Vortex image information enhancement method based on molten iron stirring process

Abstract

The present invention discloses a method for enhancing vortex image information based on a molten iron stirring process, belonging to the field of metallurgical technology. The method comprises the following steps: detecting molten iron information in an image, detecting the molten iron vortex surface from a complex background through an iterative segmentation method, and obtaining a segmented molten iron tank liquid surface image; applying a Gabor filter image recognition method to the segmented molten iron tank liquid surface image to obtain a molten iron vortex image with enhanced vortex lines; predicting and enhancing the molten iron vortex image based on vortex characteristics, and acquiring lines with difference values ​​within a limited range to obtain a target molten iron vortex image. The present invention can reduce the impact of a complex background on a detection target, enhance grayscale detail differentiation by constructing a Gabor filter enhancement template, predict the current frame based on the molten iron vortex characteristics, and obtain vortex lines after an interval time t, thereby better improving the quality of the vortex image and the time proportion of the strong feature image.

Description

Technical Field

[0001] The present invention belongs to the technical field of metallurgy, and more particularly, relates to a method for enhancing vortex image information of molten iron. Background Art

[0002] Because the sulfur content cannot be measured during the KR stirring desulfurization reaction of molten iron, the desulfurization rate sometimes fails to meet the standard after stirring. In this case, multiple desulfurization cycles are required, or large amounts of desulfurizer are injected each time to achieve saturated desulfurization, resulting in wasteful desulfurization. Previously, vortex information technology has been applied to molten iron desulfurization, using image processing to identify vortexes on the liquid surface and calculate vortex diameters. Testing revealed that during a 10-minute stirring period, there were 3 minutes and 50 seconds of clear vortex images and 4 minutes of completely blurred images from the injection of desulfurizer. The original system determined the vortex radius based on the clear vortex image, but for approximately 2 minutes and 10 seconds, the vortex image was slightly obstructed by fog. Using this semi-blurred image to extend the time during which vortex diameter information can be detected has become a research direction.

[0003] In the existing technology, although data such as the vortex radius is obtained through computer processing, during the KR stirring and desulfurization process, the camera image information is easily interfered with by factors such as iron filings, light changes, and mist generated by the sprayed desulfurizer, resulting in the loss of vortex lines and difficulty in identifying line information. It is impossible to accurately know the state of the molten iron during the stirring process of the molten iron tank, and the extracted information is unlikely to bring much help to the control system. Summary of the Invention

[0004] In response to the above defects or improvement needs of the prior art, the present invention proposes a vortex image information enhancement method based on the molten iron stirring process, thereby solving the technical problems of image information being disturbed and vortex characteristics being weakened.

[0005] To achieve the above objectives, the present invention provides a method for enhancing vortex image information based on a molten iron stirring process, comprising:

[0006] (1) Detecting the molten iron information in the image, using an iterative segmentation method to detect the molten iron vortex surface from the complex background to obtain a segmented image of the molten iron tank surface;

[0007] (2) The segmented molten iron ladle liquid surface image is subjected to the Gabor filter image recognition method to obtain the molten iron vortex image after the vortex lines are enhanced;

[0008] (3) The molten iron vortex image is predicted and enhanced according to the vortex characteristics, and the lines with difference values ​​within a limited range are obtained to obtain the target molten iron vortex image.

[0009] In some optional embodiments, step (1) comprises:

[0010] (1.1) Convert the original input image into a grayscale image I, and divide the grayscale image I into several non-overlapping blocks;

[0011] (1.2) The threshold T is determined by the average gray value of the selected area, and the new threshold is calculated based on the threshold T, which divides all blocks into two groups of pixel values;

[0012] (1.3) If the difference between the new threshold T1 and the old threshold T is zero, the new threshold T1 is used as the optimal threshold for the binary image. Otherwise, the iteration is continued until the difference between the new threshold T1 and the old threshold T is zero.

[0013] (1.4) Using the optimal threshold, the grayscale image I is binarized into the grayscale image I1, and the boundary of the segmented background area of ​​the grayscale image I1 is sharpened;

[0014] (1.5) Fill the sharpened background area into the grayscale image I1 to remove the original image background and interference factors.

[0015] In some optional embodiments, Get the new threshold T1, where G1 is the total grayscale value of all block images with grayscale values ​​lower than the threshold T, G2 is the total grayscale value of all block images with grayscale values ​​higher than the threshold T, g1 is the total number of block images with grayscale values ​​lower than the threshold T, and g2 is the total number of block images with grayscale values ​​higher than the threshold T.

