An online method and system for locating a molten metal jet boundary curve

Abstract

The application discloses an online positioning method and system for a molten metal jet boundary curve, which comprises the following steps: obtaining a visible light image containing a nozzle and a molten metal jet area of a furnace, performing Hough transformation on the binary visible light image to obtain a rough stream boundary curve, determining the slope of a straight line where the nozzle is located according to the upward angle of the nozzle, searching for the straight line where the nozzle is located in the rough stream boundary curve according to the slope of the straight line where the nozzle is located, obtaining a nozzle position function, establishing a decision model to estimate the nozzle diameter, obtaining the upper and lower positioning points of the nozzle in parallel by using the nozzle position function, performing Hough transformation on the binary edge image based on the upper and lower positioning points of the nozzle, and screening the optimal boundary curve with the stream width expanding with the jet distance, so that the technical problem of low positioning precision of the existing molten metal jet boundary curve is solved, the nozzle position can be accurately positioned, and the accurate molten metal jet boundary curve is obtained.

Description

Technical Field

[0001] The present invention mainly relates to the field of metallurgy technology, and in particular to an online positioning method and system for a molten metal jet boundary curve. Background Art

[0002] A molten metal jet is a stream of molten metal or metal concentrate produced during pyrometallurgy, ejected at high speed from the outlet of a high-temperature, closed furnace. Examples include the molten iron stream in blast furnace ironmaking, the molten steel stream in various steelmaking processes, the blister copper stream in pyrometallurgical copper smelting, and the molten aluminum stream from aluminum electrolysis. The aperture of the jet nozzle is key information for characterizing the forward flow state of metallurgical products. The curvature of the jet boundary curve is an important parameter for detecting the outlet flow rate of the smelted product. The divergence of the jet stream is the primary basis for adjusting the outflow duration of the metallurgical melt and determining the timing of outlet unblocking. Therefore, accurately extracting the upper and lower boundary curves of the molten metal jet under different outflow states is of great significance for real-time detection of metallurgical product production efficiency and product flow, as well as monitoring the operating status of the metallurgical process.

[0003] However, during the metallurgical process, molten metal liquid is often mixed with a small amount of furnace gas and ejected from the reactor. There is intermittent and randomly distributed high-concentration dust on its surface, which blurs the edges of the stream and reduces the brightness of the stream surface. Due to the high pressure inside the furnace, the jet is ejected at high speed, and its end decomposes and expands under the influence of gravity and liquid surface tension. The above factors put the jet in a harsh and difficult to predict state with a large dynamic range of initial velocity, a wide range of diameter variation, and uneven surface brightness, making the online positioning of the boundary curve of the molten metal jet full of challenges. Currently, there are three main methods for jet boundary curve positioning: outer contour curve extraction after threshold segmentation based on the light and dark contrast of the jet background, slender jet trajectory fitting based on intelligent algorithms, and significant foreground target extraction based on deep learning.

[0004] The threshold segmentation contour extraction method extracts a binary edge image after grayscale image segmentation, and then extracts a data point set for the jet column's contour. This method's edge extraction is susceptible to interference from metallurgical surface dust and splashing droplets, resulting in boundary fitting errors in contour extraction, poor measurement accuracy, and low stability. The trajectory fitting method extracts information such as the jet's starting point, length, and angle as prior information, and uses a one-way search method to calculate the particle point position in the next frame of the image to obtain the optimal value of the jet trajectory. However, this method is not suitable for tracking the upper and lower boundaries of wide molten metal jets. The foreground target extraction method uses a simple linear clustering iterative algorithm to perform superpixel segmentation on the image and detects significant foreground targets using a deep conditional random field model. This method is highly feasible, but it is labor-intensive and difficult to transfer across scenarios, requiring parameter reselection for images of different resolutions and angles.

[0005] After comprehensively comparing the advantages and disadvantages of the aforementioned detection methods, this paper proposes a method for online positioning of the boundary curve of a molten metal jet. This method accurately locates the nozzle position during molten metal ejection in pyrometallurgical processes and solves the problem of online positioning of the boundary curve of a high-speed jet in harsh environments. This method provides accurate data support for real-time metallurgical output flow detection and non-contact jet velocity measurement, providing a powerful basis for identifying metallurgical outflow conditions.

[0006] Patent publication number CN104501737A discloses a device and method for locating the boundary of a liquid jet spray. The patent describes an image acquisition and processing device consisting of a light-sheet laser, a synchronous controller, a CCD camera, and a computer. The device uses the OTSD method to extract boundaries from the grayscale integral matrix of a multi-frame image. However, the patent is not suitable for boundary extraction of a molten jet under complex dust interference.

[0007] Patent publication number CN102930543A discloses a method for searching fire monitor jet trajectories based on a particle swarm algorithm. This patent uses jet trajectory information extracted from the first image as prior information and uses a one-way search method based on the particle swarm algorithm to obtain the optimal solution for the jet trajectory. However, this solution is not applicable to metallurgical production sites with large jet divergence and a wide dynamic range of initial velocity, and its measurement accuracy is limited. Summary of the Invention

[0008] The online positioning method and system for the boundary curve of a molten metal jet provided by the present invention solve the technical problem of low positioning accuracy of the boundary curve of a molten metal jet in the prior art.

[0009] To solve the above technical problems, the present invention proposes an online positioning method for the boundary curve of a molten metal jet, which includes:

[0010] A visible light image including a nozzle of a furnace and a molten metal jet region is acquired, and the visible light image is binarized to obtain a binary edge image.

[0011] The binary edge image is subjected to Hough transform to obtain the rough stream boundary curve.

[0012] According to the upward angle of the nozzle, determine the slope of the straight line where the nozzle is located.

