Ladle mechanical arm intelligent temperature measuring and positioning system based on visual detection and identification

Abstract

The invention provides a hot metal ladle mechanical arm intelligent temperature measurement positioning system based on visual detection identification, and relates to the technical field of visual detection, and the system comprises a feature point identification module which is used for determining a temperature measurement point candidate region of a hot metal ladle; the temperature acquisition module is used for acquiring a temperature measurement data column; the temperature space binding module is used for constructing a ladle temperature distribution diagram; and the temperature measurement positioning module is used for obtaining a temperature positioning result. The technical problems of high potential safety hazard of high-temperature operation, inaccurate temperature measurement point positioning and difficulty in real-time acquisition of temperature distribution caused by dependence on manual positioning and single-point temperature measurement in a traditional ladle temperature measurement process can be solved, temperature measurement and positioning are automatically performed through a visual detection technology, manual intervention is avoided, and the accuracy of temperature measurement is improved. And the temperature measurement precision and the operation safety are improved.

Description

Technical Field

[0001] This application relates to the field of visual inspection technology, and in particular to an intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual inspection and recognition. Background Technology

[0002] In the foundry industry, molten iron ladles, as crucial material carriers, are often used to hold and transport molten iron at high temperatures to designated pouring locations. Due to the high temperature, hazardous conditions, and complexity of the working environment, traditional manual operation not only poses significant safety risks but also, in high-temperature environments, the accuracy and real-time performance of manual temperature measurements are often limited by the thermometer's adaptability and the operator's experience. These traditional methods also rely heavily on manual intervention, leading to low production efficiency. With the development of industrial automation and intelligent inspection technologies, attempts have been made to introduce robotic arms to replace manual operation. However, existing robotic arm temperature measurement systems largely rely on preset trajectories and lack visual inspection guidance. This results in insufficient accuracy in temperature measurement point positioning and low matching between the robotic arm's movement trajectory and the actual state of the molten iron ladle in complex, dynamic high-temperature environments. For example, during transportation or pouring, the molten iron ladle may experience changes in liquid level due to shaking or tilting. If traditional robotic arms cannot perceive key features such as the ladle opening edge and liquid level in real time, there is a risk that the temperature probe may not contact the target area or may collide with the ladle. Therefore, there is an urgent need for an intelligent temperature measurement and positioning system that integrates visual inspection, precise control of robotic arms, and multimodal data collaboration to solve the temperature measurement bottleneck in high-temperature and high-risk scenarios. Summary of the Invention

[0003] The purpose of this application is to provide an intelligent temperature measurement and positioning system for a robotic arm of a molten iron ladle based on visual detection and recognition, in order to solve the technical problems of high safety hazards in high-temperature operations, inaccurate temperature measurement point positioning, and difficulty in obtaining real-time temperature distribution caused by reliance on manual positioning and single-point temperature measurement in the traditional molten iron ladle temperature measurement process.

[0004] In view of the above problems, this application provides an intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition.

[0005] This application provides an intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition. The system includes: a feature point recognition module, used to acquire images of the molten iron ladle surface in real time through visual detection, obtain an image sequence of the molten iron ladle for feature point recognition, and determine candidate areas for temperature measurement points of the molten iron ladle; a temperature acquisition module, used to retrieve the pose data of the molten iron ladle robotic arm and perform spatial matching and identification with the candidate areas for temperature measurement points, extract the movement trajectory of the robotic arm, control the temperature probe of the robotic arm to move along the movement trajectory of the robotic arm to the target temperature measurement point to perform molten iron ladle temperature acquisition, and obtain a series of temperature measurement data; a temperature spatial binding module, used to separate spatial points based on the movement trajectory of the robotic arm, determine multiple spatial coordinates, bind the series of temperature measurement data with the multiple spatial coordinates, and construct a temperature distribution map of the molten iron ladle; and a temperature measurement and positioning module, used to perform bidirectional visual correlation verification according to the temperature distribution map of the molten iron ladle and the image sequence of the molten iron ladle, and perform intelligent temperature measurement and positioning of the molten iron ladle robotic arm according to the bidirectional verification result to obtain the temperature positioning result.

[0006] The technical solution provided in this application has at least the following technical effects or advantages: The aforementioned intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition acquires images of the ladle surface in real time using visual detection technology. This generates image sequences and identifies feature points to determine candidate regions for temperature measurement. Subsequently, the system acquires the robotic arm's pose data, spatially matches it with the candidate temperature measurement regions, extracts the robotic arm's motion trajectory, and controls the robotic arm's temperature probe to precisely move along these trajectories to the target temperature measurement point for temperature acquisition, obtaining a series of temperature data. Based on this, further spatial point separation is used to bind the robotic arm's motion trajectory with multiple spatial coordinates, constructing a temperature distribution map of the molten iron ladle. Finally, the temperature distribution map is visually correlated and bidirectionally verified with the molten iron ladle image data. Based on the verification results, the robotic arm is intelligently positioned to obtain the final temperature positioning result. The entire process achieves accurate measurement and positioning of the molten iron ladle temperature, improving the automation and safety of the operation.

[0007] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of the intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition, as described in this application. Figure 2 This is a flowchart illustrating the feature point recognition module of the intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition, as described in this application.

[0010] Explanation of reference numerals in the attached diagram: Feature point recognition module 11, temperature acquisition module 12, temperature space binding module 13, temperature measurement and positioning module 14. Detailed Implementation

[0011] This application provides an intelligent temperature measurement and positioning system for a robotic arm of a molten iron ladle based on visual detection and recognition. This system solves the technical problems of high safety hazards in high-temperature operations, inaccurate temperature measurement point positioning, and difficulty in obtaining real-time temperature distribution caused by the reliance on manual positioning and single-point temperature measurement in traditional molten iron ladle temperature measurement processes. It achieves the technical effect of automatically measuring and locating temperature through visual detection technology, avoiding manual intervention, and improving the accuracy of temperature measurement and operational safety.

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0013] For examples, please refer to the appendix. Figure 1 This application provides an intelligent temperature measurement and positioning system for a robotic arm of a molten iron ladle based on visual detection and recognition. The system specifically includes the following modules: The feature point recognition module is used to acquire images of the surface of the molten iron ladle in real time through visual inspection, obtain the image sequence of the molten iron ladle, identify feature points, and determine the candidate areas for temperature measurement points of the molten iron ladle.

