Crystallizer nozzle centering method and system based on machine vision

Abstract

The present invention proposes a method and system for centering a crystallizer water inlet based on machine vision, which belongs to the technical field of centering a crystallizer water inlet. The method comprises: S1, installing a first image acquisition device on a continuous casting platform to acquire images of a water inlet of a tundish car and a water inlet of a crystallizer; S2, using guided image filtering to perform edge enhancement; S3, using an ant algorithm to extract edges of a first image set to obtain a second image set; S4, using morphological closing operations to perform smoothing and centerline extraction on the second image set to obtain a centerline equation; S5, centering the water inlet of the crystallizer in a width direction according to the centerline equation to obtain a width centering result; S6, symmetrically installing two second image acquisition devices on a center extension line of the water inlet of the crystallizer, and using the second image acquisition devices to perform thickness centering results; S7, adjusting the position of the tundish car according to the width centering results and the thickness centering results.

Description

Technical Field

[0001] The present invention relates to the technical field of crystallizer water inlet centering, and in particular to a crystallizer water inlet centering method and system based on machine vision. Background Art

[0002] With the continuous development of industrial production, the crystallization process plays an important role in the chemical industry and other fields. The design and position of the crystallizer nozzle have an important impact on the efficiency of the crystallization process and the product quality. Reasonable design is required to ensure that the solvent can be effectively discharged while avoiding the loss of crystallized products. In the continuous casting process, the centering technology of the tundish car nozzle in the width and thickness directions of the crystallizer nozzle is to ensure that the molten steel flows smoothly from the ladle into the nozzle position of the crystallizer, so as to ensure the uniform solidification of the molten steel in the continuous casting process and the high-quality production of the continuous casting billet.

[0003] The centering problem of the crystallizer water inlet has always been one of the key difficulties in process control. In the existing crystallizer water inlet, manual intervention is required to manually adjust and monitor the crystallizer water inlet. In addition, the crystalline material produced during the crystallization process may block the water inlet, resulting in the inability to discharge the solvent smoothly, which in turn leads to the inability to accurately identify the edge of the crystallizer water inlet, affecting the accuracy of the crystallizer water inlet centering.

[0004] Therefore, finding a centering method that can not only automatically center the crystallizer water inlet, but also accurately identify the edge of the crystallizer water inlet to ensure that the solution in the crystallizer can be smoothly discharged is a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] In view of this, the present invention proposes a method and system for centering the water nozzle of the crystallizer based on machine vision, which monitors the water nozzle of the continuous casting machine crystallizer in real time through machine vision, extracts the center line to center the width direction of the water nozzle of the crystallizer, and uses a laser sensor to center it in the thickness direction, thereby reducing human errors and improving production efficiency and product quality in the continuous casting process.

[0006] The technical solution of the present invention is achieved in this way:

[0007] In a first aspect, the present invention provides a method for centering a crystallizer nozzle based on machine vision, comprising the following steps:

[0008] S1. Installing a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, and obtaining an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer; wherein the continuous casting platform is also provided with a tundish car and a crystallizer;

[0009] S2, using guided image filtering to perform edge enhancement on the initial image of the tundish car outlet and the initial image of the crystallizer outlet, respectively, to obtain a first image set;

[0010] S3, using an ant algorithm to extract edges from the first image set to obtain a second image set;

[0011] S4, using morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain the centerline equation of the edge image of the water inlet of the tundish car and the centerline equation of the edge image of the water inlet of the crystallizer;

[0012] S5, aligning the mold water inlet in the width direction according to the center line equation of the edge image of the water inlet of the tundish car and the center line equation of the edge image of the water inlet of the crystallizer, and obtaining a width centering result;

[0013] S6. Two second image acquisition devices are symmetrically installed on the central extension line of the crystallizer water outlet, and the second image acquisition devices are used to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result;

[0014] S7. Adjust the position of the tundish car on the continuous casting platform according to the width alignment results and the thickness alignment results.

[0015] Based on the above technical solution, preferably, step S3 specifically includes:

[0016] S31, select any image in the first image set, select m nodes in the image, take the m nodes as ants, and set the initial value of the pheromone amount of the pixels of the m ants to 0.0001;

[0017] S32, calculating the local strength of each ant's location, formulating transition rules based on the local strength, and moving m ants to acceptable neighborhoods according to the volatility of pheromones;

[0018] S33, using a heuristic function to represent the perception of the ants, and selecting the neighborhood of each ant according to the heuristic function;

[0019] S34, select one ant from the m ants and move it to a position in its 8 neighborhoods according to the heuristic function, update the pheromone concentration of each ant, and estimate the pheromone concentration of each node;

[0020] S35 , determining a second image according to the pheromone concentration of the node and the movement path of the ant, and repeating steps S31 - S32 until all second images corresponding to the first image set are determined, and a second image set is formed.