[0013] According to the slope of the nozzle line, the nozzle line with the greatest correlation with the binary edge image is searched within the rough stream boundary curve to obtain the nozzle position function.

[0014] A decision model is established to estimate the nozzle diameter, and the nozzle position function is established in parallel to obtain the upper and lower positioning points of the nozzle.

[0015] Based on the upper and lower positioning points of the nozzle, the binary edge image is subjected to Hough transform, and the optimal boundary curve of the stream width increasing with the injection distance is screened.

[0016] Furthermore, the binary edge image is subjected to Hough transform to obtain the rough stream boundary curve including:

[0017] Based on the parabola-like characteristic of the boundary curve of the molten metal jet, a Hough transform is performed on the binary edge image to obtain a first Hough transform coefficient matrix. The first Hough transform coefficient matrix includes first Hough transform coefficients and the number of first Hough transform space curves corresponding to the first Hough transform coefficients.

[0018] A first boundary curve is obtained according to the first Hough transform coefficient corresponding to the largest first Hough transform space curve number.

[0019] A second boundary curve is obtained according to first Hough transform coefficients corresponding to a maximum number of first Hough transform space curves having a preset intercept deviation from the first boundary curve.

[0020] An upper boundary curve and a lower boundary curve of the rough stream boundary curve are obtained according to the first boundary curve and the second boundary curve.

[0021] Furthermore, based on the upward angle of the nozzle, the calculation formula for determining the slope of the straight line where the nozzle is located is:

[0022]

[0023] Where θ is the upward angle of the nozzle, and k is the slope of the straight line where the nozzle is located.

[0024] Furthermore, according to the slope of the straight line where the nozzle is located, the straight line where the nozzle is located with the greatest correlation with the binary edge image is searched within the rough stream boundary curve, and the nozzle position function is obtained, including:

[0025] The first traversal window of the straight line traversing the nozzle is determined according to the rough stream boundary curve. The abscissa and ordinate values ​​of the midpoint of the first traversal window satisfy:

[0026]

[0027] Where x and y are the horizontal and vertical coordinates of the midpoint of the first traversal window, respectively, and satisfy y = kx + b, k and b are the slope and intercept of the straight line where the nozzle is located, respectively, C1, A1 and B1 are the upper boundary curves of the rough stream boundary curve y1 = C1 + A1 (x-B1) 2 The coefficients of C2, A2 and B2 are the lower boundary curve of the rough stream boundary curve y2=C2+A2(x-B2) 2 The coefficient of , width is the height of the visible light image.

[0028] In the first traversal window, according to the slope of the straight line where the nozzle is located, the straight line where the nozzle is located is traversed through the first traversal window along the horizontal coordinate direction with a preset step size, and the Pearson correlation coefficient between the points on the straight line where the nozzle is located and the edge points in the binary edge image is calculated. The intercept of the straight line where the nozzle is located when the Pearson correlation coefficient is the largest is selected as the intercept coefficient of the nozzle position function, and the slope coefficient of the nozzle position function is equal to the slope of the straight line where the nozzle is located.

[0029] Furthermore, a decision model is established to estimate the nozzle diameter, and the nozzle position function is used in conjunction to obtain the upper and lower positioning points of the nozzle, including:

[0030] According to the preset upper boundary intercept and the preset lower boundary intercept, the intercept coefficients in the upper boundary curve coefficient and the lower boundary curve coefficient are changed respectively to obtain the upper boundary approximation curve and the lower boundary approximation curve.

[0031] A second traversal window of the nozzle position straight line traversal is determined according to the upper boundary approximation curve and the lower boundary approximation curve, where the nozzle position straight line is a straight line corresponding to the nozzle position function.

[0032] In the second traversal window, the nozzle position line is traversed along the abscissa direction at a preset step length, and the number of edge points in the binary edge image on the nozzle position line is counted to obtain a first statistical value.

[0033] The overlap ratio of the upper boundary approximation curve and the lower boundary approximation curve with the edge points in the binary edge image within the jet development range starting from the nozzle is counted to obtain a second statistical value, wherein the jet development range is equal to the nozzle diameter.

[0034] An upper boundary approximation curve and a lower boundary approximation curve corresponding to when both the first statistical value and the second statistical value are maximized are obtained as the upper boundary positioning curve and the lower boundary positioning curve.

[0035] The upper and lower positioning point coordinates of the nozzle are obtained according to the intersection of the upper boundary positioning curve, the lower boundary positioning curve and the nozzle position straight line.

[0036] Furthermore, the expressions of the upper boundary approximation curve and the lower boundary approximation curve are:

[0037]

[0038] where y 逼近1 and y 逼近2 are the expressions of the upper boundary approximation curve and the lower boundary approximation curve, respectively. l1 and l2 are the preset upper boundary intercept and the preset lower boundary intercept, respectively.

[0039] And the horizontal and vertical coordinates of the midpoint of the second traversal window satisfy:

[0040]

[0041] where x 逼近 and y 逼近 They are the horizontal and vertical coordinates of the midpoint of the second traversal window, and satisfy y 逼近 =k final x 逼近 +b final , k final and b final are the slope coefficient and intercept coefficient of the nozzle position function, respectively.

[0042] Furthermore, based on the upper and lower positioning points of the nozzle, the binary edge image is subjected to Hough transform, and the optimal boundary curve of the stream width increasing with the injection distance is screened, including:

[0043] Based on the upper and lower positioning points of the nozzle, a Hough transform is performed on the binary edge image to obtain a second Hough transform coefficient matrix, which includes second Hough transform coefficients and the number of second Hough transform space curves corresponding to the second Hough transform coefficients.

[0044] The second Hough transform coefficient matrix is ​​sorted according to the number of the second Hough transform space curves.