[0016] In some optional embodiments, step (2) includes:

[0017] (2.1) Normalizing the segmented molten iron ladle surface image and converting it into a standard image with a given mean and variance;

[0018] (2.2) The standard image is processed by the Gabor filter to obtain the grayscale image I2;

[0019] (2.3) Divide the grayscale image I2 into several non-overlapping blocks and extract the ridge direction and ridge frequency of each block;

[0020] (2.4) Obtaining the curvature of the grain lines within the block;

[0021] (2.5) A Gabor enhancement template is constructed to enhance the grayscale image I2, and the molten iron vortex image with enhanced vortex lines is obtained.

[0022] In some optional embodiments, step (2.1) comprises: The segmented molten iron tank surface image is normalized, where I(i,j) represents the segmented molten iron tank surface image, G (i,j) Represents the standard image, M0 and are the pre-set image average grayscale and mean square error, M and σ 2 is the grayscale mean and variance of the segmented molten iron ladle surface image.

[0023] In some optional embodiments, Calculate the grain direction of each block, where The horizontal gradient value of each pixel is And the vertical gradient value is The pixel coordinates of the center point of the block are (i, j), θ(i, j) is the direction angle of the grain line, w represents the major diameter of the grain line, and (u, v) represents the coordinates of (i, j) in the frequency domain.

[0024] In some optional implementation schemes, the method for extracting the ridge frequency of each block is:

[0025] Based on the obtained grain direction, the grayscale values ​​of all pixels in each block are projected in the direction perpendicular to the grain direction to form a one-dimensional sine wave. The extreme points of the one-dimensional sine wave correspond to the grain lines of the vortex. Assuming T(i,j) is the average number of pixels between two adjacent peaks of the one-dimensional sine wave, the frequency

[0026] In some optional embodiments, Get the curvature of each block, where φ x (i,j,u,v)=|cos(2θ(i,j))-cos(2θ(u,v))|,φ y (i,j,u,v)=|sin(2θ(i,j))-sin(2θ(u,v))|, where (i,j) is the coordinate of the center pixel of the block, (u,v) represents the coordinate of (i,j) in the frequency domain, C(i,j) represents the curvature of the grain line in the block, θ represents the direction angle, N represents the size of the local area where the block is located, and φ represents the direction of the Gabor filter.

[0027] In some optional embodiments, Construct a Gabor enhancement template G, where x φ =xcosφ+ysinφ,y φ =-xcosφ+ysinφ, φ is the direction of the Gabor filter, f is the frequency of the line, [x φ ,y φ ] indicates that the coordinate axis [x, y] rotates counterclockwise by an angle φ, σ x and σ y are the Gaussian envelope constants along the x and y axes, respectively.

[0028] In some optional embodiments, step (3) includes:

[0029] (3.1) Collecting the grain line information of the enhanced molten iron vortex image;

[0030] (3.2) Screening the collected grain line information and eliminating useless data;

[0031] (3.3) Predicting the ridge information of the molten iron vortex image, and predicting the ridge information of the image after the interval t;

[0032] (3.4) According to the real image ridge information after the interval time t, the vortex image processed by the real information and the predicted information is calculated, and the real ridges and the predicted ridges are distinguished and recorded.

[0033] In some optional embodiments, step (3.3) includes:

[0034] Depend on and Calculate the predicted vortex line dynamics, where s is the moving distance of the pixel coordinates (i, j) of the center point of each block, the direction is θ(i, j), a[0] is the instantaneous acceleration of the center point (i, j), w is the long diameter of the line, x is the long diameter of the predicted line, R is the distance between the center point of the block line and the center point of the molten iron vortex surface, and v is the fluid velocity;

[0035] The predicted ridge line is calculated by the center point of each block (i, j), and the pixel points with a long diameter of x are filled in front and behind the predicted coordinate point direction θ(i, j) according to the curvature C(i, j) in a single ridge line.