[0014] Specifically, the feature point recognition module uses high-resolution visual inspection equipment to acquire image data of the ladle surface in real time, including visible light and near-infrared images. Subsequently, by performing binarization, edge detection, and resampling on these image sequences, surface features of the ladle (such as edges, temperature fields, and curved surfaces) are identified. Then, by fusing these features, candidate regions for potential temperature measurement points on the ladle surface are extracted as target areas for subsequent temperature measurements, thereby ensuring the accuracy of temperature acquisition on the ladle surface.

[0015] Furthermore, such as Figure 2 As shown, the feature point recognition module includes: Simultaneous acquisition of molten iron ladle images using a dual-spectrum high-temperature camera yields a sequence of images, including visible light and near-infrared images. Adaptive segmentation is performed on the visible light images to obtain multiple binary images. Edge detection of the ladle opening is performed by traversing these binary images to extract an initial edge contour point set. Temperature feature analysis is conducted on the near-infrared images to extract temperature field distribution data. The near-infrared images are resampled to the visible light image resolution to construct a pixel mapping network. The temperature field distribution data is then fused with the initial edge contour point set according to the pixel mapping network to determine candidate regions for temperature measurement points on the molten iron ladle.

[0016] In a preferred embodiment, in the feature point recognition module, image data of molten iron ladles are first acquired simultaneously from the visible light and near-infrared bands using a dual-spectrum high-temperature camera, resulting in a series of molten iron ladle image sequences. This sequence includes visible light images and near-infrared images. The visible light images provide information about the appearance of the molten iron ladle surface, while the near-infrared images provide temperature-related heat distribution data. Subsequently, the grayscale value of each pixel in the visible light image is calculated using a grayscale conversion formula to obtain the grayscale value of each pixel. The visible light image is then divided into several sub-blocks (e.g., 16×16 pixels), and a threshold is calculated independently for each sub-block. For each sub-block, the Otsu algorithm (maximum inter-class variance method) is used to determine a threshold by maximizing the inter-class variance between the foreground and background. Then, based on the local threshold of each sub-block, pixels with gray values ​​greater than the threshold are classified as foreground, and pixels with gray values ​​less than or equal to the threshold are classified as background. This transforms each sub-block of the visible light image into a binary image. White areas in the binary image typically represent the surface or edges of the molten iron ladle, while black areas represent the background or other regions. Next, an edge detection algorithm (such as Canny edge detection) is used to process each binary image. The Canny algorithm calculates the image gradient, identifies areas with drastic gray-level changes, and thus locates edges in the image. Connectivity analysis is then performed on the detected edges to ensure that the edge detection results are continuous and uninterrupted, avoiding breaks or noise interference. After connectivity analysis, morphological dilation or erosion operations are used to further enhance edge visibility, remove small noise points, and highlight the ladle opening edge. Contour extraction algorithms (such as FindContours) are then used to extract the ladle opening contour lines from the edge image. These contour lines represent the edges of the ladle. By extracting points from these contour lines, an initial edge contour point set is formed. These initial edge contour points serve as key features for subsequent analysis, providing geometric information about the ladle. Next, the grayscale values ​​in the near-infrared image are converted to radiance using a radiometric calibration formula. The radiance is then input into Planck's radiation law to calculate the temperature value of each pixel, thus forming temperature field distribution data. This temperature field distribution data reflects the temperature changes in different areas of the ladle surface, helping to determine the temperature regions that need to be measured. Furthermore, in order to effectively fuse temperature information with physical features in the visible light image, the near-infrared image is resampled to the same resolution as the visible light image. In this process, the size information of the near-infrared image and the visible light image is first obtained, and coordinate mapping is performed based on these size differences. Then, the pixel values ​​of the near-infrared image are transformed into the coordinate system of the visible light image using bilinear interpolation. Bilinear interpolation calculates the interpolation in the neighborhood of the target pixel and generates a new pixel value by weighting the four nearest pixel values ​​in the near-infrared image.This process effectively aligns temperature information from near-infrared images with physical features from visible light images, ensuring spatial consistency between the two images. In this way, a pixel mapping network can be constructed, allowing each target image pixel to accurately obtain its corresponding temperature value from the near-infrared image while maintaining consistent image resolution, ensuring smooth subsequent image-temperature data fusion. Finally, based on the constructed pixel mapping network, the temperature field distribution data from the near-infrared image is fused with the initial edge contour point set extracted from the visible light image, accurately identifying candidate temperature measurement points on the ladle surface. These areas provide target points for subsequent temperature acquisition and measurement, ensuring that temperature measurements cover key areas of the ladle surface.

[0017] Furthermore, the feature point recognition module includes: Based on the temperature field distribution data, isothermal corner points are extracted as the first feature point set; based on the visible light image, edge corner points of the ladle opening are extracted as the second feature point set; the first feature point set and the second feature point set are registered and fused according to the pixel mapping relationship network to generate a thermal-visual fusion map; effective temperature measurement is performed based on the temperature field distribution data to set an effective temperature measurement range; the inner side of the molten iron ladle is uniformly marked based on the initial edge contour point set to set a reference ring; regional overlap analysis is performed on the thermal-visual fusion map according to the reference ring and the effective temperature measurement range to determine the candidate region for temperature measurement points of the molten iron ladle.