[0021] On the basis of the above technical solution, preferably, the transition rule in step S32 is formulated according to the pheromone concentration and the heuristic function, wherein the expression of the heuristic function is as follows:

[0022]

[0023] Among them, η u，v represents the heuristic function, η 1 represents the difference in the image intensity of the first node, η 2 represents the difference in the image intensity of the second node, η 3 Indicates the difference in the image intensity of the third node, Iη max It represents the maximum intensity value among the three image nodes. The first node, the second node and the third node are the three nodes where the ants move arbitrarily.

[0024] Based on the above technical solution, preferably, in step S32, the probability calculation formula of the ant moving to an acceptable neighborhood according to the volatility of the pheromone is:

[0025]

[0026] in, represents the probability of an ant moving to an acceptable neighborhood (i, j) at the node (u, v) for the nth time, represents the amount of pheromone at the (u, v) node for the n-1th time, η u，v represents the heuristic information value of node (u, v), σ u，v represents the adjustment parameter at the (u, v) node, α represents the importance of pheromone, β represents the importance of information in the heuristic function, γ represents the coefficient of the adjustment parameter, Ω(i, j) represents the acceptable neighborhood of the ant, and Ω(u, v) represents the set of neighboring pixels to which the ant can move from the current pixel;

[0027] Adjustment parameter σ u，v The formula is as follows:

[0028]

[0029] Among them, d i，j represents the ideal property value of the tuning parameter, represents the sub-block gradient ratio, Indicates that the kth ant is in the Nth c The threshold value of times.

[0030] More preferably, step S4 specifically includes:

[0031] S41, performing a dilation operation and an erosion operation on the second image set respectively to obtain a third image set;

[0032] S42, respectively obtaining the grayscale value and the grayscale centroid position of the third image set, and obtaining the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the tundish car and the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the crystallizer;

[0033] S43. Use the least squares method to fit the grayscale centroid position of the pixel points on the water outlet edge image of the tundish car and the grayscale centroid position of the pixel points on the water outlet edge image of the crystallizer, respectively, to obtain the centerline equation of the water outlet edge image of the tundish car and the centerline equation of the water outlet edge image of the crystallizer.

[0034] On the basis of the above technical solution, preferably, step S6 specifically includes:

[0035] S61, symmetrically installing two second image acquisition devices on the central extension line of the horizontal width direction of the crystallizer of the continuous casting platform;

[0036] S62, aligning the water outlet of the tundish car with the water outlet of the crystallizer in the width direction, so that the water outlet of the tundish car is on the center line of the water outlet of the crystallizer;

[0037] S63, using the second image acquisition device to measure the distance between the busbar of the water outlet of the tundish car and the second image acquisition device, to obtain a first distance and a second distance; wherein the first distance and the second distance are respectively the distances between the water outlet of the tundish car and the two second image acquisition devices;

[0038] S64, centering the crystallizer water inlet in the thickness direction according to the first distance and the second distance to obtain a thickness centering result.

[0039] On the basis of the above technical solution, preferably, the second image acquisition device is a laser sensor, and laser ranging is used to center the crystallizer water outlet in the thickness direction.

[0040] In a second aspect, the present invention provides a crystallizer nozzle centering system based on machine vision, which adopts the crystallizer nozzle centering method as described above, comprising:

[0041] The first acquisition module is configured to install a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, so as to obtain an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer;

[0042] An edge processing module is configured to use guided image filtering to perform edge enhancement on the initial image of the tundish water outlet and the initial image of the crystallizer water outlet respectively to obtain a first image set, and use an ant algorithm to perform edge extraction on the first image set to obtain a second image set;

[0043] A centerline positioning module is configured to use morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain a centerline equation of the tundish car water inlet edge image and a centerline equation of the crystallizer water inlet edge image;

[0044] A width centering module, which is configured to center the mold water outlet in the width direction according to the center line equation of the tundish car water outlet edge image and the center line equation of the mold water outlet edge image to obtain a width centering result;

[0045] A second acquisition module is configured to symmetrically install two second image acquisition devices on the central extension line of the crystallizer water outlet, and use the second image acquisition devices to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result;

[0046] A thickness centering module, configured to use a second image acquisition device to center the crystallizer nozzle in the thickness direction to obtain a thickness centering result;

[0047] The position adjustment module is configured to adjust the position of the tundish car according to the width centering result and the thickness centering result.

[0048] In a third aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to implement the crystallizer water inlet centering method as described above.

[0049] In a fourth aspect, the present invention provides an electronic device, comprising: at least one processor, at least one memory, a communication interface, and a bus;

[0050] Wherein, the processor, memory, and communication interface communicate with each other via the bus;

[0051] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the crystallizer water inlet centering method as described above.

[0052] The crystallizer water inlet centering method of the present invention has the following beneficial effects compared with the prior art:

[0053] (1) The position and shape of the water outlet of the tundish car and the water outlet of the crystallizer are obtained by setting a first image acquisition device, and the edge extraction and detection of the initial image are performed by guided image filtering and ant algorithm, and then the width direction of the water outlet of the crystallizer is realized by using morphological closing operation and center line equation, and the water outlet of the crystallizer is centered in the thickness direction by using a second image acquisition device. The edge image of the water outlet of the crystallizer can be accurately identified and the water outlet of the crystallizer can be automatically centered, thereby improving the accuracy of the water outlet of the crystallizer;

[0054] (2) The ant colony algorithm and the heuristic function are combined to increase the local search capability of the ant algorithm. By using the heuristic function to provide a larger perception range, the ants can perceive the surrounding environment more comprehensively. The ants are moved and the pheromone concentration is updated according to the heuristic function. The pheromone distribution searched by the ant colony algorithm is used to realize the edge extraction of the image, thereby improving the effect and accuracy of edge extraction.