[0045] The second Hough transform coefficient matrix is ​​traversed and sorted in descending order according to the number of second Hough transform space curves, and the optimal boundary curve is obtained by the corresponding second Hough transform coefficient when judging that the stream width increases with the injection distance. The judgment condition that the stream width increases with the injection distance is: D(x0+a)≥D(x0), where D(x0+a) and D(x0) are the stream widths when the horizontal coordinates are x0+a and x0, respectively, and a represents the characteristic variable of the jet flow direction. When the jet flows horizontally to the left, a=-1, and when the jet flows horizontally to the right, a=1.

[0046] Furthermore, the formula for calculating the stream width is:

[0047]

[0048] Among them, A1' and B1' are the upper boundary curves in the optimal boundary curve The coefficients of A2' and B2' are the lower boundary curves in the optimal boundary curve The coefficient of and are the upper and lower positioning point coordinates of the nozzle, D(x) is the stream width when the horizontal coordinate is x, and Δx=|x up -x down |, where x up 、xdown satisfy:

[0049]

[0050]

[0051] The online positioning system for the boundary curve of a molten metal jet provided by the present invention comprises:

[0052] A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the online positioning method for the molten metal jet boundary curve provided by the present invention are implemented.

[0053] The present invention takes the molten metal jet under harsh metallurgical conditions as the research object and proposes a method and system for online positioning of the boundary curve of the molten metal jet under complex environments. The present invention uses a visible light image acquisition system to acquire real-time images of the molten metal high-speed jet and proposes a two-stage Hough transform method for positioning the upper and lower boundary curves of the jet. In view of the characteristics of the molten metal jet, such as large divergence, wide stream diameter, and unstable surface, the rough upper and lower boundary curves of the stream are obtained by Hough transform using the difference in the top intercept of the parabola. To locate the nozzle position, the function coefficient at which the correlation between the straight line where the nozzle is located and the edge is maximized is evaluated based on the Pearson correlation coefficient. To obtain the nozzle diameter, a decision model is established to make the rough curve approximate the upper and lower point positions of the aperture, thereby obtaining the intersection point as the upper and lower positioning points of the nozzle, and calculating the accurate real-time area of ​​the nozzle. Finally, from the precise positioning point of the nozzle, the Hough transform is applied again to obtain the precise upper and lower boundary curves of the stream. The present invention solves the problem of difficulty in extracting the upper and lower boundary curve coefficients of a molten metal jet with a large dynamic range of initial velocity, a wide stream diameter, and uneven surface dust concentration. It realizes the detection of the trajectory, divergence, and nozzle position of the metallurgical jet, which is of great significance for the collection of process variables such as real-time output flow rate and product production efficiency and the monitoring of abnormal operating conditions in the reactor.

[0054] The effects of the present invention specifically include:

[0055] (1) According to the shape characteristics of the metallurgical reactor nozzle, a coefficient identification model of the nozzle position function was constructed, which effectively located the reactor nozzle position in the visible light molten metal jet image.

[0056] (2) Based on the image edge integrity information of the nozzle, a computational model for measuring the diameter of the reactor nozzle was constructed. Furthermore, to address the measurement errors caused by the inaccurate edge of the camera field of view due to the friction between the molten metal jet and the nozzle, the present invention comprehensively considers the breakup principle and position of liquid jet dynamics, proposes an estimation method for nozzle diameter calculation, and obtains the accurate nozzle area online.

[0057] (3) The accurate up-down coordinates of the nozzle in the field of view are calculated according to the area and position of the nozzle, and a jet up-down boundary positioning model based on two-stage Hough transformation is proposed. According to the characteristics that the jet flow width grows with the development of the jet, the optimal boundary curve is screened out.

[0058] (4) The online positioning method and system of the molten metal jet boundary curve based on two-stage Hough transformation are first proposed, which realizes the online accurate positioning of the up-down boundary curve of the molten metal jet under the harsh and difficult-to-measure conditions of uneven dust interference on the surface of the metallurgical production site. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 It is a visible light image acquisition system schematic diagram of the second embodiment of the present application.

[0060] Figure 2 It is a molten metal jet boundary curve online positioning method flow chart of the second embodiment of the present application.

[0061] Figure 3 It is a molten metal jet spraying schematic diagram of the second embodiment of the present application.

[0062] Figure 4 It is a tapping hot metal flow boundary curve positioning effect diagram of the third embodiment of the present application.

[0063] Figure 5 It is a taphole diameter detection result diagram of the third embodiment of the present application.

[0064] Figure 6 It is a structure block diagram of the online positioning system of the molten metal jet boundary curve of the present application.

[0065] REFERENCE SIGNS:

[0066] 1, reaction furnace wall; 2, reaction furnace nozzle; 3, molten metal jet; 4, industrial high-speed camera; 5, comprehensive cable; 6, camera field of view range; 7, metal aggregate groove; 8, computer; 10, reaction furnace nozzle outlet; 11, undisturbed jet section; 12, jet surface disturbance wave; 13, nozzle diameter; 14, jet flow diameter; 20, storage; 30, processor. DETAILED DESCRIPTION

[0067] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present application is not limited to the following specific embodiments.

[0068] The embodiments of the present application will be described in detail below with reference to the accompanying drawings, but the present application can be implemented in various different ways limited and covered by the claims.

[0069] Embodiment one

[0070] The online positioning method of the molten metal jet boundary curve provided by the embodiment one of the present application comprises:

[0071] In step S101, a visible light image of a nozzle and a molten metal jet area of a furnace is acquired, and the visible light image is binarized to obtain a binarized edge image.

[0072] In step S102, Hough transformation is performed on the binarized edge image to obtain a rough stream boundary curve.

