Positioning method and system for oil outlet of converter station transformer

Abstract

The application provides a positioning method and system for a converter station transformer oil outlet, which comprises the following steps: firstly, detecting an artificial visual mark and solving the mark pose, and calculating the initial pose of the oil outlet according to the pre-calibration relationship between the mark and the oil outlet; then, determining a local region of interest according to the initial pose, extracting the oil outlet center, end face normal vector and axis direction from the depth data in the region; finally, taking the initial pose as the initial value of iteration, taking the geometric features of the oil outlet as the constraint, and correcting the pose through a geometric error model to output the final six-degree-of-freedom pose of the oil outlet. The application combines the rapidity of mark recognition and the accuracy of geometric positioning, can effectively inhibit the mark installation error and environmental interference, quickly obtains the accurate position and attitude of the oil outlet in a complex equipment background, and provides reliable pose input for automatic oil taking operation.

Description

Technical Field

[0001] This invention relates to the field of equipment positioning technology, and in particular to a positioning method and system for the oil intake port of a converter station transformer. Background Technology

[0002] Converter transformers are critical equipment in converter stations, and their operating status directly affects the safe and stable operation of the power grid. During equipment operation, it is usually necessary to regularly sample the insulating oil and use methods such as dissolved gas analysis to determine whether there are potential faults such as abnormal discharge, overheating, and insulation aging inside the equipment. Currently, oil sampling operations for converter transformers are mainly carried out manually. Operators need to locate the target equipment according to the equipment number and site markings, and manually complete operations such as confirming the oil sampling valve, connecting the sampling connector, collecting the oil sample, and sealing it.

[0003] With the development of digital converter stations and intelligent operation and maintenance technologies, the use of mobile robots equipped with robotic arms and automated oil sampling devices to replace manual labor in on-site oil sampling is gradually becoming a trend. Robots can autonomously reach the vicinity of the target equipment according to task instructions and use their robotic arms to complete the connection of oil sampling connectors and oil sample collection, helping to reduce the risks of on-site operations for personnel and improve the automation and standardization of the oil sampling process.

[0004] However, in the process of automated oil extraction by robots, relying solely on the autonomous navigation of the mobile robot cannot meet the requirements of the oil extraction operation. Mobile chassis navigation typically only allows the robot to reach the vicinity of the converter transformer, while the oil extraction port is usually located in a small area on the side or bottom of the equipment, surrounded by valves, pipelines, flanges, and other structures. The robotic arm's end effector needs to further obtain the accurate position and orientation of the oil extraction port to complete the precise docking between the oil extraction connector and the port. Therefore, how to achieve high-precision positioning of the robot at the converter transformer's oil extraction port is a key issue in automated oil extraction systems.

[0005] Existing robot localization methods mainly include localization methods based on fixed-position teaching, localization methods based on visual identifiers, and localization methods based on target geometric features, specifically: The fixed-position teaching method involves manually controlling a robotic arm to move to a target position and recording the end-effector's pose. During actual operation, the robot repeatedly performs the oil-collecting action based on pre-saved position parameters. This method is simple to implement and does not require complex vision processing algorithms, but it relies on the robot's installation environment and the stability of the equipment position. When the mobile robot's parking position changes, errors occur in the robotic arm's repetitive movements, or the local structural position changes after maintenance of the converter transformer equipment, the original taught position may not guarantee an accurate correspondence between the oil-collecting connector and the oil-collecting port, leading to a decrease in positioning accuracy.

[0006] Visual tagging-based localization methods involve placing artificial markers with specific coded information near the target device, using cameras to identify marker features, and calculating the target's position based on the spatial relationship between the marker and the target device. This method offers advantages such as clear target identification and fast localization speed, avoiding extensive searches by the robot in complex equipment environments. However, this type of method typically uses artificial markers as the direct measurement object, and the actual position of the target device needs to be indirectly deduced from the marker positions. When there are installation deviations in the artificial markers, changes in their fixing structure, surface contamination, or environmental factors affecting image acquisition, the localization error will be further propagated to the final target position, making it difficult to guarantee precise alignment between the robot's end effector and the oil intake port.

