Visual positioning method of steel ladle sliding nozzle mechanism

Abstract

The invention belongs to the technical field of casting production, and particularly relates to a visual positioning method of a steel ladle sliding nozzle mechanism, which comprises the following steps: acquiring three-dimensional point cloud data of the bottom of the steel ladle sliding nozzle mechanism; preprocessing the obtained point cloud data; feature extraction is carried out on the preprocessed point cloud data, the extracted features are registered with pre-stored template features, and a transformation relation is determined; overall transformation is conducted on the three-dimensional point cloud data based on the transformation relation, three-dimensional construction is conducted through the transformed three-dimensional point cloud data, and the accurate position of a nozzle at the bottom of the steel ladle sliding nozzle mechanism in the steel ladle sliding nozzle mechanism is calculated according to a constructed three-dimensional model; and sending the calculated position coordinates to a mechanism control system. The technical problems of molten steel fluidity and product quality caused by inaccurate positioning of the position of the steel ladle sliding nozzle in the prior art are solved, high-precision positioning of the nozzle is realized, and butt-joint assembly of the long nozzle and the nozzle at the bottom of the steel ladle sliding nozzle mechanism is facilitated.

Description

Technical Field

[0001] The invention belongs to the technical field of casting production, and in particular relates to a visual positioning method for a ladle sliding nozzle mechanism. Background Art

[0002] During the continuous casting process, the ladle sliding gate mechanism is a key device used to control the flow of molten steel in steel production. It is located at the bottom of the ladle and controls the flow of molten steel by controlling the opening and closing of the sliding gate. The sliding gate mechanism is usually made of refractory materials that can withstand high temperatures and corrosion. By adjusting the position of the sliding gate, the flow rate and flow of molten steel can be controlled to meet the needs of the smelting process.

[0003] A long shroud is installed at the bottom of the ladle sliding nozzle. The long shroud is a long strip-shaped steel outlet. When the molten steel reaches the required casting temperature and composition, the molten steel can flow out of the ladle through the long shroud and enter the tundish and continuous casting machine for casting. The long shroud can protect the molten steel from contact with air during the process of flowing from the ladle to the tundish, thereby preventing the molten steel from secondary oxidation of the ladle and preventing the steel flow from splashing.

[0004] At present, the connection between the ladle sliding nozzle and the long nozzle is done manually or by a robot arm relying on sensors or encoders to position the long nozzle and install it on the ladle sliding nozzle. Manual operation is done by manually identifying the position of the lower nozzle with the naked eye and then operating the hydraulic robot arm to complete the installation of the long nozzle. The lower nozzle installation working environment is harsh, accompanied by high temperature and high dust, and there is also the risk of molten steel splashing. The labor intensity of the workers is high and it will also cause harm to the body. The use of a robot arm for installation affects the positioning accuracy and stability of the ladle sliding nozzle mechanism.

[0005] The position accuracy of the ladle sliding nozzle is crucial to ensure the fluidity of molten steel and product quality. Therefore, steel mills urgently need a ladle sliding nozzle positioning method to solve the above problems. Summary of the invention

[0006] In response to the above problems, the present invention provides a visual positioning method for a ladle sliding nozzle mechanism, which is used to perform real-time online positioning of the ladle sliding nozzle mechanism in a real-time complex environment. By performing real-time visual positioning of the ladle sliding nozzle mechanism as a whole, the precise positions of the slide plate mechanism and the lower water outlet can be obtained, thereby providing a prerequisite for the subsequent automatic installation of the long water outlet.

[0007] The present invention provides a visual positioning method for a ladle sliding nozzle mechanism, the method comprising:

[0008] Step 1: Obtain the 3D point cloud data of the bottom of the ladle sliding gate mechanism;

[0009] Step 2: pre-processing the acquired point cloud data of the bottom of the ladle sliding gate mechanism;

[0010] Step 3: Extract features from the preprocessed point cloud data to obtain the features of the bottom of the ladle sliding nozzle mechanism, align the extracted features with the pre-stored template features, and determine the transformation relationship between the extracted features and the template features;

[0011] Step 4: transform the three-dimensional point cloud data as a whole based on the transformation relationship, use the transformed three-dimensional point cloud data to perform three-dimensional construction, and calculate the exact position of the water inlet at the bottom of the ladle sliding water inlet mechanism according to the constructed three-dimensional model;

[0012] Step 5: Send the calculated bottom water outlet position coordinates of the ladle sliding water outlet mechanism to the mechanism control system.