[0055] (3) By performing a morphological closing operation on the second image set and obtaining the centerline equation of the edge image through the position of the grayscale centroid and the least squares method, the position and shape of the water outlet of the tundish car and the water outlet of the crystallizer are determined according to the position of the grayscale centroid, and the centerline is fitted using the least squares method to align the water outlet of the crystallizer in the width direction, thereby improving the accuracy of the centering of the water outlet of the crystallizer. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a flow chart of the method for centering the water inlet of a crystallizer of the present invention;

[0058] Figure 2 It is a specific flow chart of the method for centering the water inlet of a crystallizer of the present invention;

[0059] Figure 3 It is a schematic diagram of the width direction of the crystallizer water inlet centering method of the present invention;

[0060] Figure 4 It is a structural diagram of different factions in the ant algorithm of the crystallizer water inlet centering method of the present invention;

[0061] Figure 5 The ant algorithm flow chart of the crystallizer water inlet centering method of the present invention is as follows;

[0062] Figure 6 It is a schematic diagram of an ant neighborhood of the method for centering the water inlet of a crystallizer of the present invention;

[0063] Figure 7 It is a schematic diagram of the width direction of the crystallizer water inlet centering method of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] like Figure 1 and Figure 2 As shown, the present invention provides a method for centering a crystallizer nozzle based on machine vision, comprising the following steps:

[0066] S1. Installing a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, and obtaining an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer; wherein the continuous casting platform is also provided with a tundish car and a crystallizer;

[0067] S2, using guided image filtering to perform edge enhancement on the initial image of the tundish car outlet and the initial image of the crystallizer outlet, respectively, to obtain a first image set;

[0068] S3, using an ant algorithm to extract edges from the first image set to obtain a second image set;

[0069] S4, using morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain the centerline equation of the edge image of the water inlet of the tundish car and the centerline equation of the edge image of the water inlet of the crystallizer;

[0070] S5, aligning the mold water inlet in the width direction according to the center line equation of the edge image of the water inlet of the tundish car and the center line equation of the edge image of the water inlet of the crystallizer, and obtaining a width centering result;

[0071] S6. Two second image acquisition devices are symmetrically installed on the central extension line of the crystallizer water outlet, and the second image acquisition devices are used to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result;

[0072] S7. Adjust the position of the tundish car on the continuous casting platform according to the width alignment results and the thickness alignment results.

[0073] like Figure 3As shown, in the embodiment of the present application, the position and shape information of the water outlet 2 of the tundish car and the water outlet 3 of the crystallizer are accurately obtained by the first image acquisition device, and the edge enhancement and detection are performed using the guided image filtering and optimization method, so that the contours of the water outlet 2 of the tundish car and the water outlet 3 of the crystallizer are clearly visible, and the centerline equations of the edge images of the water outlet 2 of the tundish car and the water outlet 3 of the crystallizer are respectively obtained through morphological closing operation and centerline extraction, so as to realize the precise positioning of the water outlet 3 of the crystallizer in the width direction, and the water outlet 3 of the crystallizer is centered in the width direction and the thickness direction according to the centerline equation, and the centering results of the width and thickness can be obtained, and finally the position of the tundish car 1 is adjusted according to the width centering results and the thickness centering results to ensure its centering with the water outlet of the crystallizer, thereby improving the accuracy and stability of the continuous casting process production process.

[0074] In one embodiment of the present application, step S2 uses a guided filter to perform guided filtering on the initial image, and step S2 specifically includes:

[0075] S21, prepare a guide image I u and a target image p u , where the guide image I u and a target image p u can be the same image; where the target image p u is the initial image;

[0076] S22, assuming that in the selected window ω centered at pixel k k The linear transformation in is used as the guided filter, and the image q is obtained after filtering by the guided filter. u , the formula for filtering using the guided filter is:

[0077]

[0078] Among them, q u Represents the target image p u The filtered image, φ k and α k are linear coefficients, which are constants, I u represents the guide image, which can be set according to the actual situation, and u represents the number of initial images;

[0079] S23, for image q u Modeling is performed to remove noise, texture, etc., and the first image q is obtained i , the formula is as follows:

[0080] q i =q u -n u

[0081] Among them, n u Indicates noise.

[0082] S24, determining the minimized noise variance to ensure the quality of the filtered image, wherein the minimized noise variance is:

[0083]

[0084] S25, in window ω k Minimize the cost function in , where the formula of the cost function is as follows:

[0085]

[0086] Among them, E(φ k , α k ) represents the cost function, ε is the regularization parameter, and the cost function can be solved by the ridge regression model to obtain the optimal filtering result. After solving the cost function, we can get:

[0087]

[0088] The window ω k Middle guide image I u The mean and variance are μ k and σ k 2 , window ω k The number of pixels in is |ω|.