[0073] In step S103, the slope of a straight line where the nozzle is located is determined according to the upward angle of the nozzle.

[0074] In step S104, the straight line where the nozzle is located is searched in the rough stream boundary curve according to the slope of the straight line where the nozzle is located, and a nozzle position function is obtained, which has the largest correlation with the binarized edge image.

[0075] In step S105, a decision model is established to estimate the nozzle diameter, and the upward and downward positioning points of the nozzle are obtained by combining the nozzle position function.

[0076] In step S106, Hough transformation is performed on the binarized edge image based on the upward and downward positioning points of the nozzle, and the optimal boundary curve with the stream width expanding with the jet distance is screened.

[0077] The online positioning method of the molten metal jet boundary curve provided by the embodiment of the present application acquires a visible light image of a nozzle and a molten metal jet area of a furnace, and binarizes the visible light image to obtain a binarized edge image, performs Hough transformation on the binarized edge image to obtain a rough stream boundary curve, determines the slope of a straight line where the nozzle is located according to the upward angle of the nozzle, searches the straight line where the nozzle is located in the rough stream boundary curve according to the slope of the straight line where the nozzle is located, and obtains a nozzle position function which has the largest correlation with the binarized edge image, establishes a decision model to estimate the nozzle diameter, combines the nozzle position function to obtain the upward and downward positioning points of the nozzle, and performs Hough transformation on the binarized edge image based on the upward and downward positioning points of the nozzle, and screens the optimal boundary curve with the stream width expanding with the jet distance, thereby solving the technical problem of low positioning precision of the existing molten metal jet boundary curve, accurately positioning the nozzle position, obtaining the upward and downward boundary function curves of the wide jet, and having the advantages of safe operation, strong real-time performance, accurate positioning result and the like.

[0078] Specifically, the embodiment of the present invention takes the molten metal jet under harsh metallurgical conditions as the research object, and proposes a method and system for online positioning of the boundary curve of the molten metal jet under complex environments. The embodiment of the present invention uses a visible light image acquisition system to obtain real-time images of the molten metal high-speed jet, and proposes a two-stage Hough transform method for positioning the upper and lower boundary curves of the jet. In view of the characteristics of the molten metal jet, such as large divergence, wide stream diameter, and unstable surface, the rough upper and lower boundary curves of the stream are obtained by Hough transform using the difference in the top intercept of the parabola. In order to locate the nozzle position, the function coefficient when the correlation between the straight line where the nozzle is located and the edge is maximized is evaluated based on the Pearson correlation coefficient. In order to obtain the nozzle diameter, a decision model is established to make the rough curve approach the upper and lower point positions of the aperture, so as to obtain the intersection point as the upper and lower positioning points of the nozzle, and calculate the accurate real-time area of ​​the nozzle. Finally, from the precise positioning point of the nozzle, the Hough transform is applied again to obtain the precise upper and lower boundary curves of the stream. The embodiments of the present invention solve the problem of difficulty in extracting the upper and lower boundary curve coefficients of a molten metal jet with a large dynamic range of initial velocity, a wide stream diameter, and uneven high dust concentration on the surface. It realizes the detection of the trajectory, divergence, and nozzle position of the metallurgical jet, which is of great significance for the real-time collection of process variables such as output flow rate and product production efficiency and the monitoring of abnormal operating conditions in the reactor.

[0079] Example 2

[0080] This proposal proposes an online positioning method and system for the boundary curve of a molten metal jet. Figure 1 This is a schematic diagram of a visible light image acquisition system. The system includes a reactor wall 1, a reactor nozzle 2, a molten metal jet 3, an industrial high-speed camera 4, an integrated cable 5, a camera field of view 6, a metal aggregate ditch 7, and a computer 8. The camera field of view includes the reactor nozzle and the entire process of the jet ejecting into the metal aggregate ditch. Figure 2 1 is a diagram illustrating the implementation steps of the online positioning method for the boundary curve of a molten metal jet according to an embodiment of the present invention, comprising the following steps:

[0081] (1) A visible light image acquisition system is used to obtain a visible light image of the furnace nozzle and the molten metal jet area, and a high-noise-resistant edge detection algorithm is used to obtain an edge binary image of the original image.

[0082] (2) Using the top intercept difference of the stream boundary parabola curve, the rough stream boundary curve is located from the first stage Hough transform.

[0083] (3) Traverse and search for the function coefficient with the greatest correlation between the straight line where the nozzle is located and its edge points to obtain the optimal coefficient of the nozzle position function.

[0084] (4) Combining the edge integrity information of the nozzle image with the dynamic properties of the jet vertical gravity injection of air, a decision model is established to estimate the nozzle diameter, and the nozzle position function is used in conjunction to obtain the accurate upper and lower positioning points of the nozzle.

[0085] (5) Perform Hough transform again from the nozzle positioning point to screen the optimal boundary curve of the stream width as the injection distance increases.

[0086] The specific implementation plan is as follows:

[0087] (1) Obtain a visible light image of the molten metal fluid and locate the edge of the original image:

[0088] First, the visible light image acquisition system is installed facing the nozzle position and at a suitable distance from the molten metal. It is necessary to ensure that the industrial camera's field of view includes the entire process of the jet from the nozzle to the end, and to take into account the strong radiation of the molten metal fluid, which may damage the camera lens and reduce the camera power. The camera can be placed in a protective tube, and air cooling is used to cool the lens and blow away the dust on the surface, which not only reduces the camera's operating temperature but also prevents dust from covering the lens surface and affecting the image acquisition quality. After collecting the visible light image, the present invention uses an edge detection algorithm based on wavelet modulus maximum processing to obtain a binary image of the image. Specifically, it includes the following steps:

[0089] Step 1: Convert the RGB image captured by the camera into a grayscale image. The conversion formula is shown in formula (1), where Gray(x,y) represents the grayscale value of the pixel at the horizontal coordinate x and the vertical coordinate y, and R(x,y), G(x,y), and B(x,y) represent the red, green, and blue channel component values ​​at the coordinates, respectively.