[0036] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0037] The image processing method of the present invention reduces the impact of complex backgrounds on the detection target and allows the molten iron vortex surface to be detected from the complex background. By constructing a Gabor filter enhancement template, grayscale detail differentiation is enhanced, improving the clarity of the vortex lines while preserving the original line structure. The current frame is predicted based on the molten iron vortex characteristics to obtain the vortex lines after an interval of time t, thereby avoiding the potential blurring of the lines after the interval of time t and further improving the quality of the vortex image. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of molten iron surface image segmentation provided by an embodiment of the present invention;

[0039] Figure 2 This is a diagram of steps for dividing an area and filling an original image provided by an embodiment of the present invention;

[0040] Figure 3This is an enhanced swirl line process effect diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0042] In this embodiment, the present invention is based on a method for obtaining enhanced vortex image information during the KR desulfurization stirring process, namely, a method for enhancing and identifying vortex image information based on the direction and curvature of vortex lines and vortex characteristics, including:

[0043] Step 101): The molten iron ladle arrives at the slag removal station, and a high-temperature resistant high-definition industrial camera located below the stirring and desulfurization platform transmits image information during the KR desulfurization and stirring process;

[0044] In an embodiment of the present invention, a camera system is adopted, which includes a high-temperature resistant high-definition industrial camera, which is connected to the host through a data transmission line and a conversion device. The camera is installed at the bottom side of the stirring and desulfurization platform to obtain the complete liquid level video frame image of the molten iron tank during stirring. The camera has a built-in dust accumulation device, and the system operation is set up indoors.

[0045] Step 102): Input the molten iron ladle liquid level image and convert it into a grayscale image I. Divide the grayscale image I into W×W non-overlapping blocks, where the default value of W×W is 16×16;

[0046] Step 103): Obtain the grayscale value within the block;

[0047] Step 104): Divide each block image into two groups of pixel values ​​according to the threshold T and calculate a new threshold value, where the new threshold value T1 is equal to the sum of the average values ​​of the two groups of pixels divided by 2;

[0048] In an embodiment of the present invention, the threshold T is equal to the average grayscale value of the selected area, where the selected area refers to a rectangular area containing the entire liquid surface of the molten iron ladle, and may refer to the entire image captured by the camera system.

[0049] In an embodiment of the present invention, step 104) further includes calculating a new threshold method, and the calculation formula is:

[0050]

[0051] Where T1 is the new threshold, G1 is the total grayscale value of all block images with grayscale values ​​lower than the threshold T, G2 is the total grayscale value of all block images with grayscale values ​​higher than the threshold T, g1 is the total number of block images with grayscale values ​​lower than the threshold T, and g2 is the total number of block images with grayscale values ​​higher than the threshold T.

[0052] Step 105): If the difference between the new threshold and the old threshold is zero, the new threshold is the optimal threshold for the binary image. Otherwise, continue iterating until the condition is met. The number of iterations is an empirical value.

[0053] Step 106): binarize the grayscale image I using the optimal threshold to obtain a grayscale image I1;

[0054] Step 107): sharpen the boundary of the segmented background area of ​​the grayscale image I1. The sharpening operation is an opening operation, and the convolution kernel size is 9×9, so that the blurred boundary is processed to obtain a smooth black boundary.

[0055] Step 108): Fill the sharpened background area into the grayscale image I1 to remove the background and interference factors of the grayscale image I1;

[0056] In the embodiment of the present invention, step 108) further includes a step of sharpening the background and drawing it into the grayscale image I1. For details, see Figure 2 , traverse the grayscale image I1. If the grayscale value is 0, the grayscale value of the pixel where the grayscale value is 0 is 0. Otherwise, the grayscale value of the pixel where the grayscale value is not 0 is 255.

[0057] Step 109): The image of the molten iron surface is obtained by the above steps, which is separated from the background and interference factors, thereby reducing the influence of the complex background on the detection target, and is named as the molten iron tank liquid surface image; Figure 1 As shown;

[0058] Step 110): Input the molten iron ladle liquid level image, and perform image normalization processing to convert all images into standard images with a given mean and variance, thereby reducing the difference in grayscale values ​​between different vortex images;

[0059] In an embodiment of the present invention, step 110) further includes calculating an image normalization method, and the calculation formula is:

[0060]

[0061] Where, I(i,j) represents the image before normalization (i.e., the image of the molten iron tank surface in step 109), G (i,j) Represents the normalized image (i.e., standard image), M0 and are the pre-set image average grayscale and mean square error, M and σ 2are the grayscale mean and variance of the image before normalization.