[0089] S26. For different windows, the optimal parameter values ​​are calculated in the image, and the solution obtained by the ridge regression module is applied to obtain the first image q u , the first image q u The expression is:

[0090]

[0091] It can be understood that guided filtering is an adaptive filtering method that can adjust filtering parameters according to the pixel values ​​in the local window, thereby helping to reduce the noise in the image, especially Gaussian noise and salt and pepper noise. Unlike some traditional smoothing filters, guided filtering can reduce noise while retaining the edge information of the image.

[0092] In the embodiment of the present application, by minimizing the noise variance and the cost function, the guided filter can effectively remove the noise in the image, and the guided filtering process on the initial image can reduce the noise while maintaining the details and edge information of the image.

[0093] In an embodiment of the present application, step S3 specifically includes:

[0094] S31, select any image in the first image set, select m nodes in the image, take the m nodes as ants, and set the initial value of the pheromone amount of the pixels of the m ants to 0.0001;

[0095] S32, calculating the local strength of each ant's location, formulating transition rules based on the local strength, and moving m ants to acceptable neighborhoods according to the volatility of pheromones;

[0096] S33, using a heuristic function to represent the perception of the ants, and selecting the neighborhood of each ant according to the heuristic function;

[0097] S34, select one ant from the m ants and move it to a position in its 8 neighborhoods according to the heuristic function, update the pheromone concentration of each ant, and estimate the pheromone concentration of each node;

[0098] S35 , determining a second image according to the pheromone concentration of the node and the movement path of the ant, and repeating steps S31 - S32 until all second images corresponding to the first image set are determined, and a second image set is formed.

[0099] It is understandable that the ant colony algorithm has a certain degree of adaptability, and can automatically adjust the ant's search path according to the characteristics of the image and the pheromone concentration, which is suitable for the detection of different types of images and edge features; at the same time, the ant colony algorithm has a certain degree of parallel performance, and can accelerate the edge extraction process through reasonable design and implementation, and improve the efficiency and practicality of the algorithm. This application combines the ant colony algorithm with the heuristic function to update the pheromone, combines the search ability of the ant colony algorithm with the accuracy of edge extraction, and can effectively extract the edge of the image and improve the accuracy and precision of detection.

[0100] In the embodiment of the present application, the image pixels of the first image set are assumed to be nodes of the graph, and a global search of the image is realized by selecting m nodes as ant colonies. The initial value of the pixel pheromone amount of the ant is set to 0.0001 to ensure that the ant has a certain pheromone guidance during the search process. The local intensity of each ant's position is calculated, so that the ant is more inclined to choose the area with higher local intensity during the movement, thereby increasing the local search ability of the ant algorithm. At the same time, the transition rules are formulated according to the local intensity, which can make the ant more intelligent when moving. The next position is selected, which helps to improve the search efficiency of the algorithm. The use of heuristic functions can provide a larger perception range, so that the ants can perceive the surrounding environment more comprehensively. The ants are moved according to the heuristic function and the pheromone concentration is updated, so that the ants can adjust the path in real time according to the changes in the environment during the search process. At the same time, the pheromone concentration is updated and the pheromone concentration of each node is estimated so that the algorithm can more accurately reflect the characteristics and edge information of different areas in the image. The edge extraction is performed according to the pheromone concentration of the node and the moving path of the ant. The pheromone distribution searched by the ant colony algorithm is used to realize the edge extraction of the image, thereby obtaining the second image, which helps to improve the effect and accuracy of edge extraction.

[0101] Preferably, the heuristic function uses a 5×5 window to improve the efficiency of ant search. The user can set the window size according to actual usage, and this application does not make any specific restrictions on this.

[0102] Specifically, the transition rule in step S32 is formulated according to the pheromone concentration and the heuristic function, wherein the expression of the heuristic function is as follows:

[0103]

[0104] Among them, η u，v represents the heuristic function, η 1 represents the difference in the image intensity of the first node, η 2 represents the difference in the image intensity of the second node, η 3 Indicates the difference in the image intensity of the third node, Iη max It represents the maximum intensity value among the three image nodes. The first node, the second node and the third node are the three nodes where the ants move arbitrarily.

[0105] Furthermore, the calculation formula of the difference in image intensity is as follows:

[0106] η 1=[In(u-2,v-2)+In(u-1,v-2)+In(u,v-2)+In(u+1,v-2)+In(u+2,v-2)]-[In(u-2,v+2)+In(u-1,v+2)+In(u,v+2)+In(u+1,v+2)+In(u+2,v+2)]

[0107] η 2 =[In(u-2,v-1)+In(u-1,v-1)+In(u,v-1)+In(u+1,v-1)+In(u+2,v-1)]-[In(u-2,v+1)+In(u-1,v+1)+In(u,v+1)+In(u+1,v+1)+In(u+2,v+1)]

[0108] η 1 =[In(u-2,v)+In(u-1,v)]-[In(u+1,v)+In(u+2,v)]

[0109] Here, In(u, v) represents the intensity at the position (u, v).