[0018] In one optional implementation, in the feature point recognition module, isothermal intervals (such as 5℃ or 10℃) are first set on the temperature field distribution data to generate isotherm maps. For example, if the temperature range is 1200-1400℃, the isotherms are closed or open curves at 1200℃, 1210℃, 1220℃...1400℃. Then, a corner detection algorithm (such as Harris corner detection or SUSAN corner detection) is applied to the isotherm map to identify the intersection points or curvature change points of the isotherms. Taking Harris corner detection as an example, the Harris algorithm calculates the gradient covariance matrix of the temperature value in the horizontal and vertical directions, judges the corner response value, and takes the points with response values ​​greater than the threshold as isothermal corners to form the first feature point set. These isothermal corners refer to key points in the temperature field where the temperature gradient changes significantly or where isotherms intersect, reflecting the local extrema or boundaries of the temperature field, and are potential candidates for temperature measurement points. Subsequently, the initial edge contour point set extracted from the visible light image was fitted with the bag opening contour using the least squares method to obtain a smoother circular or elliptical contour. On the fitted contour, the Shi-Tomasi corner detection algorithm (based on the second derivative of image grayscale) was used to identify the corner points on the contour. The Shi-Tomasi algorithm calculates the corner response function of pixels and retains points with response values ​​greater than a threshold as bag opening edge corner points, forming the second feature point set. These bag opening edge corner points are geometric feature points of the bag opening contour in the visible light image, usually located at geometric abrupt changes in the bag opening contour (such as the connection between the bag opening and the bag body), providing a geometric reference for subsequent fusion with the temperature field. Then, based on the constructed pixel mapping relationship, the coordinates of the first feature point set were transformed into the coordinate system of the visible light image. In the visible light image coordinate system, the transformed first feature point set and the second feature point set were superimposed to generate a thermal-visual fusion map. In this thermal-visual fusion map, each feature point simultaneously contains temperature information and geometric information, providing multi-dimensional data support for subsequent region analysis. Then, based on the temperature field distribution data, the temperature of the molten iron ladle surface is effectively measured. This involves calculating the mean and standard deviation of the temperature field distribution data, and then adding or subtracting multiple standard deviations (usually two or three) from the mean to obtain an effective temperature measurement interval. Simultaneously, based on the ladle opening edge contour point set, the center coordinates and radius of the ladle opening are calculated, and then offset inward by a preset distance to form a ring-shaped area as a reference ring. Points are also marked on the reference ring at equal intervals (e.g., one point every 10°). These points form the reference ring point set, representing the theoretical position below the liquid surface and the target area for temperature measurement. Finally, the shortest distance from each feature point in the thermal-visual fusion image to the reference ring is calculated. If the distance is less than a threshold (e.g., 5 pixels, corresponding to a physical distance of approximately 2 cm), the point is considered to be near the reference ring. These points are then judged in conjunction with the effective temperature measurement interval, and only feature points with temperature values ​​within that interval are retained.By summarizing feature points that simultaneously meet the criteria of being located near the reference ring and having valid temperature, a candidate region for temperature measurement points of the molten iron ladle is formed. These temperature measurement points will serve as the target region for molten iron ladle temperature acquisition to ensure the accuracy and representativeness of the temperature data.

[0019] The temperature acquisition module is used to retrieve the pose data of the molten iron ladle robotic arm and spatially match and identify the candidate area of ​​the temperature measurement point, extract the movement trajectory of the robotic arm, control the temperature measurement probe of the robotic arm to move along the movement trajectory of the robotic arm to the target temperature measurement point to perform molten iron ladle temperature acquisition, and obtain a series of temperature measurement data.

[0020] Specifically, the main function of the temperature acquisition module is to precisely move the temperature probe of the robotic arm to the designated temperature measurement point by coordinating with the pose data of the molten iron ladle robotic arm, thereby performing the temperature acquisition task. In the temperature acquisition module, firstly, based on the determined candidate temperature measurement point area, the system retrieves the pose data of the robotic arm (including the robotic arm's posture, position, and motion state) and performs collision detection between the robotic arm's current position and the candidate temperature measurement point area to ensure that the robotic arm can accurately move to the target temperature measurement point. Through this spatial matching, a robotic arm motion trajectory is generated to ensure that the temperature probe can move smoothly and accurately to the target temperature measurement point. When the robotic arm's temperature probe reaches the target temperature measurement point, the system controls the temperature probe to collect the temperature of the molten iron ladle. This process, through precise control of the robotic arm's movements, ensures that the temperature probe can closely contact the target area and accurately collect temperature data. This measurement data is recorded and stored, forming a temperature data series, providing a basis for subsequent temperature analysis. Through the temperature acquisition module, the system can achieve high-precision temperature data acquisition, avoiding errors in manual operation and ensuring that the temperature measurement process of the molten iron ladle in a high-temperature environment is both safe and efficient.

[0021] Furthermore, the temperature acquisition module includes: A three-dimensional analysis is performed on the molten iron ladle robotic arm to construct the robotic arm's base coordinate system. The candidate regions for temperature measurement points are transformed to the robotic arm's base coordinate system according to a calibration matrix to obtain the coordinates of the temperature measurement points. Collision detection is performed between the pose data of the molten iron ladle robotic arm and the coordinates of the temperature measurement points. The separation distance between the pose data and the coordinates of the temperature measurement points is determined to be less than a preset separation threshold. When the separation distance is less than the preset separation threshold, trajectory replanning is triggered to search for a path and generate a collision-free motion trajectory. Layered detection is performed based on the collision-free motion trajectory to determine the robotic arm's motion trajectory.

[0022] In a preferred embodiment, the temperature acquisition module first constructs a 3D map of the robotic arm's workspace based on 3D point cloud data (obtainable via LiDAR) of the ladle robotic arm and its surrounding environment. This map clarifies the location and boundaries of obstacles (such as the ladle itself and the casting trough). Then, with the center of the ladle robotic arm's mounting base as the origin, three axes of the robotic arm's base coordinate system are defined. Typically, the X-axis represents the robotic arm's forward direction, the Y-axis is perpendicular to the X-axis, and the Z-axis is vertically upward. This base coordinate system is aligned with the 3D map to ensure that the robotic arm's motion control is consistent with the actual environment. Subsequently, a calibration plate is placed in the ladle's working area, and the robotic arm, carrying a camera, moves to multiple different poses to acquire images of the calibration plate. For each image, the corner coordinates of the calibration board are detected and converted into 3D coordinates in the camera coordinate system using camera intrinsic parameters (focal length, principal point coordinates). The pose of the robotic arm's end effector in the base coordinate system is then recorded. By combining the pose of the robotic arm's end effector in the base coordinate system with the pose of the calibration board in the camera coordinate system, the hand-eye calibration matrix is ​​solved using least squares or SVD decomposition. This hand-eye calibration matrix is ​​then used to transform the coordinates of the candidate temperature measurement point region to the robotic arm's base coordinate system, obtaining the precise coordinates of the temperature measurement point in the robotic arm's base coordinate system, making subsequent localization and path planning more accurate. Next, collision detection is performed on the robotic arm. During this process, the pose data of the robotic arm (i.e., the current posture and position of the robotic arm) is compared with the coordinates of the temperature measurement point. Euclidean distance is used to calculate the separation distance between the pose data and the temperature measurement point coordinates. This separation distance is used to determine whether the robotic arm may collide with surrounding objects during temperature acquisition. If the separation distance between the pose data and the coordinates of the temperature measurement point is less than the preset safe separation threshold, it indicates that the robotic arm is at risk of collision and path adjustment is required. At this time, the Rapid Expanding Random Tree (RRT) or A* algorithm will be used, combined with the robotic arm's kinematic constraints (such as joint angle restrictions and speed restrictions) and environmental obstacles (such as molten iron ladle and factory structure), to search for a collision-free path from the current pose to the temperature measurement point in the robotic arm's workspace. During the search process, the searched path will be smoothed (such as B-spline interpolation) to reduce abrupt changes in joint acceleration, thereby generating a motion trajectory that conforms to the robotic arm's dynamic characteristics to adapt to the dynamically changing environment of the molten iron ladle. Then, to improve the accuracy and safety of temperature measurement, a layered detection strategy is adopted to adjust the movement of the robotic arm. The layered detection strategy divides the movement trajectory of the robotic arm into two layers: coarse detection and fine detection. The coarse detection layer involves the robotic arm approaching the temperature measurement point at a higher speed along the replanned trajectory. When it is within a certain range (e.g., 20cm) of the temperature measurement point, it enters the fine detection layer. The fine detection layer involves fine-tuning the trajectory at a lower speed, combined with feedback from the force sensor, to ensure that the temperature measurement probe is in perpendicular contact with the surface of the molten iron ladle (to avoid damage to the probe caused by lateral forces).Through this layered detection, the movement trajectory of the robotic arm is finally determined. This trajectory not only meets the requirements of no collision, but also ensures that the temperature probe accurately contacts the target area, thus ensuring the safety, accuracy, and efficiency of the entire process.