[0062] Step 111): Process the standard image using a Gabor filter to enhance grayscale detail differentiation and obtain a grayscale image I2;

[0063] In an embodiment of the present invention, step 111) further includes setting parameters of the Gabor filter, wherein the parameters are:

[0064] Ksize=21; Sigma=5; Theta=0 (any value from 0 to 360); Lambd=12; Gamma=300;

[0065] Psi=74;

[0066] Ksize represents the size of the filter; Sigma represents the standard deviation of the Gaussian function; Theta represents the Gaussian function; Lambd represents the wavelength of the cosine function; Gamma represents the aspect ratio of the two-dimensional Gaussian function; Psi represents the phase of the cosine function.

[0067] Step 112): Divide the grayscale image I2 into W×W non-overlapping blocks, where the default value of W×W is 16×16;

[0068] Step 113): extracting the grain direction and grain frequency of each block;

[0069] In an embodiment of the present invention, step 113) further includes calculating the ridge direction extraction method for each block, and the calculation formula is:

[0070]

[0071]

[0072]

[0073] Among them, the horizontal gradient value of each pixel is And the vertical gradient value is The pixel coordinates of the center point of the block are (i, j), θ(i, j) is the direction angle of the grain line, and w represents the major diameter of the grain line.

[0074] In the embodiment of the present invention, in step 113), the method for extracting the ridge frequency of each block is:

[0075] Based on the obtained ridge direction, the grayscale values ​​of all pixels of each image are projected along the vertical direction of the ridge direction. The projection forms a one-dimensional sine wave, and the extreme points of the one-dimensional sine wave correspond to the ridges of the vortex. Assuming T(i,j) is the average number of pixels between two adjacent peaks of the one-dimensional sine wave, the frequency

[0076] Step 114): Obtaining the curvature of the grain lines within the block;

[0077] In an embodiment of the present invention, step 114) further includes calculating the ridge curvature extraction method of each block, and the calculation formula is:

[0078]

[0079] φ x (i,j,u,v)=|cos(2θ(i,j))-cos(2θ(u,v))|

[0080] φ y (i,j,u,v)=|sin(2θ(i,j))-sin(2θ(u,v))|

[0081] Among them, (i, j) is the coordinate of the center pixel of the block, (u, v) represents the coordinate of (i, j) in the frequency domain, C(i, j) represents the curvature of the grain line in the block, θ represents the direction angle, N represents the size of the local area where the block is located (N×N) (here 3×3), and φ represents the direction of the Gabor filter.

[0082] Step 115): construct a Gabor enhancement template;

[0083] In an embodiment of the present invention, step 115) further includes a method for calculating and constructing a Gabor enhancement template, and the calculation formula is:

[0084]

[0085] x φ =xcosφ+ysinφ

[0086] y φ =-xcosφ+ysinφ

[0087] Where φ is the direction of the Gabor filter, f is the frequency of the lines, [x φ ,y φ ] indicates that the coordinate axis [x, y] rotates counterclockwise by an angle φ, σ x and σ y are the Gaussian envelope constants along the x and y axes, respectively.

[0088] Step 116): After constructing the vortex enhancement template, the grayscale image I2 obtained in step 111) is enhanced to obtain a vortex image with enhanced vortex lines. Figure 3 ;

[0089] Step 117): Collecting the ridge line information of the molten iron vortex image of the current frame after processing in the above steps;

[0090] In an embodiment of the present invention, in step 117), the spiral line information is collected and information is extracted from the spiral line and processed. The lines are numbered according to the distance from the inside to the outside. The line coordinates of each line are established according to the pixel coordinates (i, j) of the center point of each block in step 113), and a database is established based on the line coordinates, line direction, and line curvature.

[0091] Step 118): Filter the collected grain line information and remove useless data;

[0092] Among them, step 118) further illustrates the method of filtering the grain line information and eliminating useless data as follows:

[0093] Organize the collected data and remove data with too many null values;

[0094] For vortex line data with fewer null values, the missing null values ​​are replaced by the score of the historical average rank of the vortex direction and curvature similar to the vortex line rank of the frame image.

[0095] It should be noted that according to Helmholtz's first theorem: the vortex intensity on each cross section of the same vortex tube is the same (Helmholtz's first theorem applies to both ideal fluids and viscous fluids), and each pixel point on the same stripe line has the same identity.

[0096] That is, when the camera system and the molten iron ladle are fixed relative to each other, the surface flow of the molten iron fluid is traceable on a two-dimensional image. A fluid simulation is set up to capture the outward diffusion trajectories of the molten iron surface from multiple near-center points. The diffusion trajectories are observed at the same relative positions and the trajectory formula is output on the two-dimensional image.