[0013] Furthermore, the process of aligning the extracted features with the pre-stored template features in step 3 includes: transforming the point cloud features, matching the transformed point cloud features and the template features after each transformation, and calculating the matching degree between the two until the matching degree between the two meets a predetermined condition.

[0014] Furthermore, the three-dimensional construction process in step 4 includes: using a grid generation method to divide the transformed three-dimensional point cloud data into a uniform three-dimensional voxel grid, and counting the number of points contained in each voxel, and according to the set voxel resolution, converting the point cloud data into a dense voxel grid representation to form a three-dimensional model of the bottom of the ladle sliding nozzle mechanism.

[0015] Furthermore, in step 4, the process of overall transformation of the three-dimensional point cloud data based on the transformation relationship includes: for each point in the three-dimensional point cloud, determining its position coordinates in the point cloud coordinate system, and projecting the point into the local coordinate system of the ladle sliding gate mechanism according to the obtained transformation relationship.

[0016] Furthermore, the step 3 uses a nearest neighbor matching method to match the transformed point cloud features and template features. The matching process includes: for each transformed point cloud feature point, using a nearest neighbor search algorithm to find the nearest neighbor point of each point cloud feature point in the template feature, taking each pair of matching points as a group of matching point pairs, calculating the Euclidean distance between all matching point pairs, and taking the average value as a measure of the current matching degree and comparing it with a preset threshold to determine whether the predetermined conditions are met, repeating the transformation, nearest neighbor matching and matching degree calculation until the matching degree reaches the predetermined conditions.

[0017] Further, the process of judging whether the predetermined condition is met includes: if the obtained average value is ≤ a set threshold value, the matching degree meets the predetermined condition; if the obtained average value is greater than the set threshold value, the matching degree does not meet the predetermined condition.

[0018] Furthermore, in step 2, the point cloud data preprocessing is performed using a wavelet denoising method.

[0019] Furthermore, the step 2 specifically includes:

[0020] Step 2.1: Perform wavelet transform on the point cloud data at the bottom of the ladle sliding gate mechanism, decompose the point cloud data to obtain signal components at various frequencies, and reflect the characteristics of the point cloud data at different spatial frequencies based on the signal components;

[0021] Step 2.2: Apply threshold processing to each decomposed frequency component and determine the appropriate threshold for noise removal;

[0022] Step 2.3: Recombine the processed frequency components to form denoised point cloud data.

[0023] The beneficial effects of the present invention are:

[0024] First, the present invention can accurately determine the water nozzle position at the bottom of the ladle sliding water nozzle mechanism through a complete visual acquisition, preprocessing, feature extraction, registration, three-dimensional construction and position calculation process. This comprehensive method ensures high-precision data acquisition and processing, effectively improves the accuracy of the entire positioning system, and can stably acquire and process point cloud data under different environmental conditions (such as high temperature, smoke, etc.), maintaining a high level of stability and reliability. From data acquisition to position calculation and transmission, each step is automatically connected, making the operation process of the entire ladle sliding water nozzle mechanism more intelligent and efficient, reducing manual intervention and errors;

[0025] Second, in the preferred implementation, the present invention introduces a normal vector estimation method and a nearest neighbor search algorithm to accurately extract and match point cloud features. This method not only improves the efficiency of feature extraction, but also ensures high-precision alignment between point cloud features and template features by optimizing transformation parameters. The unique matching evaluation and optimization strategy enables the registration process to maintain high robustness in complex backgrounds.

[0026] Third, in the preferred implementation, the present invention converts point cloud data into a dense three-dimensional model by voxel gridding. This method not only simplifies the data complexity, but also retains the necessary geometric information. Combined with the centroid calculation method, it realizes high-precision calculation of the nozzle position and provides reliable position information for the control system.

[0027] Fourth, in the preferred implementation, the present invention uses a wavelet denoising method for point cloud preprocessing, and separates noise and effective features by decomposing signal components of different frequencies. This method removes noise while retaining point cloud details, laying a solid foundation for efficient processing in subsequent steps. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of the visual positioning method of the ladle sliding nozzle mechanism of the present invention;

[0029] Figure 2 It is a point cloud diagram of a ladle sliding nozzle mechanism of a visual positioning method of a ladle sliding nozzle mechanism of the present invention;

[0030] Figure 3 It is a point cloud diagram of the water outlet after extraction by the visual positioning method of the sliding water outlet mechanism of the ladle of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the technical solution of the present application, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0032] The terms such as up, down, left, right, front and back in this application document are based on the positional relationships shown in the drawings. If the drawings are different, the corresponding positional relationships may also change accordingly, so they cannot be understood as limiting the scope of protection.