[0023] Furthermore, the temperature acquisition module includes: Multiple path points are set according to the robotic arm's motion trajectory, and the joint angular displacement parameters of the robotic arm are constructed according to the motion sequence of the multiple path points. Displacement boundary condition constraints are set based on the joint angular displacement parameters of the robotic arm. Temperature measurement analysis is performed according to the candidate temperature measurement point area to determine the target temperature measurement point. The temperature measurement probe of the robotic arm is controlled to move along the robotic arm's motion trajectory to the target temperature measurement point according to the displacement boundary condition constraints and to perform contact analysis to obtain the contact force data at the end of the robotic arm. A high-frequency temperature acquisition command is triggered based on the contact force data at the end of the robotic arm, and the temperature of the molten iron ladle is acquired through the high-frequency temperature acquisition command to obtain a series of temperature measurement data.

[0024] In one optional implementation, in the temperature acquisition module, multiple path points are first set according to the planned robotic arm motion trajectory, at equal intervals or key positions (such as turning points or approach points to the target point) according to the motion sequence. These path points represent the intermediate steps of the robotic arm moving from its current position to the target temperature measurement point. Then, based on the robotic arm's Denavit-Hartenberg (DH) parameter model, a mapping relationship from joint angles to end-effector pose is established. For multi-DOF robotic arms (such as 6-DOF robotic arms), inverse kinematics usually has multiple solutions, and a solution that satisfies physical constraints (such as joint angles within allowable ranges) must be selected. If the robotic arm has no closed-loop solution (such as a non-spherical wrist structure), an initial guess value is set during the solution process, and then numerical methods (such as Newton-Raphson iteration) are used to iteratively minimize the end-effector pose error until the error is less than a threshold, thus obtaining the joint angles. If the robotic arm has a closed-loop solution (such as PUMA 560), the joint angles are directly calculated through geometric relationships. For example, the wrist center coordinates are calculated first, and then the joint angles are solved sequentially. By summarizing the solved joint angles and arranging them according to the motion sequence, the joint angular displacement parameters of the robotic arm are constructed. These parameters describe the angular changes of each joint, ensuring that each joint of the robotic arm accurately executes its actions along a predetermined path during movement. Subsequently, to ensure that the robotic arm does not exceed its workspace during operation and avoids collisions with the surrounding environment or unnecessary mechanical damage, displacement boundary conditions are set based on the joint angular displacement parameters. These constraints define the maximum range of motion of the robotic arm, the range of joint motion angles, and other physical limitations, ensuring that the robotic arm always remains within a safe working area and avoiding collisions caused by misoperation. After setting the path points and boundary conditions, in the candidate temperature measurement point area, temperature-stable points (e.g., small temperature gradients) and those located near the reference loop are prioritized as target temperature measurement points. For example, the temperature standard deviation of candidate points is calculated using temperature field distribution data, and the point with the smallest standard deviation is selected. Then, based on the displacement boundary conditions and joint angular displacement parameters, the robotic arm is controlled to move precisely along the motion trajectory to the target temperature measurement point. When approaching the target temperature measurement point (e.g., within 5cm), the torque sensor at the end of the robotic arm is activated to collect contact force data in real time, forming end-effector contact force data. When the end-effector contact force data exceeds the contact force threshold, it indicates that the probe has achieved stable contact with the surface of the molten iron ladle. At this point, a high-frequency temperature acquisition command is triggered, causing the robotic arm to perform high-frequency temperature measurement operations while maintaining stable contact. Through high-frequency temperature acquisition, the system can quickly acquire temperature data from the surface of the molten iron ladle and form a continuous series of temperature measurement data. This data will provide a foundation for subsequent temperature analysis and processing.Through this series of steps, it can be ensured that the robotic arm can move precisely to the target temperature measurement point, make stable contact and efficiently collect temperature data, avoiding the errors and safety hazards of traditional manual measurement, thereby achieving a more accurate and automated temperature acquisition process.

[0025] Furthermore, the temperature acquisition module includes: When the contact force data at the end of the robotic arm exceeds a preset contact force, the robotic arm end contact data is obtained, triggering a pressure sensing signal. The pressure sensing signal controls the robotic arm end contact force data to maintain continuous contact, triggering a high-frequency temperature acquisition command. Based on the high-frequency temperature acquisition command, the robotic arm is controlled to perform multi-cycle acquisition of molten iron ladle temperature data. Temperature stability analysis is performed based on the multi-cycle temperature data to obtain temperature stability parameters. The temperature stability parameters are arranged according to the temperature measurement sequence to obtain the temperature measurement data series.