[0090] Gray(x,y)=0.3R(x,y)+0.59G(x,y)+0.11B(x,y)(1)

[0091] Step 2: Establish and select the smoothing scaling function θ(x,y) to satisfy formula (2):

[0092]

[0093] Get the scale function θ under scale s s (x,y), as shown in formula (3):

[0094]

[0095] For the scale function, we get the wavelet function in the (x, y) direction As shown in formula (4):

[0096]

[0097] Step 3: Convolve the grayscale image with the wavelet function in the (x, y) direction to obtain wavelet coefficient components at different scales. As shown in formula (5):

[0098]

[0099] Calculate the wavelet modulus M based on the gradient component s f(x,y) and direction angle A s f(x,y), as shown in formula (6):

[0100]

[0101] Step 4: Traverse the image and determine the maximum value of the gradient amplitude in each direction. Select an appropriate threshold thre and convert the image into a binary edge image. P(x,y) is the binary pixel value of the coordinate (x,y), as shown in formula (7):

[0102]

[0103] At metal smelting sites, the vibration of the reactor causes small droplets to break away from the stream, and the jet splashes in the gutter, resulting in a large amount of useless information in the edge detection image. Therefore, an operation is performed on the edge image. Small edge interference in the edge image is removed to obtain the final binary image.

[0104] (2) The first stage Hough transform locates the rough stream boundary curve:

[0105] The Hough transform is an algorithm widely used for image features. When the function form of the object shape needs to be detected in a known image, the optimal function coefficient can be obtained by voting. According to the characteristics of the vertical gravity injection air jet trajectory, the molten metal liquid mass is thrown out from the nozzle in a parabola. Therefore, the boundary curve function can be formalized as a quadratic curve standard formula (8):

[0106] y=C+A(xB) 2 (8)

[0107] Assume that n edge points are obtained in equations (1)-(7), and their edge positions are (x0, y0), (x1, y1)…(x n ,y n ), when the Hough transform is used, it is substituted into formula (8), so that A, B, and C are used as parameter coordinate systems, and n parameter equations are formed, as shown in formula (9):

[0108]

[0109] At this time in the parametric coordinate system, n space curves in formula (9) converge on the collinear point. According to the number of space curves on the collinear point from large to small, reorder and store in the n*4 matrix, the first three columns store the values of A, B and C on the collinear point, and the fourth column stores the number of space curves thereon.

[0110] Since the intercept in the quadratic function form is the coefficient for approximately representing the nozzle position, it is necessary to find the difference between the intercepts in the coefficient matrix. The first row of the matrix is the coefficient row with the most coincident curves on the collinear point, which is the first boundary curve of the stream, and the intercept thereof is C1. Then traverse the matrix in descending order of the number of curves. When the intercept C2 and C1 satisfy |C1-C2|≥thre', the optimal coefficient of the second boundary curve is selected. According to the comparison of the intercepts, the two boundary curves are integrated to obtain the rough upper and lower boundary curves of the correct position. At this time, the boundary curve function is formula (10):

[0111]

[0112] In formula (10), C1≤C2 is satisfied.

[0113] (3) Traverse the search nozzle located straight line and its edge point related function coefficient, obtain the optimal coefficient of the nozzle position function.

[0114] In the edge image of the camera field of view, the nozzle edge is the only information representing the nozzle position. Therefore, it is necessary to construct an optimal coefficient evaluation model of the straight line where the nozzle is located, detect the straight line coefficient most similar to the edge, and determine the slope k and intercept b of the straight line. In the metal smelting site, the nozzle of the reaction furnace is upward at a fixed angle. When the camera is directly opposite the reaction furnace, the nozzle is located on a straight line in the field of view. Assuming that the upward angle is θ, the straight line where the nozzle is located is y=kx+b, at this time the slope k of the straight line is formula (11):

[0115]

[0116] In order to determine the accurate straight line intercept b, the embodiments of the present application search for the straight line where the nozzle is located in the rough stream boundary curve according to the slope of the straight line where the nozzle is located, and obtain the nozzle position function. Specifically, the embodiments of the present application use the Pearson correlation coefficient to analyze the correlation between the straight line and the edge points in the edge image. In order to estimate the region of the nozzle in the image and compare the correlation between the straight line and the edge points within the region, it is necessary to determine the horizontal and vertical range of the region. First, determine the horizontal range of the nozzle region, as formula (12):

[0117]

[0118] where k and b are the slope and intercept of the straight line where the nozzle is located, and width is the height of the visible light image.

[0119] Then the longitudinal range of the nozzle area is determined so that the straight line is limited between the upper and lower boundary curves determined by the first stage Hough transform, which can be expressed as y1≤y≤y2.

[0120] Within this range, the Pearson correlation coefficient r is compared between the points on the straight line where the nozzle is located and the edge points in the binary edge image. The calculation formula is as follows:

[0121]

[0122] Where m and n are the horizontal and vertical coordinates that satisfy the range of formula (12) and formula (13) respectively. is the pixel value of the coordinate (m,n) in the binary edge image, To determine whether the coordinates (m, n) are on the straight line, the indicator variable is calculated as follows:

[0123]

[0124] The calculation formula is as follows (15):

[0125]

[0126] When the correlation between the straight line where the nozzle is located and the edge image is the highest, the straight line intercept value b when the Pearson correlation coefficient is the largest is selected. final At this time, the straight line where the nozzle is located is expressed as y=kx+b final , the nozzle position straight line equation corresponding to the nozzle position function can be obtained.