[0033] In this application, the terms "install", "connected", "connected", "connected", "fixed" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, an integral connection, a mechanical connection, an electrical connection, or mutual communication, a direct connection, or an indirect connection through an intermediate medium, the internal connection of two components, or the interaction relationship between two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0034] As the instruction manual Figure 1-3 The present invention records a visual positioning method for a ladle sliding nozzle mechanism, which is used for real-time online positioning of the ladle sliding nozzle mechanism in a real-time complex environment. The method does not require the addition of additional markers, avoids secondary modification of the existing ladle sliding nozzle mechanism, and obtains the precise position of the slide plate mechanism and the lower nozzle by performing real-time visual positioning of the ladle sliding nozzle mechanism as a whole, thereby providing a prerequisite for the subsequent automatic installation of the long nozzle.

[0035] The visual positioning method includes:

[0036] Step 1: Obtain the 3D point cloud data of the bottom of the ladle sliding gate mechanism.

[0037] Specifically, the bottom of the ladle sliding nozzle mechanism is photographed using a visual acquisition device to obtain the three-dimensional point cloud data of the bottom nozzle of the ladle sliding nozzle mechanism. The visual acquisition device can be installed on a robotic arm, and the visual acquisition device uses a laser rapid scanning device, which includes a laser light source, an optical scanning system, a receiver, and a data processing and analysis system.

[0038] Laser light sources usually use lasers as light sources to produce high-intensity, monochromatic, and well-coherent laser beams. Laser light sources of different wavelengths can be used for different applications, such as infrared lasers, visible light lasers, and ultraviolet lasers.

[0039] The optical scanning system includes one or more lenses and scanning mirrors to control the scanning path and scanning speed of the laser beam on the target object or scene. The optical scanning system can realize different scanning modes, such as point scanning, line scanning and surface scanning.

[0040] The receiver includes a photodetector and a corresponding signal processing circuit, which is used to receive the optical signal reflected by the laser beam from the target object or scene, and convert the optical signal into a digital signal or an analog signal for subsequent processing and analysis.

[0041] The data processing and analysis system obtains the three-dimensional shape, color, texture and other information of the target object or scene, and processes and analyzes the received light signals; common data processing and analysis methods include point cloud processing, triangular mesh generation, texture mapping, registration, etc.

[0042] Step 2: Preprocess the acquired point cloud data of the bottom of the ladle sliding gate mechanism.

[0043] In step 2, the point cloud data preprocessing is performed using the wavelet denoising method, which specifically includes:

[0044] Step 2.1: Perform wavelet transform on the point cloud data at the bottom of the ladle sliding gate mechanism, decompose the point cloud data into signal components at various frequencies, and reflect the characteristics of the point cloud data at different spatial frequencies based on the signal components.

[0045] Signal components include low-frequency components and high-frequency components. The low-frequency components contain the main morphology and global structural information of the point cloud data, while the high-frequency components contain the details, local changes, and noise information of the point cloud data. Wavelet transform can be used to observe the features of point cloud data at different scales, which helps to identify the global morphology and local details in the data. By decomposing the point cloud data, noise and useful features can be separated into different frequency components, providing a basis for subsequent noise removal.

[0046] Step 2.2: Apply threshold processing to each decomposed frequency component and determine the appropriate threshold for noise removal.

[0047] Specifically, threshold processing is applied to the high-frequency components (noise and details) obtained by decomposition to determine the appropriate threshold. The threshold can be set through experience, statistical methods or adaptive algorithms. The signal amplitude in the high-frequency components is screened, and the components below the threshold (mainly noise) are set to zero or reduced to remove the noise, while the components above the threshold (important details) are retained. Through threshold processing, the noise in the high-frequency components is effectively removed, and the random errors and irregular fluctuations in the point cloud data are reduced.

[0048] Step 2.3: Recombine the processed frequency components to form denoised point cloud data.

[0049] Specifically, the low-frequency and high-frequency components after threshold processing are recombined. During the recombining process, the original spatial characteristics of the point cloud data need to be accurately restored so that the denoised point cloud maintains the geometric structure and positional relationship of the original data, and finally generates the denoised point cloud data. This process synthesizes each frequency component into a complete data set through inverse wavelet transform.

[0050] Step 3: Extract features from the preprocessed point cloud data to obtain the features of the bottom of the ladle sliding nozzle mechanism, align the extracted features with the pre-stored template features, and determine the transformation relationship between the extracted features and the template features.

[0051] Step 3 specifically includes:

[0052] Step 3.1: Using the normal vector estimation method, the geometric features of the point cloud are detected and the geometric structure of the point cloud is extracted as a set of feature points to obtain the geometric features of the bottom of the ladle sliding gate mechanism.