[0026] In one optional implementation, when the contact force data at the end of the robotic arm is greater than a preset contact force, it indicates that the temperature probe has stably contacted the surface of the molten iron ladle and the contact is good, meeting the temperature measurement requirements. At this time, the contact force data at the end of the robotic arm is recorded to form robotic arm end-of-arm contact data, and a pressure sensing signal is triggered to further ensure that the temperature probe continues to be in contact with the surface of the molten iron ladle, thereby stabilizing the acquisition of temperature data. Once the contact force meets the set standard, a high-frequency temperature acquisition command is triggered. This high-frequency temperature acquisition command requires the temperature probe of the robotic arm to continuously acquire temperature data for multiple cycles under stable contact, forming multi-cycle temperature data. Subsequently, the multi-cycle temperature data is filtered (e.g., by moving average or wavelet denoising) to eliminate high-frequency noise (e.g., temperature jumps at the moment the probe contacts the molten iron ladle), and the standard deviation of the filtered multi-cycle temperature data is calculated. When the standard deviation of N consecutive sampling points is less than or equal to a preset stability threshold, this segment is determined to be a temperature stable segment. At this time, specific temperature data is extracted from this temperature stable segment as temperature stable segment parameters. Finally, the obtained temperature stability parameters are arranged according to the acquisition sequence to generate a complete temperature measurement data series. This data series represents the stable temperature values ​​of the molten iron ladle over multiple time periods, reflecting the overall surface temperature of the ladle. Through these steps, the system not only ensures stable contact between the robotic arm's end-effector and the ladle surface but also ensures the accuracy and stability of data acquisition through high-frequency temperature acquisition commands, thus achieving a highly efficient temperature measurement process. This automated temperature acquisition method can greatly improve the temperature control accuracy in the casting process, reduce human error, and increase production efficiency.

[0027] The temperature space binding module is used to separate spatial points based on the movement trajectory of the robotic arm, determine multiple spatial coordinates, bind the temperature measurement data column with the multiple spatial coordinates, and construct a temperature distribution map of the molten iron ladle.

[0028] Specifically, the main task of the temperature spatial binding module is to accurately correlate the movement trajectory of the robotic arm with the positional data of the molten iron ladle surface, ensuring that each temperature measurement data is paired with its corresponding spatial coordinates. First, the system separates spatial points based on the robotic arm's movement trajectory and determines multiple spatial coordinates, each spatial point representing a specific location on the molten iron ladle surface. Then, the temperature measurement data columns are bound to the corresponding spatial coordinates, ensuring that each temperature value clearly indicates its specific location. After these bindings are completed, a three-dimensional temperature distribution map of the molten iron ladle is constructed by triangulating and labeling these bound data. This map displays the temperature distribution at different locations on the ladle surface, helping to understand the temperature changes across the entire surface area, thus providing crucial information for production adjustments.

[0029] Furthermore, the temperature space binding module includes: The robotic arm's motion trajectory is spatially separated according to the multiple path points to obtain multiple spatial coordinates. Density clustering is performed based on these spatial coordinates, and the temperature measurement data column is segmented according to the clustering results to determine K temperature characterization regions, where K is an integer greater than 1. Each of the K temperature characterization regions contains K temperature measurement data column segments. The K temperature measurement data column segments are spatiotemporally encoded and bound to the multiple spatial coordinates to generate three-dimensional spatial point cloud data, which has temperature attributes. The three-dimensional spatial point cloud data is triangulated based on the temperature attributes to construct a temperature grid for the surface of the molten iron ladle. Temperature gradient column annotations are performed on the temperature grid for the surface of the molten iron ladle to construct a temperature distribution map of the molten iron ladle.

[0030] In a preferred embodiment, in the temperature spatial binding module, firstly, different spatial coordinate points are extracted from the robot arm's motion trajectory based on multiple path points. These path points represent different positions traversed by the robot arm at different time steps. Through these path points, the entire motion trajectory can be separated into multiple spatial point regions. These spatial coordinates will serve as the basis for subsequent analysis, used to accurately describe the spatial distribution of the robot arm's measurement positions. Subsequently, to effectively analyze the distribution of temperature data, density clustering is performed on the path points based on the spatial coordinates. Density clustering is a clustering analysis method used to group path points that are relatively close together into the same group. During the clustering process, a neighborhood radius ϵ (e.g., 5 cm) and a minimum number of points MinPts (e.g., 10) are set, and points with similar and densely distributed temperatures in space are divided into the same region. In this way, the temperature measurement data column can be divided into K temperature characterization regions, where K is an integer greater than 1. These regions will represent different temperature distribution areas on the surface of the molten iron ladle. Next, the temperature measurement data segments from these regions are spatiotemporally encoded and bound to their corresponding spatial coordinates. This involves aligning the timestamps of the temperature measurement data segments with the acquisition timestamps of the spatial coordinates, and then mapping the temperature measurement data segments to their corresponding spatial coordinates. This creates a one-to-one correspondence between the temperature measurement data and the spatial coordinates, generating three-dimensional spatial point cloud data. These data points possess temperature attributes. Then, the obtained three-dimensional spatial point cloud data is triangulated. Triangulation is a common geometric analysis method for spatial data. The system divides the point cloud data in three-dimensional space into multiple triangular units, and then identifies units with abrupt temperature changes from these triangular units, constructing a temperature grid for the surface of the molten iron ladle. Finally, for each triangular facet in the temperature grid on the surface of the molten iron ladle, the temperature difference and spatial distance between it and its adjacent facets are calculated. By dividing the temperature difference by the spatial distance, the temperature gradient is obtained. These temperature gradients are then plotted as temperature gradient bars (similar to a 3D bar chart). The height and color of the bars represent the magnitude of the gradient (e.g., red indicates a high gradient, and blue indicates a low gradient). The plotted temperature gradient bars are then labeled at their corresponding positions on the temperature grid on the surface of the molten iron ladle, constructing a temperature distribution map of the molten iron ladle. This temperature distribution map clearly shows the temperature changes in different areas of the molten iron ladle surface, facilitating temperature control and adjustment according to actual conditions, thereby improving the accuracy of temperature measurement.

[0031] Furthermore, the temperature space binding module includes: Local temperature calculations are performed on the three-dimensional spatial point cloud data according to temperature attributes to obtain local temperature gradient amplitudes. Based on the local temperature gradient amplitudes, size constraint parameters for triangulation are set, and the three-dimensional spatial point cloud data is triangulated according to the size constraint parameters to generate an initial triangular network. Temperature change analysis is performed by traversing the initial triangular network to identify temperature abrupt change regions. The initial triangular network is optimized based on the temperature abrupt change regions to construct the surface temperature grid of the molten iron ladle.