[0127] (4) Combining the edge integrity information of the nozzle image with the dynamic properties of the jet vertical gravity injection of air, a decision model is established to estimate the nozzle diameter, and the nozzle position function is used in conjunction to obtain the accurate upper and lower positioning points of the nozzle.

[0128] During the metal smelting process, the high-speed molten metal jet constantly rubs against the nozzle, causing the nozzle to wear during operation, resulting in different shapes of occlusions in the camera field of view. If only the complete nozzle edge in the upper and lower boundaries is considered, the measurement error of the nozzle diameter will be large. Therefore, it is necessary to consider both the edge information and the initial width information of the stream: According to the jet breakup principle, when a liquid jet is ejected from the nozzle, the propagation of the liquid surface disturbance and the influence of gravity or the shear force between the air and the jet surface will cause the unstable disturbance to become wider as the jet develops, such as Figure 3 As shown. Figure 3As known from the field experience formula, when the jet is vertically ejected from the outlet 10 of the nozzle at high speed, the jet is in the undisturbed flow section 11 within the distance from the nozzle to the length of the nozzle diameter 13, then the jet develops into the jet surface disturbance wave 12, and the jet flow diameter 14 is approximately equal to the nozzle diameter 13. Therefore, in order to accurately detect the nozzle diameter, a decision model needs to be established to meet two conditions and improve the accuracy of approximating the nozzle diameter:

[0129] Condition 1: In the space contained by the upper and lower boundary curves, the number of intersection points of the straight line where the nozzle is located and the edge image is the largest.

[0130] Condition 2: In the range of s / a≤1 from the nozzle, the overlap rate of the boundary curve and the edge point is the highest. S is the distance of the jet development, and a is the nozzle diameter.

[0131] In order to meet condition 1, the edge intercepts of the upper and lower boundary curves are approximated from the boundary points of the image downward and upward, respectively, as shown in equation (16):

[0132]

[0133] where y 逼近1 and y 逼近2 are the expressions of the upper boundary approximation curve and the lower boundary approximation curve, respectively, and l1 and l2 are the preset upper boundary intercept and the preset lower boundary intercept, respectively. At this time, x is traversed in the second traversal window with a step size of , and the number of edge points contained in the binary edge image on the straight line of the nozzle position is counted to obtain the first statistical value. Wherein, the horizontal coordinate and the vertical coordinate of the point in the second traversal window satisfy:

[0134]

[0135] where x 逼近 and y 逼近 are the horizontal coordinate and the vertical coordinate of the point in the second traversal window, respectively, and satisfy y 逼近 =k final x 逼近 +b final , k final and b final are the slope coefficient and the intercept coefficient of the nozzle position function, respectively.

[0136] In order to determine whether the point on the straight line in the space of the upper and lower boundary approximation curves is the nozzle edge point, the indicator variable I(x) is given, which is defined as equation (18):

[0137]

[0138] The sum N of the indicator variables on the straight line is represented as equation (19):

[0139]

[0140] Where x begin 、x end To satisfy the approximation constraint in formula (17).

[0141] In order to satisfy condition 2, the upper and lower positioning points (x1, y1) and (x2, y2) of the nozzle are the intersection points of the upper and lower approximation boundary curves and the straight line where the nozzle is located, which are respectively expressed in equation (20):

[0142]

[0143] Where Δ1 and Δ2 satisfy formula (21):

[0144]

[0145] The nozzle diameter a is expressed as formula (22):

[0146]

[0147] The distances s1 and s2 of the jet bends starting from the two intersection points are expressed as (23):

[0148]

[0149] Where k1 and k2 satisfy formula (24):

[0150]

[0151] Given the indicator variable I'(x), it is defined as formula (25):

[0152]

[0153] Where

[0154] In order to compare the density of intersections between boundary curves and edge points, ρ is defined as Equation (26):

[0155]

[0156] The terminal coordinates are obtained by combining equations (22) and (23).

[0157] For the evaluation variable formula (19) of condition 1, when the boundary curve is relaxed relative to the nozzle edge point, it remains unchanged within the relaxation range; when the boundary curve intersects the nozzle edge point, N≤N max .

[0158] Regarding the evaluation variable formula (26) of condition 2, condition 2 is satisfied when the density ρ between the boundary curve and the stream edge points is the highest.

[0159] The decision model is constructed to satisfy conditions 1 and 2 at the same time, that is, to set equation (19) to the maximum value maxN and equation (26) to the maximum value maxρ.

[0160] At this time, the boundary curve function y=C+L+A(xB) is determined, and the upper and lower positioning points of the nozzle are the intersection points of the boundary curve and the straight line where the nozzle is located.

[0161] (5) Perform Hough transform again from the nozzle positioning point to screen the optimal boundary curve of the stream width as the injection distance increases

[0162] Similar to the one-stage Hough transform, the boundary curve detected by the Hough transform is repositioned as (27):

[0163]

[0164] After the coefficient matrix is ​​reordered from largest to smallest according to the number of spatial curves on collinear points, the optimal boundary curve is selected, whose upper and lower edges expand with the injection distance and have the largest number of overlapping curves on collinear points. In order to determine the expansion of the upper and lower boundary curves with the injection distance, after obtaining the center trajectory curve, the embodiment of the present invention calculates the intersection of the tangent on the center trajectory curve and the upper and lower boundary curves. At this time, the distance between the two intersection points is the stream width at point (x, y). Center trajectory curve y middle It is expressed as formula (28):

[0165]

[0166] On the stream center curve, the stream width is the perpendicular line of the stream center tangent in the boundary curve. First, the tangent slope k' on the center trajectory curve is obtained and expressed as formula (29):

[0167]