[0053] Geometric features of point clouds such as corners, edges, planes, etc.

[0054] Step 3.2: Perform a preliminary transformation on the extracted point cloud features so that the point cloud features are roughly aligned with the pre-stored template features.

[0055] Rough alignment is to perform preliminary translation, rotation and other transformations on the point cloud features based on relative position, direction, scaling and other operations, so that the point cloud features are initially aligned to the reference coordinate system of the template features.

[0056] Step 3.3: For each transformed point cloud feature point, use a nearest neighbor search algorithm (such as KD-Tree) to find the nearest point in the template feature set to form a matching point pair, calculate the Euclidean distance between all matching point pairs, and use the average of these distances as the measure of the current matching degree. Compare the current matching degree (average Euclidean distance) with the preset threshold; if the matching degree is less than or equal to the threshold, it is considered that the registration has met the predetermined standard. If the matching degree does not meet the predetermined conditions, adjust the transformation parameters (for example, use the iterative closest point algorithm ICP), continue to transform and match the feature points, repeat the above process, and repeatedly optimize the transformation parameters until the matching degree reaches the predetermined threshold condition.

[0057] The Euclidean distance formula is:

[0058]

[0059] Where: x1, y1, z1 represent the coordinates of the feature points of the point cloud; x2, y2, z2 represent the coordinates of the feature points of the template.

[0060] Through the precise matching and transformation optimization in step 3, the point cloud features and the template features are accurately aligned, so as to obtain the best transformation relationship between the two. Through matching evaluation and optimization, it is ensured that the registration result meets the accuracy requirements and improves the reliability and accuracy of the entire positioning process.

[0061] Step 4: Based on the transformation relationship, the three-dimensional point cloud data is transformed as a whole, and the transformed three-dimensional point cloud data is used for three-dimensional construction. The exact position of the water inlet at the bottom of the ladle sliding water inlet mechanism is calculated according to the constructed three-dimensional model.

[0062] Step 4 specifically includes:

[0063] Step 4.1: According to the transformation relationship between the point cloud features and the template features determined in step 3, the entire three-dimensional point cloud data is subjected to corresponding translation, rotation, scaling and other related transformation operations so that the transformed point cloud data will be aligned with the reference coordinate system of the ladle sliding gate mechanism.

[0064] Step 4.1: Through overall transformation, the point cloud data is aligned with the structural coordinate system of the ladle sliding gate mechanism to ensure data consistency and accuracy during the 3D construction process. The transformed point cloud data accurately reflects the geometric shape of the bottom of the ladle sliding gate mechanism, laying the foundation for model construction and subsequent analysis.

[0065] Step 4.2: Use the grid generation method to divide the transformed 3D point cloud data into a uniform 3D voxel grid, count the number of points contained in each voxel, and convert the point cloud data into a dense voxel grid representation according to the set voxel resolution. According to the statistical results, all valid voxels are combined to form a 3D model of the bottom of the ladle sliding gate mechanism.

[0066] Specifically, each voxel represents a small cubic unit in space, and its size is determined by the set voxel resolution. For each voxel, if the number of points it contains exceeds a certain threshold, it is considered a valid voxel and is used to construct a 3D model.

[0067] Step 4.2 uses voxel meshing to simplify point cloud data, reduce data complexity, and retain necessary geometric information. The dense voxel mesh representation helps to clearly show the three-dimensional shape of the bottom of the ladle sliding nozzle mechanism, providing an accurate spatial reference for subsequent nozzle position calculation.

[0068] Step 4.3: Using the transformed feature point coordinates, as well as the structural features and positional relationship of the ladle sliding nozzle mechanism, the centroid calculation method is used to treat the transformed feature point set as a point cloud cluster, and the centroid of these feature points is calculated as the position coordinates of the nozzle.

[0069] The centroid calculation formula of the feature point is as follows:

[0070]

[0071] Where: (x c ,y c ,z c ) represents the centroid of the feature point; x i ,y i 、z i represents the coordinates of each feature point; n represents the total number of feature points; Indicates the summation of all feature points, where i increases from 1 to n; i represents the number of the feature point currently being calculated.

[0072] Step 5: Send the calculated bottom water outlet position coordinates of the ladle sliding water outlet mechanism to the mechanism control system.

[0073] Specifically, step 5 formats the calculated water nozzle position coordinates to meet the input requirements of the mechanism control system (such as JSON, XML, or a specific control protocol format), and sends the formatted coordinates to the control system of the ladle sliding water nozzle mechanism through a communication interface (such as a serial port, a network interface, etc.). After receiving the coordinates, the control system will perform further operations based on these coordinates, such as adjusting the mechanism position, performing operational control, etc.