[0110] Furthermore, in step S32, the probability calculation formula of the ant moving to an acceptable neighborhood according to the volatility of the pheromone is:

[0111]

[0112] in, represents the probability of an ant moving to an acceptable neighborhood (i, j) at the node (u, v) for the nth time, represents the amount of pheromone at the (u, v) node for the n-1th time, η u，v represents the heuristic information value of node (u, v), σ u，v represents the adjustment parameter at the (u, v) node, α represents the importance of pheromone, β represents the importance of information in the heuristic function, γ represents the coefficient of the adjustment parameter, Ω(i, j) represents the acceptable neighborhood of the ant, and Ω(u, v) represents the set of neighboring pixels to which the ant can move from the current pixel;

[0113] Adjustment parameter σ u，v The formula is as follows:

[0114]

[0115] Among them, d i，j represents the ideal property value of the tuning parameter, represents the sub-block gradient ratio, Indicates that the kth ant is in the Nth c The threshold value of times.

[0116] Understandably, in the ant colony algorithm, the threshold of ants is usually called the exploration probability of ants or the importance factor of heuristic information. It is used to determine whether the ants adopt an exploration strategy or a tracking strategy using existing pheromones when selecting the next target. The threshold of ants represents the degree of preference of ants for exploration. When the threshold of ants is small, ants are more inclined to choose to explore new paths to find better solutions; when the threshold of ants is large, ants are more inclined to choose existing paths to use existing pheromone information.

[0117] In one embodiment of the present application, assuming that the pixels of the first image are nodes, m nodes are randomly selected as ants in the first image and the pheromone amount of each ant's pixel is taken as 0.0001. The ant is moved from its initial position (u, v) to its acceptable neighborhood Ω(i, j) according to the probability or transition rule, and a 5×5 heuristic function is used to represent the ant's perception of nodes near a few points (u, v). Figure 4 As shown, three different faction representatives are set. The figure shows different faction structures. Pixels of the same color are taken as a group (where different letters represent different color numbers). The heuristic value η of the node (u, v) is calculated by subtracting the sum of the pixel values ​​of each group from the sum of the corresponding group of the same color. u，v , and select the maximum value in the differential pair. Select one ant from the m ants to move, and each selected ant can move to its 8 neighborhoods, that is, each ant moves to the node with the highest probability, as shown in the figure. Repeat this process until L moves are made. Then, each ant has a fixed memory length, recorded as l. According to the l nodes that the ant has moved, determine the edge extraction image of the first image, that is, the second image.

[0118] In a further embodiment of the present application, the pheromone concentration of each ant is iteratively updated by moving the ant L steps, and the updated calculation formula is as follows:

[0119]

[0120] Among them, τ (u，v) (NEW) represents the updated pheromone amount at the (u, v) node, τ (u，v) (old) represents the amount of pheromone at the (u, v) node before the update, represents the pheromone difference before and after the sth ant information concentration is updated, ρ(n) represents the pheromone volatility, s represents the total number of ants at the (u, v) node, τ min represents the minimum information value, τ max Indicates the maximum pheromone value.

[0121] Furthermore, the calculation formula for the pheromone amount at each (u, v) node is as follows:

[0122]

[0123] Among them, s th represents the sth ant, and t is the preset threshold, which limits the deposition amount of pheromone.

[0124] Furthermore, adjusting the parameter σ u，v The volatility of information is also improved in a dynamic way. The volatility of information ρ(n) is calculated as follows:

[0125]

[0126] Among them, N c Indicates the number of iterative updates.

[0127] In the embodiment of the present application, the ant colony algorithm is updated by moving ants to an acceptable neighborhood, such as Figure 5 and Figure 6 As shown, Figure 5 This is the flow chart of the ant algorithm. Figure 6 The diagram is a schematic diagram of the ant neighborhood. The first image is read, and m nodes are randomly selected as ants in the first image. The ant's pheromone matrix is ​​initialized, and each pixel value of the image corresponding to the heuristic matrix is ​​calculated. An ant is selected from the m ants. The current position of the ant is (i, j). The ant moves L steps to the acceptable neighborhood according to the rules. It is determined whether all ants move. When all ants move, the pheromone matrix is ​​updated. It is determined whether the iteration is completed. If not, one of the m ants is selected and moved according to the rules until all iterations are completed. The ant algorithm is used to perform global search and local optimization of the image, combined with pheromone guidance and intelligent search of heuristic functions, so as to realize the perception and extraction of image features, thereby improving the effect and accuracy of edge extraction.