[0032] In one optional implementation, in the temperature spatial binding module, for each spatial point in the 3D spatial point cloud data, multiple neighboring points are first selected to construct a local point set. Then, a linear regression model is set, and the points in the local point set and their corresponding temperature attributes are fitted using the least squares method to determine the coefficients of the linear regression model. The coefficients of the linear regression model are the gradient components. By taking the square root of the sum of the squares of these gradient components, the local temperature gradient amplitude is obtained. Areas with larger temperature gradient amplitudes usually indicate more rapid temperature changes, which may be abrupt regions caused by changes in heat flow or other factors. Subsequently, the triangulation size constraint parameters are obtained by subtracting the normalized value of the local temperature gradient magnitude (maximum-minimum normalization method) from the product of the basic maximum side length and the basic maximum side length. Generally, in regions with drastic temperature changes (i.e., large local temperature gradient magnitudes), smaller triangular elements are used to achieve higher temperature resolution; while in regions with gentler temperature changes, the triangular elements can be appropriately enlarged to reduce computation and improve efficiency. These constraint parameters ensure that the triangulated mesh has high accuracy in important regions (areas with large temperature changes) and low computational complexity in regions with smaller temperature changes. Then, according to the set size constraint parameters, the 3D spatial point cloud data is triangulated, transforming the point cloud data into a triangular mesh. During this process, the system connects each spatial point to its neighboring points, generating an initial triangular mesh. The vertices of each triangular element in the initial triangular mesh correspond to temperature data points in 3D space. These triangular elements provide the foundation for subsequent temperature analysis and optimization. Then, the initial triangular network is traversed. For each traversed triangular cell, the temperature difference and gradient difference between it and its neighboring triangular cells are calculated. When both the temperature difference and gradient difference are greater than the corresponding threshold, the triangular cell is marked as a temperature abrupt change region. After identifying the temperature abrupt change regions, the initial triangular network is optimized. The optimization goal is to adjust the mesh structure to achieve higher resolution in the temperature abrupt change regions, so as to more accurately represent these temperature changes. In this process, the triangular cells in the abrupt change regions are subdivided (e.g., each triangle is split into 4 sub-triangles), and the maximum side length is reduced to the basic maximum side length of the abrupt change, which is usually half of the basic maximum side length. At the same time, the Laplacian smoothing algorithm is used to adjust the coordinates of the triangle vertices in non-abrupt change regions to eliminate mesh distortion caused by local refinement. The optimized initial triangular network will serve as the temperature mesh for the surface of the molten iron ladle, used to represent the spatial characteristics of the temperature distribution on the surface of the molten iron ladle, ensuring that each temperature data point accurately reflects the temperature at its location.

[0033] The temperature measurement and positioning module is used to perform bidirectional visual correlation verification between the temperature distribution map of the molten iron ladle and the image sequence of the molten iron ladle, and to perform intelligent temperature measurement and positioning of the molten iron ladle robotic arm based on the bidirectional verification results to obtain temperature positioning results.

[0034] Specifically, in the temperature measurement and positioning module, the temperature distribution map of the molten iron ladle is first spatially registered by mapping it onto the visible light image in the molten iron ladle image sequence, generating a thermal-visible light visual image. Then, the temperature distribution map is mapped onto the near-infrared image in the molten iron ladle image sequence, generating a thermal-infrared image. These mapping and registration steps ensure spatial alignment between the temperature data and the actual image information, allowing the temperature data to accurately reflect the surface features of the molten iron ladle. Subsequently, temperature-edge correlation analysis is performed based on the thermal-visible light visual image to extract consistent features between temperature and image edges. Simultaneously, temperature-radiation correlation analysis is performed based on the thermal-infrared image to obtain consistent features between temperature and radiation. Afterward, through dual-modal joint analysis of the temperature-edge and temperature-radiation consistent features, the repositioning vector of the robotic arm is accurately calculated, and bidirectional verification is performed based on the repositioning vector to generate a bidirectional verification result. Then, intelligent temperature measurement and positioning is performed based on the bidirectional verification result; that is, the candidate regions for temperature measurement points are sorted according to the confidence level in the bidirectional verification result, and the candidate region with the highest confidence level is selected as the final temperature positioning result. Based on this temperature positioning result, high-precision and high-efficiency temperature measurement can be achieved, ensuring that the temperature measurement of the robotic arm can accurately reflect the temperature state of the molten iron ladle surface during the casting process, thereby improving production quality.

[0035] Furthermore, the temperature measurement and positioning module includes: The temperature distribution map of the molten iron ladle is mapped to the visible light image for spatial registration to generate a thermal-visible light visual image; the temperature distribution map of the molten iron ladle is mapped to the near-infrared image for spatial registration to generate a thermal-infrared image; temperature-edge correlation analysis is performed based on the thermal-visible light visual image to obtain temperature-edge consistency features; temperature-radiation correlation analysis is performed based on the thermal-infrared image to obtain temperature-radiation consistency features; bimodal joint analysis is performed according to the temperature-edge consistency features and the temperature-radiation consistency features, and the robotic arm is repositioned according to the analysis results to obtain the robotic arm repositioning vector; bidirectional verification is performed based on the robotic arm repositioning vector to generate bimodal confidence scores, and the bimodal confidence scores are added to the bidirectional verification results.

[0036] In a preferred embodiment, in the temperature measurement and positioning module, a coordinate transformation chain between the temperature distribution map and the image is first established using a calibration matrix (the transformation from camera to robotic arm base coordinates) and a pixel mapping network (the mapping from visible light / near-infrared image pixels to robotic arm base coordinates). Based on this coordinate transformation chain, the ladle temperature distribution map and the visible light image can be spatially registered, and information at the same location can be mapped to generate a thermal-visible light visual image. Simultaneously, using the same method, the ladle temperature distribution map and the near-infrared image are registered and mapped to generate a thermal-infrared image. Subsequently, for the thermal-visible light visual image, the temperature gradient direction at the edge points is calculated, and its consistency with the visible light edge direction is verified. When the gradient direction deviation is less than or equal to the gradient direction deviation threshold, it indicates that the temperature gradient direction and the visible light edge direction are consistent. At this point, the consistent edge points and temperatures are extracted to form a temperature-edge consistency feature. For thermal-infrared images, a fitting model is constructed using radiation intensity and temperature converted through a radiometric calibration formula. The absolute error of the fit is then calculated, and regions with absolute errors within a threshold are selected to form temperature-radiation consistency features. Next, bimodal joint analysis is performed on the obtained temperature-edge and temperature-radiation consistency features, and the analysis results are compared with stable intervals. If the comparison shows that the position is not within a stable interval, robotic arm repositioning is performed, generating a robotic arm repositioning vector. This vector describes the precise direction and distance of movement of the robotic arm from its current position to the target temperature measurement point. Then, forward and reverse verification are performed on the repositioning vector. In the forward verification stage, the robotic arm pose is adjusted based on the repositioning vector, and visible / near-infrared images and temperature data are reacquired to calculate a new bimodal consistency feature overlap rate. In the reverse verification stage, the original positioning data is retained, and the feature overlap rate after repositioning is simulated. By averaging the two feature overlap rates, a bimodal confidence score is obtained. This bimodal confidence score reflects the reliability of the match between image data and temperature data and is used to evaluate the reliability of the verification process. Finally, bimodal confidence is added to the bidirectional verification results to provide more accurate temperature measurement positioning, ensuring the accuracy and stability of the robotic arm during the temperature measurement process.