[0168] The stream width D at the point (x0, y0) is the solution of equation (30):

[0169]

[0170] Among them, A1' and B1' are the upper boundary curves in the optimal boundary curve The coefficients of A2' and B2' are the lower boundary curves in the optimal boundary curve The coefficient of and are the upper and lower positioning point coordinates of the nozzle, D(x) is the stream width when the horizontal coordinate is x, and Δx=|xup x down |, wherein x up , x down satisfies:

[0171]

[0172]

[0173] The simultaneous equations (29) and (30) satisfy the width screening condition when x1≥x0, D(x1)≥D(x0). The determination condition for the width of the stream to expand with the jet distance is: D(x0+a)≥D(x0), wherein x0 is the abscissa of any point in the development range of the stream, and a is a characteristic variable representing the flow direction of the jet. When the jet flows horizontally to the left, a=-1; when the jet flows horizontally to the right, a=1.

[0174] Finally, the embodiment of the present application traverses the coefficient matrix according to the number of spatial curves in descending order, checks whether each boundary curve is expanded, and screens the optimal upper and lower boundary curves.

[0175] Specifically, the embodiment of the present application acquires a visible light image containing a reaction furnace nozzle and a molten metal jet area by using an industrial high-speed camera. In order to reduce the noise interference generated by uneven dust covering the stream, the edge position of the jet stream is located, and a binary edge image of the visible light image is obtained by using a high-anti-noise edge detection algorithm. In view of the characteristics that the molten metal jet is large in divergence degree, wide in stream diameter, and uneven in surface brightness, the embodiment of the present application proposes an on-line positioning method for the upper and lower boundary curves of the jet based on two-stage Hough transformation. First, the first-stage Hough transformation is used to obtain the rough upper and lower boundary curves of the stream by using the parabola top intercept difference. Since the jet is sprayed from a fixed nozzle, in order to locate the accurate nozzle position, the embodiment of the present application traverses the search for the optimal function coefficient of the maximum correlation between the straight line where the nozzle is located and the nozzle edge point based on the characteristics that the reaction furnace nozzle is upwardly inclined at a fixed angle in the field of view. Then, the decision model is established by combining the edge integrity information of the nozzle image and the jet dynamic breaking position, the error of the two measurement conditions is corrected, the accurate diameter of the nozzle is calculated, and then the accurate upper and lower positioning points of the nozzle are obtained by using the nozzle position function. Finally, the second-stage Hough transformation is performed from the nozzle positioning point to screen the optimal boundary curve of the upper and lower edges expanding with the jet distance.

[0176] Embodiment three

[0177] In this embodiment, the 1050m 3 high blast furnace of a certain iron mill is taken as a test platform, the on-line positioning method and system for the boundary curve of the molten metal jet of the present application are applied to the molten iron quality monitoring system of the No. 1 iron notch of the large-scale blast furnace. The system of the present application takes the real-time image of the jet spraying by the visible light image acquisition system of the molten iron quality monitoring system, Figure 4 The edge image, taphole position, and upper and lower stream boundary curves of the jet stream detected in an embodiment of the present invention during a certain tapping process are shown. As can be seen from the figure, the curve positioning effect of the present invention is consistent with the outer boundary of the molten iron flow under normal tapping conditions. In order to illustrate the effect of taphole diameter measurement, Figure 5 The comparison between the tapping diameter detection fluctuation curve of the present invention and the theoretical tapping diameter wear curve in one tapping heat is shown, further demonstrating the accuracy of the tapping diameter positioning of the present invention in the blast furnace ironmaking process.

[0178] Reference Figure 6 The online positioning system of the molten metal jet boundary curve proposed in an embodiment of the present invention includes a memory 20, a processor 30, and a computer program stored in the memory 20 and executable on the processor 30, wherein the processor 30 implements the steps of the online positioning method of the molten metal jet boundary curve proposed in this embodiment when executing the computer program.

[0179] The specific working process and working principle of the online positioning system of the molten metal jet boundary curve of this embodiment can refer to the working process and working principle of the online positioning method of the molten metal jet boundary curve of this embodiment.

[0180] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An online positioning method for a molten metal jet boundary curve, characterized in that: The method comprises: Acquire a visible light image including a nozzle of a furnace and a molten metal jet region, and binarize the visible light image to obtain a binarized edge image; Performing Hough transform on the binary edge image to obtain a rough stream boundary curve, wherein performing Hough transform on the binary edge image to obtain the rough stream boundary curve includes: Based on the parabola-like characteristic of the boundary curve of the molten metal jet, a Hough transform is performed on the binary edge image to obtain a first Hough transform coefficient matrix, where the first Hough transform coefficient matrix includes first Hough transform coefficients and the number of first Hough transform space curves corresponding to the first Hough transform coefficients; Obtaining a first boundary curve according to a first Hough transform coefficient corresponding to the largest number of first Hough transform space curves; Obtaining a second boundary curve according to first Hough transform coefficients corresponding to a maximum number of first Hough transform space curves having a preset intercept deviation from the first boundary curve; Obtaining an upper boundary curve and a lower boundary curve of the rough stream boundary curve according to the first boundary curve and the second boundary curve; According to the upward angle of the nozzle, determine the slope of the straight line where the nozzle is located; According to the slope of the straight line where the nozzle is located, the straight line where the nozzle is located with the greatest correlation with the binary edge image is searched within the rough stream boundary curve to obtain the nozzle position function; A decision model is established to estimate the nozzle diameter, and the nozzle position function is established in parallel to obtain the upper and lower positioning points of the nozzle; Based on the upper and lower positioning points of the nozzle, the binary edge image is subjected to Hough transform, and the optimal boundary curve of the stream width expanding with the injection distance is screened. The optimal boundary curve of the stream width expanding with the injection distance is screened based on the upper and lower positioning points of the nozzle, including: Based on the upper and lower positioning points of the nozzle, a Hough transform is performed on the binary edge image to obtain a second Hough transform coefficient matrix, where the second Hough transform coefficient matrix includes second Hough transform coefficients and the number of second Hough transform space curves corresponding to the second Hough transform coefficients; sorting the second Hough transform coefficient matrix according to the number of the second Hough transform space curves; The second Hough transform coefficient matrix is ​​traversed and sorted in descending order according to the number of second Hough transform space curves, and the optimal boundary curve is obtained by the corresponding second Hough transform coefficient when judging that the stream width increases with the injection distance. The judgment condition that the stream width increases with the injection distance is: D(x0+a)≥D(x0), where D(x0+a) and D(x0) are the stream widths when the horizontal coordinates are x0+a and x0, respectively, and a represents the characteristic variable of the jet flow direction. When the jet flows horizontally to the left, a=-1, and when the jet flows horizontally to the right, a=1.