[0097] It should be noted that, due to the viscous nature of molten iron, its viscosity increases with decreasing temperature, and it rises sharply around 1250°C. At the same temperature, viscosity is affected by factors such as Ti content, sulfur content, alkalinity, and silicon content.

[0098] Specifically, the identification process constrains the vortex region and the coordinate ranges for the start and end points of particle motion to ensure that the selected region is minimally affected by the viscosity of the molten iron under the current conditions. The smaller contact area between the agitator and the molten iron also has a certain impact. The edges of the molten iron ladle are significantly affected, so lines in the database with coordinates within a certain range from the center and near the edge are eliminated.

[0099] Step 119): Predicting the ridge lines of the molten iron vortex image in the current frame of the vortex, and predicting the image ridge line information after the interval time t;

[0100] The method for predicting the ridge lines of the molten iron vortex image in step 119) is:

[0101] It should be noted that the stirring of molten iron is a forced vortex, that is, the stirring fluid can cause the fluid to rotate, but there is no relative motion between the small fluid groups. The fluid is rotated by consuming energy provided by the outside world. This force forces the fluid to move centripetally, thus achieving the fluid's rotational motion. The velocity distribution of the fluid is:

[0102] v=CR

[0103] Where C is a constant. It can be seen that as the radius of curvature increases, the fluid flow rate gradually increases.

[0104] According to the motion characteristics of the grain lines, the dynamic formula of the grain lines for further calculation and prediction of the vortex is:

[0105]

[0106]

[0107] Where s is the distance moved by the pixel coordinates (i, j) of each block center point in step 113), and the direction is θ(i, j) in step 113). a[0] is the instantaneous acceleration of the center point (i, j), w represents the major diameter of the grain line, x represents the predicted major diameter of the grain line, and R represents the distance between the center point of the block grain line and the center point of the molten iron vortex surface.

[0108] The predicted ridge line is calculated by the center point of each block (i, j), and the pixel points with a long diameter of x are filled in front and behind the predicted coordinate point direction θ(i, j) according to the curvature C(i, j) in a single ridge line.

[0109] It should be noted that according to Thomson's theorem: the velocity circulation along any closed fluid circumference composed of fluid particles does not change with time. Therefore, the prediction algorithm is designed to collect the current frame image and the time image with an interval of t as input data. Based on the instantaneous nature of the theorem, it is found through experiments that the interval t should not be too large.

[0110] Step 120): Based on the real image ridge information after the interval time t, a vortex image processed by the real information and the predicted information is given, and the real ridges and the predicted ridges are distinguished and recorded.

[0111] The present invention is based on the vortex characteristics of molten iron discovered through research. Through the image processing method of the present invention, vortex lines after an interval time t are obtained, so as to avoid the possible blurring of the lines after the interval time t and better increase the number of strong characteristic vortex images.

[0112] By processing the images provided by the camera system during the KR stirring desulfurization reaction of molten iron using the technology provided by the present invention, more clear, characteristic images with strong vortex lines can be obtained, and the vortex line data can directly reflect the quality of the operating status. By using more interference-free, strong characteristic vortex images, the monitoring system can more accurately obtain the state of the molten iron during the stirring process. After timely adjustment of the control parameters, the waste of desulfurizer in supersaturated conditions and the frequency of secondary desulfurization will be greatly reduced. This not only effectively reduces the time and cost of a set of processes, but also improves the production efficiency of the entire production system, providing powerful support for the company's production operations.

[0113] It should be pointed out that, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, or two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0114] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for enhancing vortex image information based on molten iron stirring process, characterized in that: include: (1) Detecting the molten iron information in the image, using an iterative segmentation method to detect the molten iron vortex surface from the complex background to obtain a segmented image of the molten iron tank surface; (2) The segmented molten iron ladle liquid surface image is subjected to the Gabor filter image recognition method to obtain the molten iron vortex image after the vortex lines are enhanced; (3) Predicting and enhancing the molten iron vortex image based on the vortex characteristics, acquiring the lines with difference values ​​within a limited range, and obtaining the target molten iron vortex image; Step (1) includes: (1.1) Convert the original input image into a grayscale image I, and divide the grayscale image I into several non-overlapping blocks; (1.2) The threshold T is determined by the average gray value of the selected area, and the new threshold is calculated based on the threshold T, which divides all blocks into two groups of pixel values; (1.3) If the difference between the new threshold T1 and the old threshold T is zero, the new threshold T1 is used as the optimal threshold for the binary image. Otherwise, the iteration is continued until the difference between the new threshold T1 and the old threshold T is zero. (1.4) Using the optimal threshold, the grayscale image I is binarized into the grayscale image I1, and the boundary of the segmented background area of ​​the grayscale image I1 is sharpened; (1.5) Fill the sharpened background area into the grayscale image I1 to remove the original image background and interference factors.