[0128] In an embodiment of the present application, step S4 specifically includes:

[0129] S41, performing a dilation operation and an erosion operation on the second image set respectively to obtain a third image set;

[0130] S42, respectively obtaining the grayscale value and the grayscale centroid position of the third image set, and obtaining the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the tundish car and the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the crystallizer;

[0131] S43. Use the least squares method to fit the grayscale centroid position of the pixel points on the water outlet edge image of the tundish car and the grayscale centroid position of the pixel points on the water outlet edge image of the crystallizer, respectively, to obtain the centerline equation of the water outlet edge image of the tundish car and the centerline equation of the water outlet edge image of the crystallizer.

[0132] It can be understood that the calculation formula of step S41 is as follows:

[0133]

[0134] Where A represents the second image, B represents the structural element, represents the dilation operation, and Θ represents the erosion operation.

[0135] The contour is smoothed by performing a morphological closing operation on the second image. Meanwhile, the elongated grooves and narrow discontinuities in the second image can be bridged and smaller holes can be eliminated.

[0136] The calculation formula for the gray value of a pixel is as follows:

[0137]

[0138] Among them, f(w) represents the gray value of pixel w, f th It indicates the grayscale threshold value, which can be set according to the actual situation. w The upper and lower limits of the column pixel bar are [a(x w ), b(x w )], the weight formula of the grayscale centroid is as follows:

[0139]

[0140] Among them, f(x w , y) represents the pixel coordinate (x w , the gray value at y(x w ) represents the calculated xth w The ordinate of the centroid of the column pixel bar.

[0141] In the embodiment of the present application, the pixel strip center point data extracted according to the grayscale centroid method is basically distributed near a straight line, and the equation of the straight line is:

[0142]

[0143] in, represents the equation of a straight line, and represents the slope and intercept of the straight line, Indicates the center point of a pixel bar.

[0144] The fitting criterion formula of the least squares method is as follows:

[0145]

[0146] Among them, E minrepresents the fitting criterion, Represents the equation of the ith line.

[0147] In the embodiment of the present application, the target object in the image is expanded by dilation operation to obtain a more complete target area, and the target object in the image is reduced by corrosion operation to remove noise and small discontinuous parts, thereby making the edge clearer. The position and shape of the water outlet of the tundish car and the water outlet of the crystallizer are determined by the position of the grayscale center of gravity, which improves the recognition and positioning accuracy of the water outlet of the tundish car and the water outlet of the crystallizer, and further improves the accuracy of the centering of the water outlet of the crystallizer.

[0148] In an embodiment of the present application, step S6 specifically includes:

[0149] S61, symmetrically installing two second image acquisition devices on the central extension line of the horizontal width direction of the crystallizer of the continuous casting platform;

[0150] S62, aligning the water outlet of the tundish car with the water outlet of the crystallizer in the width direction, so that the water outlet of the tundish car is on the center line of the water outlet of the crystallizer;

[0151] S63, using the second image acquisition device to measure the distance between the busbar of the water outlet of the tundish car and the second image acquisition device, to obtain a first distance and a second distance; wherein the first distance and the second distance are respectively the distances between the water outlet of the tundish car and the two second image acquisition devices;

[0152] S64, centering the crystallizer water inlet in the thickness direction according to the first distance and the second distance to obtain a thickness centering result.

[0153] like Figure 7 As shown, in the embodiment of the present application, the second image acquisition device 4 is used to realize all-round image acquisition in the horizontal width direction of the crystallizer, ensuring that the observation angle of the crystallizer water inlet is more comprehensive, thereby improving the accuracy and stability of the centering operation, and by aligning the water outlet of the tundish car with the water outlet of the crystallizer in the width direction, the water outlet of the tundish car is placed on the center line of the water outlet of the crystallizer, ensuring the centering of the water outlet of the tundish car with the water outlet of the crystallizer in the horizontal width direction; by measuring the distance between the generatrix of the water outlet of the tundish car and the second image acquisition device 4, the position of the water outlet of the crystallizer in the thickness direction is determined, which facilitates the adjustment of the water outlet of the crystallizer in the thickness direction, thereby improving the uniformity and quality of the product. By adjusting the centering in the thickness direction, the uniformity and stability of the ingot in the thickness direction are ensured.

[0154] In one embodiment of the present application, the second image acquisition device 4 is a laser sensor, and laser ranging is used to center the crystallizer water outlet in the thickness direction.

[0155] As you can understand, when using a laser sensor to measure distance, the laser emitting diode first aims at the target and emits a laser pulse. After being reflected by the target, the laser is scattered in all directions, and part of the scattered light returns to the sensor receiver, is received by the optical system, and is imaged on the avalanche photodiode. The avalanche photodiode is an optical sensor with an internal amplification function, so it can detect extremely weak light signals and convert them into corresponding electrical signals. By measuring the time it takes for the light pulse to travel back and forth to the point to be measured, multiplying it by the speed of light and dividing it by 2, the distance of the target to be measured can be calculated.