[0074] The present invention realizes high-precision positioning of the water outlet by using computer vision technology for feature extraction and matching, combined with dimension construction and position calculation. Compared with traditional mechanical sensors or encoders, the present invention is not affected by factors such as environmental interference and equipment wear, and has higher stability. In complex environments, the present invention can accurately locate the ladle sliding water outlet mechanism, thereby improving the positioning accuracy and product quality in the production process.

[0075] The above is only an embodiment of the present invention, and the common sense such as the known specific structure and characteristics in the scheme is not described in detail here. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present invention is defined by the attached claims rather than the above description, so it is intended to include all changes within the meaning and scope of the equivalent elements of the claims. Any figure mark in the claims should not be regarded as limiting the claims involved.

Claims

1. A visual positioning method for a ladle sliding nozzle mechanism, characterized in that: The method comprises: Step 1: Obtain the 3D point cloud data of the bottom of the ladle sliding gate mechanism; Step 2: pre-processing the acquired point cloud data of the bottom of the ladle sliding gate mechanism; Step 3: Extract features from the preprocessed point cloud data to obtain the features of the bottom of the ladle sliding nozzle mechanism, align the extracted features with the pre-stored template features, and determine the transformation relationship between the extracted features and the template features; Step 4: transform the three-dimensional point cloud data as a whole based on the transformation relationship, use the transformed three-dimensional point cloud data to perform three-dimensional construction, and calculate the exact position of the water inlet at the bottom of the ladle sliding water inlet mechanism according to the constructed three-dimensional model; Step 5: Send the calculated bottom water outlet position coordinates of the ladle sliding water outlet mechanism to the mechanism control system.

2. The visual positioning method of the ladle sliding gate mechanism according to claim 1, characterized in that: The process of aligning the extracted features with the pre-stored template features in step 3 includes: transforming the point cloud features, matching the transformed point cloud features and the template features after each transformation, and calculating the matching degree between the two until the matching degree between the two meets a predetermined condition.

3. The visual positioning method of the ladle sliding gate mechanism according to claim 1, characterized in that: The three-dimensional construction process in step 4 includes: using a grid generation method to divide the transformed three-dimensional point cloud data into a uniform three-dimensional voxel grid, and counting the number of points contained in each voxel, and according to the set voxel resolution, converting the point cloud data into a dense voxel grid representation to form a three-dimensional model of the bottom of the ladle sliding nozzle mechanism.

4. The visual positioning method of the ladle sliding gate mechanism according to claim 1, characterized in that: In step 4, the process of overall transformation of the three-dimensional point cloud data based on the transformation relationship includes: for each point in the three-dimensional point cloud, determining its position coordinates in the point cloud coordinate system, and projecting the point into the local coordinate system of the ladle sliding gate mechanism according to the obtained transformation relationship.

5. The visual positioning method of the ladle sliding gate mechanism according to claim 2, characterized in that: The step 3 uses the nearest neighbor matching method to match the transformed point cloud features and template features. The matching process includes: for each transformed point cloud feature point, using the nearest neighbor search algorithm to find the nearest neighbor point of each point cloud feature point in the template feature, taking each pair of matching points as a group of matching point pairs, calculating the Euclidean distance between all matching point pairs, and taking the average value as the current matching degree measure and comparing it with the preset threshold to determine whether the predetermined conditions are met, repeating the transformation, nearest neighbor matching and matching degree calculation until the matching degree reaches the predetermined conditions.

6. The visual positioning method of the ladle sliding nozzle mechanism according to claim 5, characterized in that: The process of judging whether the predetermined condition is met includes: if the obtained average value is ≤ the set threshold value, the matching degree meets the predetermined condition; if the obtained average value is greater than the set threshold value, the matching degree does not meet the predetermined condition.

7. The visual positioning method of the ladle sliding gate mechanism according to claim 1, characterized in that: In step 2, the point cloud data is preprocessed using a wavelet denoising method.

8. The visual positioning method of the ladle sliding gate mechanism according to claim 7, characterized in that: The step 2 specifically includes: Step 2.1: Perform wavelet transform on the point cloud data at the bottom of the ladle sliding gate mechanism, decompose the point cloud data to obtain signal components at various frequencies, and reflect the characteristics of the point cloud data at different spatial frequencies based on the signal components; Step 2.2: Apply threshold processing to each decomposed frequency component and determine the appropriate threshold for noise removal; Step 2.3: Recombine the processed frequency components to form denoised point cloud data.

Images