2. The online positioning method of the molten metal jet boundary curve according to claim 1, characterized in that: According to the upward angle of the nozzle, the calculation formula for determining the slope of the straight line where the nozzle is located is: Where θ is the upward angle of the nozzle, and k is the slope of the straight line where the nozzle is located.

3. The online positioning method of the molten metal jet boundary curve according to claim 2, characterized in that: According to the slope of the straight line where the nozzle is located, the straight line where the nozzle is located with the greatest correlation with the binary edge image is searched within the rough stream boundary curve. The nozzle position function is obtained as follows: The first traversal window of the straight line traversing the nozzle is determined according to the rough stream boundary curve. The abscissa and ordinate values ​​of the midpoint of the first traversal window satisfy: Where x and y are the horizontal and vertical coordinates of the midpoint of the first traversal window, respectively, and satisfy y = kx + b, k and b are the slope and intercept of the straight line where the nozzle is located, respectively, C1, A1 and B1 are the upper boundary curves of the rough stream boundary curve y1 = C1 + A1 (x-B1) 2 The coefficients of C2, A2 and B2 are the lower boundary curve of the rough stream boundary curve y2=C2+A2(x-B2) 2 The coefficient of , width is the height of the visible light image; In the first traversal window, according to the slope of the straight line where the nozzle is located, the straight line where the nozzle is located is traversed through the first traversal window along the horizontal coordinate direction with a preset step size, and the Pearson correlation coefficient between the points on the straight line where the nozzle is located and the edge points in the binary edge image is calculated. The intercept of the straight line where the nozzle is located when the Pearson correlation coefficient is the largest is selected as the intercept coefficient of the nozzle position function, and the slope coefficient of the nozzle position function is equal to the slope of the straight line where the nozzle is located.

4. The online positioning method for the boundary curve of a molten metal jet according to any one of claims 1 to 3, characterized in that: Establishing a decision model to estimate the nozzle diameter and establishing the nozzle position function in parallel to obtain the upper and lower positioning points of the nozzle include: According to the preset upper boundary intercept and the preset lower boundary intercept, the intercept coefficients in the upper boundary curve coefficient and the lower boundary curve coefficient are changed respectively to obtain the upper boundary approximation curve and the lower boundary approximation curve; Determining a second traversal window for the nozzle position straight line traversal according to the upper boundary approximation curve and the lower boundary approximation curve, wherein the nozzle position straight line is a straight line corresponding to the nozzle position function; In the second traversal window, the nozzle position straight line is traversed along the horizontal coordinate direction at a preset step length, and the number of edge points in the binary edge image on the nozzle position straight line is counted to obtain a first statistical value; Counting the overlap rates of the upper boundary approximation curve and the lower boundary approximation curve with the edge points in the binary edge image within the jet development range starting from the nozzle to obtain a second statistical value, wherein the jet development range is equal to the nozzle diameter; Obtaining an upper boundary approximation curve and a lower boundary approximation curve corresponding to when the first statistical value and the second statistical value are simultaneously maximum, as the upper boundary positioning curve and the lower boundary positioning curve; The upper and lower positioning point coordinates of the nozzle are obtained according to the intersection of the upper boundary positioning curve, the lower boundary positioning curve and the nozzle position straight line.

5. The online positioning method of the molten metal jet boundary curve according to claim 4, characterized in that: The expressions of the upper boundary approximation curve and the lower boundary approximation curve are: where y 逼近1 and y 逼近2 are the expressions of the upper boundary approximation curve and the lower boundary approximation curve, respectively, l1 and l2 are the preset upper boundary intercept and the preset lower boundary intercept, respectively; The horizontal and vertical coordinates of the midpoint of the second traversal window satisfy the following values: where x 逼近 and y 逼近 They are the horizontal and vertical coordinates of the midpoint of the second traversal window, and satisfy y 逼近 =k final x 逼近 +b final , k final and b final are the slope coefficient and intercept coefficient of the nozzle position function, respectively.

6. The online positioning method of the molten metal jet boundary curve according to claim 5, characterized in that: The formula for calculating stream width is: Among them, A1' and B1' are the upper boundary curves in the optimal boundary curve The coefficients of A'2 and B'2 are the lower boundary curves in the optimal boundary curve The coefficient of and are the upper and lower positioning point coordinates of the nozzle, D(x) is the stream width when the horizontal coordinate is x, and Δx=|x up -x down |, where x up 、x down satisfy:

7. An online positioning system for a molten metal jet boundary curve, the system comprising: A memory (20), a processor (30), and a computer program stored in the memory (20) and executable on the processor (30), wherein the processor (30) implements the steps of the method according to any one of claims 1 to 6 when executing the computer program.

Images