize the boundary parabolas on both sides of the molten iron stream.

[0064] In this disclosure, the RANSAC algorithm is used for optimization: Randomly sample a minimum sample subset from the local maximum points, such as fitting a parabola with three points; use a nonlinear optimization algorithm to estimate the parabola parameters, calculate all errors to the parabola, mark points with errors less than a preset value as inliers, and count the number of inliers; iterate multiple times, select the model with the most inliers, and refit using all inliers to obtain the boundary parabola.

[0065] S5. Calculate the vertical diameter between the continuous boundaries of the molten iron stream according to the boundary parabola, and convert the vertical diameter into an actual diameter.

[0066] Select N horizontal coordinates x at equal intervals in the dynamic region of interest i , where N is a non-zero natural number; calculate the ordinates on the boundary parabolas on both sides corresponding to each horizontal coordinate, calculate the absolute value of the difference between the two ordinates, and obtain the vertical diameter D i : D i =y b (xi) -y t (xi) , Among them, y t(xi) Indicates that the upper boundary parabola is at x i The vertical coordinate value at y b(xi) Indicates that the lower boundary parabola is at x i The vertical coordinate value at x i Indicates the corresponding horizontal coordinate value at the i-th position.

[0067] Convert the vertical diameter into the actual physical diameter D according to the calibration parameters reali The calibration parameters include the pitch angle of the video acquisition device for obtaining the real-time video stream of the molten iron stream at the blast furnace tapping hole and the pixel size of the molten iron stream video stream: D reali =D i ·s·cosα, Where s is the pixel size in mm / pixel, and α is the pitch angle of the industrial camera.

[0068] Taking into account the fluid characteristics of the molten iron stream, the diameter measurement values ​​at different positions are dynamically weighted to resolve measurement errors caused by factors such as viewing angle tilt, edge noise, and fluid fluctuations. The calculation formula for dynamic weighting is:

[0069] in, is the actual diameter finally obtained, which represents the vertical diameter measurement value (pixel unit) at the i-th position, s represents the pixel size (mm / piex), which is determined by camera calibration; α represents the camera pitch angle, represents the dynamic weight of the i-th position. The dynamic weight is determined by the position reliability and the flow characteristics. In one embodiment, the position reliability accounts for 60% of the weight and is set based on the edge detection confidence and view geometry correction. The flow characteristics account for 40% of the weight and is set based on the flow stability and spatial continuity.

[0070] In addition, the present disclosure also includes abnormal state detection. First, time series features are extracted from the actual diameter time series data. The time series features include statistical features, dynamic features, and frequency domain features. Statistical features include mean (characterizing the average level of flow diameter), variance (measuring the degree of diameter fluctuation; a sudden increase in variance may indicate flow instability or breakage), skewness (the process distribution of diameter values, whether it is too large or too small), and kurtosis (the occurrence of local extreme values). Dynamic features include short-term rate of change (the rate of change of diameter between specific frames) and acceleration (reflecting the acceleration / deceleration trend of diameter change). Frequency domain features include the energy proportion of the main frequency components (periodic fluctuation characteristics are extracted through Fourier transform).

[0071] Construct a classification model based on random forest, input the extracted time series features into the constructed classification model, and identify abnormal conditions such as "large flow", "coke jam", "splashing", and "gas leakage" and their confidence levels.

[0072] Based on the above technical solutions, the present disclosure adopts an improved Hough transform algorithm, and the flow diameter measurement error is less than 1mm, which meets various control requirements derived from subsequent industries. The present invention has strong real-time performance, and the processing time of a single frame image is less than 50ms, which can realize the real-time online measurement of the diameter of the molten iron strand; The disclosed method has strong environmental adaptability and effectively overcomes interference from high temperature, strong light, and dust through optical filtering and image enhancement algorithms; The present disclosure has good safety and is a completely non-contact measurement, avoiding the safety risk of manual approach to high-temperature molten iron; The present invention has a high degree of intelligence and can automatically identify abnormal iron-making conditions, providing decision support for process adjustments.

[0073] The present disclosure also provides an online measurement system for the diameter of molten iron strands in a blast furnace based on an improved Hough transform, comprising the following modules: Video stream acquisition module: used to obtain a real-time video stream of molten iron stream at the blast furnace tapping site, and pre-process the video stream of the molten iron stream to obtain a set of first images; an interest confirmation module connected to the video stream acquisition module, configured to set, based on the non-rigidity of the molten iron stream, a parabolic characteristic of the molten iron stream boundary in each frame of the first image, and determine a dynamic region of interest in the first image according to the parabolic characteristic; a continuous boundary confirmation module, connected to the interest confirmation module, configured to perform edge enhancement on the dynamic region of interest using adaptive bilateral filtering, and perform edge detection on the dynamic region of interest after edge enhancement using an improved Canny operator to obtain a continuous boundary; a parabola confirmation module, connected to the continuous boundary confirmation module, for extracting two boundary parabolas of the continuous boundary of the molten iron stream using an improved Hough algorithm; The strand diameter calculation module is connected to the parabola confirmation module and is used to calculate the vertical diameter between the continuous boundaries of the molten iron strand according to the boundary parabola and convert the vertical diameter into an actual diameter.

[0074] For other contents, please refer to the section on online measurement method of blast furnace molten iron strand diameter based on improved Hough transform, which will not be elaborated here.

[0075] The present disclosure also provides a device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above method.