2. The method according to claim 1, characterized in that Depend on Get the new threshold T1, where G1 is the total grayscale value of all block images with grayscale values ​​lower than the threshold T, G2 is the total grayscale value of all block images with grayscale values ​​higher than the threshold T, g1 is the total number of block images with grayscale values ​​lower than the threshold T, and g2 is the total number of block images with grayscale values ​​higher than the threshold T.

3. The method according to claim 2, characterized in that Step (2) includes: (2.1) Normalizing the segmented molten iron ladle surface image and converting it into a standard image with a given mean and variance; (2.2) The standard image is processed by the Gabor filter to obtain the grayscale image I2; (2.3) Divide the grayscale image I2 into several non-overlapping blocks and extract the ridge direction and ridge frequency of each block; (2.4) Obtaining the curvature of the grain lines within the block; (2.5) A Gabor enhancement template is constructed to enhance the grayscale image I2, and the molten iron vortex image with enhanced vortex lines is obtained.

4. The method according to claim 3, characterized in that Depend on Calculate the grain direction of each block, where The horizontal gradient value of each pixel is And the vertical gradient value is The pixel coordinates of the center point of the block are (i, j), θ(i, j) is the direction angle of the grain line, w represents the major diameter of the grain line, and (u, v) represents the coordinates of (i, j) in the frequency domain.

5. The method according to claim 4, characterized in that The method for extracting the ridge frequency of each block is: Based on the obtained grain direction, the grayscale values ​​of all pixels in each block are projected in the direction perpendicular to the grain direction to form a one-dimensional sine wave. The extreme points of the one-dimensional sine wave correspond to the grain lines of the vortex. Assuming T(i,j) is the average number of pixels between two adjacent peaks of the one-dimensional sine wave, the frequency 6. The method according to claim 5, characterized in that Depend on Get the curvature of each block, where φ x (i,j,u,v)=|cos(2θ(i,j))-cos(2θ(u,v))|,φ y (i,j,u,v)=|sin(2θ(i,j))-sin(2θ(u,v))|, where (i,j) is the coordinate of the center pixel of the block, (u,v) represents the coordinate of (i,j) in the frequency domain, C(i,j) represents the curvature of the grain line in the block, θ represents the direction angle, N represents the size of the local area where the block is located, and φ represents the direction of the Gabor filter.

7. The method according to claim 6, characterized in that Depend on Construct a Gabor enhancement template G, where x φ =xcosφ+ysinφ,y φ =-xcosφ+ysinφ, φ is the direction of the Gabor filter, f is the frequency of the line, [x φ ,y φ ] indicates that the coordinate axis [x, y] rotates counterclockwise by an angle φ, σ x and σ y are the Gaussian envelope constants along the x and y axes, respectively.

8. The method according to claim 7, characterized in that Step (3) includes: (3.1) Collecting the grain line information of the enhanced molten iron vortex image; (3.2) Screening the collected grain line information and eliminating useless data; (3.3) Predicting the ridge information of the molten iron vortex image, and predicting the ridge information of the image after the interval t; (3.4) According to the real image ridge information after the interval time t, the vortex image processed by the real information and the predicted information is calculated, and the real ridges and the predicted ridges are distinguished and recorded.

9. The method according to claim 8, characterized in that Step (3.3) includes: Depend on and Calculate the predicted vortex line dynamics, where s is the moving distance of the pixel coordinates (i, j) of the center point of each block, the direction is θ(i, j), a[0] is the instantaneous acceleration of the center point (i, j), w is the long diameter of the line, x is the long diameter of the predicted line, R is the distance between the center point of the block line and the center point of the molten iron vortex surface, and v is the fluid velocity; The predicted ridge line is calculated by the center point of each block (i, j), and the pixel points with a long diameter of x are filled in front and behind the predicted coordinate point direction θ(i, j) according to the curvature C(i, j) in a single ridge line.

Images