[0156] As shown in the figure, in the embodiment of the present application, in order to realize the alignment of the crystallizer water inlet 3 and the tundish car water outlet 3 in the thickness direction, a ready-made laser sensor is used to measure the distance, so as to realize the accurate measurement of the crystallizer water inlet 3 in the thickness direction, and to timely find and adjust when the position error is found. Specifically, two laser sensors 4 are first symmetrically installed on the extension line of the center line of the crystallizer water inlet 3 on the continuous casting platform. After the tundish car water outlet 2 and the crystallizer water outlet 3 are aligned in the width direction, the tundish car water outlet 2 is just on the center line of the crystallizer water outlet 4. At this time, the distance between the busbar of the tundish car water outlet 2 and the laser sensor is measured by the laser sensor, and the distance between the tundish car water outlet 2 and the laser sensors on both sides can be obtained respectively. By comparing the distance between the tundish car water outlet 2 on the left and right sides and the laser sensor, it can be known whether the position of the tundish car water outlet 2 in the thickness direction is biased to the left or right, and then a moving distance is provided to the control system of the tundish car 1 to realize automatic centering in the thickness direction of the crystallizer.

[0157] The present invention provides a crystallizer water inlet centering system based on machine vision, which adopts the crystallizer water inlet centering method as described above, and comprises:

[0158] The first acquisition module is configured to install a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, so as to obtain an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer;

[0159] An edge processing module is configured to use guided image filtering to perform edge enhancement on the initial image of the tundish water outlet and the initial image of the crystallizer water outlet respectively to obtain a first image set, and use an ant algorithm to perform edge extraction on the first image set to obtain a second image set;

[0160] A centerline positioning module is configured to use morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain a centerline equation of the tundish car water inlet edge image and a centerline equation of the crystallizer water inlet edge image;

[0161] A width centering module, which is configured to center the mold water outlet in the width direction according to the center line equation of the tundish car water outlet edge image and the center line equation of the mold water outlet edge image to obtain a width centering result;

[0162] A second acquisition module is configured to symmetrically install two second image acquisition devices on the central extension line of the crystallizer water outlet, and use the second image acquisition devices to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result;

[0163] A thickness centering module, configured to use a second image acquisition device to center the crystallizer nozzle in the thickness direction to obtain a thickness centering result;

[0164] The position adjustment module is configured to adjust the position of the tundish car according to the width centering result and the thickness centering result.

[0165] In the embodiment of the present application, the crystallizer water inlet centering system can automatically perform the centering operation on the crystallizer water inlet without manual intervention, thereby improving production efficiency and reducing labor costs. Through the first acquisition module and the second acquisition module, the system can collect images of the crystallizer water inlet in real time, monitor the position and shape of the water inlet of the tundish car and the crystallizer water inlet in real time, and process and analyze them to ensure the safety and stability of the continuous casting process; using edge enhancement and edge extraction technology, combined with morphological closing operations to extract the center line, it is possible to achieve high-precision centering of the crystallizer water inlet, ensure the centering accuracy of the crystallizer water inlet in the width and thickness directions, and improve the consistency and quality of the product; according to the width centering results and the thickness centering results, the position of the tundish car can be adjusted in real time to maintain the centering state of the crystallizer water inlet, so that the crystallizer water inlet centering system can quickly respond to changes in the production process, ensuring the stability of the continuous casting process and the quality of the product.

[0166] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to implement the crystallizer water inlet centering method as described above.

[0167] The present invention provides an electronic device, comprising: at least one processor, at least one memory, a communication interface and a bus; wherein the processor, the memory, and the communication interface communicate with each other through the bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the crystallizer water inlet centering method as described above.

[0168] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for centering a crystallizer nozzle based on machine vision, characterized in that: The following steps are involved: S1. Installing a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, and obtaining an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer; wherein the continuous casting platform is also provided with a tundish car and a crystallizer; S2, using guided image filtering to perform edge enhancement on the initial image of the tundish car outlet and the initial image of the crystallizer outlet, respectively, to obtain a first image set; S3, using an ant algorithm to extract edges from the first image set to obtain a second image set; S4, using morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain the centerline equation of the edge image of the water inlet of the tundish car and the centerline equation of the edge image of the water inlet of the crystallizer; S5, aligning the mold water inlet in the width direction according to the center line equation of the edge image of the water inlet of the tundish car and the center line equation of the edge image of the water inlet of the crystallizer, and obtaining a width centering result; S6. Two second image acquisition devices are symmetrically installed on the central extension line of the crystallizer water outlet, and the second image acquisition devices are used to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result; S7. Adjust the position of the tundish car on the continuous casting platform according to the width centering result and the thickness centering result; Step S3 specifically includes: S31, select any image in the first image set, select m nodes in the image, take the m nodes as ants, and set the initial value of the pheromone amount of the pixels of the m ants to 0.0001; S32, calculating the local strength of each ant's location, formulating transition rules based on the local strength, and moving m ants to acceptable neighborhoods according to the pheromone volatility rate; S33, using a heuristic function to represent the perception of the ants, and selecting the neighborhood of each ant according to the heuristic function; S34, select one ant from the m ants and move it to a position in its 8 neighborhoods according to the heuristic function, update the pheromone concentration of each ant, and estimate the pheromone concentration of each node; S35 , determining a second image according to the pheromone concentration of the node and the movement path of the ant, and repeating steps S31 - S32 until all second images corresponding to the first image set are determined, and a second image set is formed.