[0037] Furthermore, the temperature measurement and positioning module includes: Based on the temperature-edge consistency feature and the temperature-radiation consistency feature, a dual-modal joint analysis is performed to generate dual-modal feature values. When the dual-modal feature values ​​are not in the stable range, a repositioning command is triggered to reposition the robotic arm and generate compensation parameters. Based on the compensation parameters, motion compensation calculations are performed on the robotic arm to generate the robotic arm repositioning vector.

[0038] In one optional implementation, the temperature measurement and positioning module first performs a dual-modal joint analysis based on temperature-edge consistency features and temperature-radiation consistency features. During this process, the proportion of consistent feature points to the total number of edge points is calculated based on the temperature-edge consistency features, serving as the spatial coverage. The mean and standard deviation of the temperature-radiation fitting error are calculated based on the temperature-radiation consistency features. By normalizing the spatial coverage, mean fitting error, and standard deviation, and then weighting the results, a dual-modal feature value is obtained. This dual-modal feature value integrates the consistency between the two modes (temperature-edge and temperature-radiation), representing the degree of matching between the current image data and the temperature data. Subsequently, the generated bimodal feature values ​​are checked to determine if they are within a stable range. A stable range represents a preset threshold range, indicating a stable matching state between image and temperature data. If the bimodal feature values ​​are within this range, it means the robotic arm has accurately positioned itself and successfully reached the target temperature measurement location, and temperature acquisition can continue. Conversely, if the bimodal feature values ​​are not within this range, it indicates a deviation in the matching between image and temperature data. This could be due to the robotic arm failing to reach the target temperature measurement point correctly, or measurement deviation caused by other reasons. In this case, a repositioning command is triggered, requiring the robotic arm to reposition itself according to a new position adjustment strategy. The purpose of the repositioning command is to recalculate the target position of the robotic arm to ensure the temperature probe can accurately reach the new measurement location. During the repositioning process, the spatial coverage is compared with the corresponding threshold. If it is less than this threshold, it indicates a low overlap between the temperature field and the visible light edge, requiring adjustment of the robotic arm position along the edge normal direction. The mean of the fitting error is also compared with the corresponding threshold. If it is less than this threshold, it indicates a discrepancy between the temperature measurement and the radiation intensity, requiring adjustment of the robotic arm height (changing the distance between the temperature probe and the molten iron ladle). By subtracting the spatial coverage from 1 and multiplying the calculated difference by the baseline adjustment, the position compensation parameter (adjustment along the X / Y axes) is obtained. Then, by multiplying the mean of the fitting error by the baseline adjustment, the attitude compensation parameter (pitch angle of the robotic arm's end effector) is obtained. Finally, the compensation parameters are converted into displacement vectors in the base coordinate system using a forward kinematics model. This displacement vector is the robotic arm repositioning vector, used to describe the precise direction and distance of the robotic arm from its current position to the target position. This helps the robotic arm complete the repositioning task, thereby ensuring high accuracy and stability during temperature measurement and improving the reliability and efficiency of automated operations.

[0039] In summary, the intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition provided in this application has the following technical advantages: The system comprises the following modules: a feature point recognition module, which acquires real-time images of the molten iron ladle surface via visual inspection, obtains an image sequence of the molten iron ladle, identifies feature points, and determines candidate regions for temperature measurement points; a temperature acquisition module, which retrieves the pose data of the molten iron ladle robotic arm and performs spatial matching and identification with the candidate regions for temperature measurement points, extracts the robotic arm's motion trajectory, controls the robotic arm's temperature probe to move along the trajectory to the target temperature measurement point, performs temperature acquisition on the molten iron ladle, and obtains a series of temperature measurement data; a temperature spatial binding module, which performs spatial point separation based on the robotic arm's motion trajectory, determines multiple spatial coordinates, binds the series of temperature measurement data with the multiple spatial coordinates, and constructs a temperature distribution map of the molten iron ladle; and a temperature positioning module, which performs bidirectional visual correlation verification between the temperature distribution map of the molten iron ladle and the image sequence of the molten iron ladle, performs intelligent temperature positioning of the molten iron ladle robotic arm based on the bidirectional verification results, and obtains the temperature positioning results. These technologies collectively solve the technical problems of high-temperature operation safety hazards, inaccurate temperature measurement point positioning, and difficulty in real-time acquisition of temperature distribution caused by the reliance on manual positioning and single-point temperature measurement in traditional molten iron ladle temperature measurement processes. They achieve the technical effect of automatically measuring and locating temperature through visual detection technology, avoiding manual intervention, and improving the accuracy of temperature measurement and operational safety.

[0040] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0041] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart temperature measurement and positioning system for a robotic arm of a molten iron ladle based on visual detection and recognition, characterized in that, The system includes: The feature point recognition module is used to acquire images of the surface of the molten iron ladle in real time through visual inspection, obtain the image sequence of the molten iron ladle, identify feature points, and determine the candidate area of ​​the temperature measurement point of the molten iron ladle. The temperature acquisition module is used to retrieve the pose data of the molten iron ladle robotic arm and spatially match and identify the candidate area of ​​the temperature measurement point, extract the movement trajectory of the robotic arm, control the temperature measurement probe of the robotic arm to move along the movement trajectory of the robotic arm to the target temperature measurement point to perform molten iron ladle temperature acquisition, and obtain a series of temperature measurement data. The temperature space binding module is used to separate spatial points based on the movement trajectory of the robotic arm, determine multiple spatial coordinates, bind the temperature measurement data column with the multiple spatial coordinates, and construct a temperature distribution map of the molten iron ladle. The temperature measurement and positioning module is used to perform bidirectional visual correlation verification between the temperature distribution map of the molten iron ladle and the image sequence of the molten iron ladle, and to perform intelligent temperature measurement and positioning of the molten iron ladle robotic arm based on the bidirectional verification results to obtain temperature positioning results.

2. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 1, characterized in that, The feature point recognition module includes: The molten iron ladle is simultaneously acquired using a dual-spectrum high-temperature resistant camera to obtain an image sequence of the molten iron ladle, which includes visible light images and near-infrared images; Adaptive segmentation is performed based on the visible light image to obtain multiple binary images. The edges of the ladle opening are detected by traversing the multiple binary images, and an initial edge contour point set is extracted. Temperature feature analysis is performed on the near-infrared image to extract temperature field distribution data; The near-infrared image is resampled to the visible light image resolution to construct a pixel mapping relationship network; The temperature field distribution data is fused with the initial edge contour point set according to the pixel mapping relationship network to determine the candidate region for temperature measurement points in the molten iron ladle.

3. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 2, characterized in that, The feature point recognition module includes: Based on the temperature field distribution data, isothermal corner points are extracted as the first feature point set; Based on the visible light image, the corner points of the bag opening edge are extracted as the second feature point set; The first feature point set and the second feature point set are registered and fused according to the pixel mapping relationship network to generate a thermal-visual fusion map. Based on the temperature field distribution data, effective temperature measurement is performed, and an effective temperature measurement range is set. Based on the initial edge contour point set, the inner side of the molten iron ladle is uniformly marked, and a reference ring is set; Based on the reference ring and the effective temperature measurement range, the thermal-visual fusion map is analyzed for regional overlap to determine the candidate regions for temperature measurement points in the molten iron ladle.

4. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 1, characterized in that, The temperature acquisition module includes: Three-dimensional analysis was performed on the molten iron ladle robotic arm to construct the robotic arm's base coordinates; The candidate region of temperature measurement points is transformed to the base coordinate system of the robotic arm according to the calibration matrix to obtain the coordinates of the temperature measurement points; Collision detection is performed between the pose data of the ladle robot arm and the coordinates of the temperature measuring point. The coordinates of the temperature measuring point are then traversed to determine whether the separation distance between the pose data and the coordinates of the temperature measuring point is less than a preset separation threshold. When the separation distance between the pose data and the coordinates of the temperature measurement point is less than the preset separation threshold, trajectory replanning is triggered to perform path search and generate a collision-free motion trajectory. The robotic arm's motion trajectory is determined by performing layered detection based on the collision-free motion trajectory.

5. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 1, characterized in that, The temperature acquisition module includes: Multiple path points are set according to the motion trajectory of the robotic arm, and the joint angular displacement parameters of the robotic arm are constructed by the multiple path points according to the motion sequence. Displacement boundary condition constraints are set based on the joint angular displacement parameters of the robotic arm. Temperature measurement analysis is performed based on the candidate temperature measurement point area to determine the target temperature measurement point. The temperature measurement probe of the robotic arm is controlled to move along the movement trajectory of the robotic arm according to the displacement boundary condition constraint to the target temperature measurement point for contact analysis, thereby obtaining the contact force data at the end of the robotic arm. Based on the contact force data at the end of the robotic arm, a high-frequency temperature acquisition command is triggered, and the temperature of the molten iron ladle is acquired through the high-frequency temperature acquisition command to obtain a series of temperature measurement data.

6. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 5, characterized in that, The temperature acquisition module includes: When the contact force data at the end of the robotic arm is greater than the preset contact force, the contact data at the end of the robotic arm is obtained, a pressure sensing signal is triggered, and the contact force data at the end of the robotic arm is controlled to make continuous contact through the pressure sensing signal, triggering a high-frequency temperature acquisition command. Based on the high-frequency temperature acquisition command, the robotic arm is controlled to perform multi-cycle acquisition of molten iron ladle to obtain multi-cycle temperature data. Temperature stability analysis is performed based on the multi-cycle temperature data to obtain parameters for the temperature stability segment. The temperature stability parameters are arranged according to the temperature measurement time sequence to obtain the temperature measurement data column.

7. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 5, characterized in that, The temperature space binding module includes: The movement trajectory of the robotic arm is spatially separated according to the multiple path points to obtain multiple spatial coordinates; Density clustering is performed based on the multiple spatial coordinates. The temperature measurement data column is divided according to the clustering results to determine K temperature characterization regions, where K is an integer greater than 1. The K temperature characterization regions contain K temperature measurement data column segments. The K temperature measurement data segments are spatiotemporally encoded and bound to the multiple spatial coordinates to generate three-dimensional spatial point cloud data, which has temperature attributes. Based on the temperature attribute, the three-dimensional spatial point cloud data is triangulated to construct a temperature grid on the surface of the molten iron ladle. Temperature gradient bars are marked on the surface temperature grid of the molten iron ladle to construct the temperature distribution map of the molten iron ladle.

8. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 7, characterized in that, The temperature space binding module includes: Local temperature calculations are performed on the three-dimensional spatial point cloud data according to temperature attributes to obtain the local temperature gradient amplitude. Based on the local temperature gradient magnitude, the size constraint parameters for triangulation are set, and the three-dimensional spatial point cloud data is triangulated according to the size constraint parameters to generate an initial triangular network. The initial triangular network is traversed to analyze temperature changes and identify regions of abrupt temperature changes. The initial triangular network is optimized based on the temperature abrupt change region to construct the surface temperature grid of the molten iron ladle.

9. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 2, characterized in that, The temperature measurement and positioning module includes: The temperature distribution map of the molten iron ladle is mapped to the visible light image for spatial registration to generate a thermal-visible light visual image. The temperature distribution map of the molten iron ladle is mapped to the near-infrared image for spatial registration to generate a thermal-infrared image; Temperature-edge correlation analysis is performed on the thermal-visible light visual image to obtain temperature-edge consistent features; Temperature-radiation correlation analysis is performed based on the thermal-infrared images to obtain temperature-radiation consistency characteristics. A dual-modal joint analysis is performed based on the temperature-edge consistency feature and the temperature-radiation consistency feature. The robotic arm is then repositioned based on the analysis results to obtain the robotic arm repositioning vector. Based on the robotic arm repositioning vector, perform bidirectional verification, generate bimodal confidence scores, and add the bimodal confidence scores to the bidirectional verification results.

10. The intelligent temperature measurement and positioning system for a molten iron ladle robotic arm based on visual detection and recognition as described in claim 9, characterized in that, The temperature measurement and positioning module includes: Based on the temperature-edge consistency features and the temperature-radiation consistency features, a dual-modal joint analysis is performed to generate dual-modal feature values; When the bimodal feature value is not in the stable range, a repositioning command is triggered, and the robotic arm is repositioned through the repositioning command to generate compensation parameters; The robot arm is subjected to motion compensation calculation based on the compensation parameters to generate the robot arm repositioning vector.

Images