[0076] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0077] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0078] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0079] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0080] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0081] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. An online measurement method for blast furnace molten iron strand diameter based on improved Hough transform, including: Acquiring a real-time video stream of molten iron streams at a blast furnace taphole site, and preprocessing the video stream of the molten iron streams to obtain a set of first images; Setting the boundary of the molten iron stream in each frame of the first image to satisfy a parabolic characteristic based on the non-rigidity of the molten iron stream, and determining a dynamic region of interest in the first image according to the parabolic characteristic; Using adaptive bilateral filtering to perform edge enhancement on the dynamic region of interest, and using an improved Canny operator to perform edge detection on the edge-enhanced dynamic region of interest to obtain a continuous boundary; The improved Hough algorithm is used to extract two boundary parabolas of the continuous boundary of the molten iron stream; The vertical diameter between the consecutive boundaries of the molten iron stream is calculated based on the boundary parabola, and the vertical diameter is converted into an actual diameter.

2. The method according to claim 1, wherein The preprocessing includes: optically filtering the collected video stream of the molten iron stream, and then performing image enhancement processing; the image enhancement processing includes illumination compensation based on Retinex theory and dust processing based on a symmetric zero area algorithm.

3. The method according to claim 1, wherein The determining of the dynamic region of interest in the first image according to the parabola characteristic includes: Determining the molten iron stream region in each frame of the first image using an adaptive threshold method; The molten iron stream area is dynamically adjusted according to the parabolic characteristics to obtain a dynamic region of interest.

4. The method according to claim 1, wherein The step of performing edge detection on the dynamic region of interest after edge enhancement using an improved Canny operator includes: Calculating the gradient magnitude and direction of the dynamic region of interest; Determine an adaptive gradient high threshold and an adaptive gradient low threshold according to the mean and standard deviation of the gradient amplitude; The candidate edge points are obtained by using high threshold, low threshold and non-maximum suppression, and the candidate edge points and the points corresponding to the adaptive gradient high threshold and adaptive gradient low threshold are connected to form a continuous boundary.

5. The method according to claim 4, wherein The method of obtaining candidate edge points by using high threshold, low threshold and non-maximum suppression includes: Non-maximum suppression is used to check the gradient direction of each pixel in the dynamic region of interest, and the local maximum points in the gradient direction are retained. The local maximum points of the gradient between the high threshold and the low threshold are selected as a set of discrete candidate edge points.

6. The method according to claim 5, wherein: The improved Hough algorithm specifically comprises the following steps: Discretize the parameter space of the parabola parameters; For each candidate edge point, the parameter in the parameter space whose gradient direction is consistent with the parabola tangent direction is selected for voting; Perform hierarchical clustering on the peaks formed after parameter voting in the parameter space to obtain the boundary parabolas on both sides of the molten iron stream; The boundary parabolas on both sides of the molten iron stream are fitted and optimized.

7. The method according to claim 6, wherein: The peaks formed after voting on the parameters in the parameter space are subjected to hierarchical clustering to obtain the boundary parabolas on both sides of the molten iron stream, including: Perform 2D Gaussian smoothing on the parameters in the parameter space; Finding Satisfaction V p >V thresh The local maximum point of V p Indicates the number of votes at the peak point, V thresh Indicates the preset threshold for filtering peaks; Each local maximum point is used as the initial cluster, and the closest clusters are aggregated until the distance between the initial clusters is greater than the preset minimum separation threshold; The cluster with the largest number of votes in the aggregated initial cluster is selected as the boundary parabola on both sides of the molten iron stream.

8. The method according to claim 1, wherein Calculating the vertical diameter between the continuous boundaries of the molten iron stream according to the boundary parabola and converting the vertical diameter into the actual diameter includes: Select N horizontal coordinates at equal intervals in the dynamic region of interest, and calculate the vertical coordinates on the boundary parabolas on both sides corresponding to each horizontal coordinate, where N is a non-zero natural number; Calculate the absolute value of the difference between the two vertical coordinates to obtain the vertical diameter; Converting the vertical diameter into an actual physical diameter according to calibration parameters, wherein the calibration parameters include the pitch angle of a video acquisition device for acquiring a real-time video stream of molten iron streams at a blast furnace tapping site and the pixel size of the video stream of the molten iron streams; The actual physical diameter is dynamically weighted according to the position to obtain the actual diameter.

9. The method according to claim 1, wherein Also includes: Performing time series feature extraction on the time series data constructed from the actual diameter; The time series features include statistical features, dynamic features and frequency domain features; Input the time series features into the constructed random forest classification model and output the abnormal status and confidence level.

10. An online measurement system for blast furnace molten iron strand diameter based on an improved Hough transform algorithm, for implementing the methods described in claims 1 to 9, comprising the following modules: Video stream acquisition module: used to obtain a real-time video stream of molten iron stream at the blast furnace tapping site, and pre-process the video stream of the molten iron stream to obtain a set of first images; an interest confirmation module connected to the video stream acquisition module, configured to set, based on the non-rigidity of the molten iron stream, a parabolic characteristic of the molten iron stream boundary in each frame of the first image, and determine a dynamic region of interest in the first image according to the parabolic characteristic; a continuous boundary confirmation module, connected to the interest confirmation module, configured to perform edge enhancement on the dynamic region of interest using adaptive bilateral filtering, and perform edge detection on the dynamic region of interest after edge enhancement using an improved Canny operator to obtain a continuous boundary; a parabola confirmation module, connected to the continuous boundary confirmation module, for extracting two boundary parabolas of the continuous boundary of the molten iron stream using an improved Hough algorithm; The strand diameter calculation module is connected to the parabola confirmation module and is used to calculate the vertical diameter between the continuous boundaries of the molten iron strand according to the boundary parabola and convert the vertical diameter into an actual diameter.

Images