2. A method for centering a mold nozzle based on machine vision as claimed in claim 1, characterized in that: In step S32, the transition rule is formulated according to the pheromone concentration and the heuristic function, wherein the expression of the heuristic function is as follows: Among them, η u，v represents the heuristic function, η1 represents the difference in the image intensity of the first node, η2 represents the difference in the image intensity of the second node, η3 represents the difference in the image intensity of the third node, In max It represents the maximum intensity value among the three image nodes. The first node, the second node and the third node are the three nodes where the ants move arbitrarily.

3. A method for centering a crystallizer nozzle based on machine vision as claimed in claim 1, characterized in that: In step S32, the probability calculation formula of the ant moving to an acceptable neighborhood according to the volatility of the pheromone is: in, represents the probability of an ant moving to an acceptable neighborhood (i, j) at the node (u, v) for the nth time, represents the amount of pheromone at the (u, v) node for the n-1th time, η u，v represents the heuristic information value of node (u, v), σ u，v represents the adjustment parameter at the (u, v) node, α represents the importance of pheromone, β represents the importance of information in the heuristic function, γ represents the coefficient of the adjustment parameter, Ω(i, j) represents the acceptable neighborhood of the ant, and Ω(u, v) represents the set of neighboring pixels to which the ant can move from the current pixel; Adjustment parameter σ u，v The formula is as follows: Among them, d i，j represents the ideal property value of the tuning parameter, represents the sub-block gradient ratio, Indicates that the kth ant is in the Nth c The threshold value of times.

4. A method for centering a crystallizer nozzle based on machine vision as claimed in claim 1, characterized in that: Step S4 specifically includes: S41, performing a dilation operation and an erosion operation on the second image set respectively to obtain a third image set; S42, respectively obtaining the grayscale value and the grayscale centroid position of the third image set, and obtaining the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the tundish car and the grayscale centroid position of the pixel point on the image of the edge of the water outlet of the crystallizer; S43. Use the least squares method to fit the grayscale centroid position of the pixel points on the water outlet edge image of the tundish car and the grayscale centroid position of the pixel points on the water outlet edge image of the crystallizer, respectively, to obtain the centerline equation of the water outlet edge image of the tundish car and the centerline equation of the water outlet edge image of the crystallizer.

5. The method for centering a crystallizer nozzle based on machine vision according to claim 1, characterized in that: Step S6 specifically includes: S61, symmetrically installing two second image acquisition devices on the central extension line of the horizontal width direction of the crystallizer of the continuous casting platform; S62, aligning the water outlet of the tundish car with the water outlet of the crystallizer in the width direction, so that the water outlet of the tundish car is on the center line of the water outlet of the crystallizer; S63, using the second image acquisition device to measure the distance between the busbar of the water outlet of the tundish car and the second image acquisition device, to obtain a first distance and a second distance; wherein the first distance and the second distance are respectively the distances between the water outlet of the tundish car and the two second image acquisition devices; S64, centering the crystallizer water inlet in the thickness direction according to the first distance and the second distance to obtain a thickness centering result.

6. A method for centering a mold nozzle based on machine vision as claimed in claim 5, characterized in that: The second image acquisition device is a laser sensor, and laser ranging is used to center the crystallizer water outlet in the thickness direction.

7. A crystallizer nozzle centering system based on machine vision, characterized in that: The method for centering a crystallizer nozzle according to any one of claims 1 to 6 comprises: The first acquisition module is configured to install a first image acquisition device on the continuous casting platform to acquire images of the water outlet of the tundish car and the water outlet of the crystallizer, so as to obtain an initial image of the water outlet of the tundish car and the initial image of the water outlet of the crystallizer; An edge processing module is configured to use guided image filtering to perform edge enhancement on the initial image of the tundish water outlet and the initial image of the crystallizer water outlet respectively to obtain a first image set, and use an ant algorithm to perform edge extraction on the first image set to obtain a second image set; A centerline positioning module is configured to use morphological closing operations to perform smoothing and centerline extraction on the second image set, respectively, to obtain a centerline equation of the tundish car water inlet edge image and a centerline equation of the crystallizer water inlet edge image; A width centering module, which is configured to center the mold water outlet in the width direction according to the center line equation of the tundish car water outlet edge image and the center line equation of the mold water outlet edge image to obtain a width centering result; A second acquisition module is configured to symmetrically install two second image acquisition devices on the central extension line of the crystallizer water outlet, and use the second image acquisition devices to center the crystallizer water outlet in the thickness direction to obtain a thickness centering result; A thickness centering module, configured to use a second image acquisition device to center the crystallizer nozzle in the thickness direction to obtain a thickness centering result; The position adjustment module is configured to adjust the position of the tundish car according to the width centering result and the thickness centering result.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to implement the crystallizer water nozzle centering method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: at least one processor, at least one memory, a communication interface, and a bus; Wherein, the processor, memory, and communication interface communicate with each other via the bus; The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the crystallizer water inlet centering method according to any one of claims 1 to 6.

Images