[0007] Localization methods based on target geometric features directly utilize the device's own structural information for positioning. For example, they use the target's contours, edges, planes, circular structures, or spatial point cloud features to calculate the target's center position and orientation. This method requires no additional markers, and the positioning results directly reflect the device's location, exhibiting good physical consistency. However, converter transformers have complex structures, and multiple similar metal components often exist near the oil tap. The target area is easily affected by shading, reflections, oil contamination, and changes in ambient light, making target feature extraction difficult. Furthermore, this method typically requires a relatively accurate initial search range; when the robot is far from the target or the target's location is unknown, the positioning efficiency and stability are limited.

[0008] Furthermore, existing robot vision localization methods typically separate target recognition and end-effector motion control into two independent processes: first, the target position is acquired through visual algorithms, and then the robotic arm's movement is controlled based on the calculation results. When there are deviations in the visual measurement results, the subsequent control process can only execute the movement according to the existing target position, lacking the ability to correct the visual localization results using the target's own geometric constraints. Therefore, in the oil sampling scenario of converter transformers, relying solely on a single localization method is insufficient to simultaneously meet the requirements of rapid target recognition, high-precision localization, and stable docking.

[0009] Therefore, to address the challenges of target location identification, insufficient positioning accuracy, and difficulty in achieving precise docking of the robot end effector during automatic oil extraction from converter transformers, a fusion positioning method that combines target identification capabilities with equipment geometric constraint information is needed to improve the autonomy and positioning reliability of the robot's oil extraction operation. Summary of the Invention

[0010] The purpose of this invention is to provide a positioning method and system for the oil intake port of a converter station transformer, aiming to solve the problems of difficulty in target location identification, insufficient positioning accuracy, and difficulty in achieving precise docking of the robot end effector during the automatic oil intake process of the converter transformer.

[0011] In a first aspect, the present invention provides a method for locating the oil intake port of a converter station transformer, the method comprising: A color image containing the artificial visual sign and the oil extraction port is acquired, and the color image is sequentially converted to grayscale and then binarized to obtain a binary image. The corner points of the artificial visual sign are then extracted from the binary image. The pose matrix of the artificial visual sign is constructed based on the pixel coordinates and three-dimensional coordinates of the corner points of the sign, and the initial pose of the oil extraction port is calculated based on the pose matrix. The initial center position of the oil sampling port is obtained based on the initial pose, and the initial center position is projected onto the color image. The pixel coordinates of the initial center position are obtained based on the projection result, and the local region of interest is determined based on the pixel coordinates of the initial center position. The pixels in the local region of interest are converted into three-dimensional points in the camera coordinate system, and all the three-dimensional points are used to form a local point cloud of the oil intake port. The center of the oil intake port, the end face normal vector, and the axis direction are extracted from the local point cloud. Using the initial pose of the oil intake port as the initial value for fusion localization, the center of the oil intake port, the end face normal vector, and the axial direction are used as geometric constraints to iteratively correct the initial pose of the oil intake port until the preset termination condition is met, thus obtaining the final pose of the oil intake port.

[0012] Secondly, the present invention provides a positioning system for the oil intake port of a converter station transformer, the system comprising: The image processing module is used to acquire a color image containing artificial visual signs and an oil extraction port, and to convert the color image to grayscale and then to binarize it to obtain a binary image, and to extract the corner points of the artificial visual signs from the binary image; The initial pose calculation module is used to construct the pose matrix of the artificial visual sign based on the pixel coordinates and three-dimensional coordinates of the sign corner points, and to calculate the initial pose of the oil extraction port based on the pose matrix. The region recognition module is used to obtain the initial center position of the oil sampling port according to the initial pose, project the initial center position onto the color image, obtain the pixel coordinates of the initial center position according to the projection result, and determine the local region of interest according to the pixel coordinates of the initial center position. The feature extraction module is used to convert the pixels in the local region of interest into three-dimensional points in the camera coordinate system, and to form a local point cloud of the oil intake port from all the three-dimensional points, and to extract the center of the oil intake port, the end face normal vector and the axis direction from the local point cloud; The pose correction module is used to use the initial pose of the oil tap as the initial value for fusion positioning, and uses the center of the oil tap, the end face normal vector, and the axis direction as geometric constraints to iteratively correct the initial pose of the oil tap until the preset termination condition is met, thus obtaining the final pose of the oil tap.

[0013] Thirdly, the present invention provides a storage medium that stores one or more programs, which, when executed by a processor, implement the above-described positioning method for the oil tap of a converter station transformer.

[0014] Fourthly, the present invention provides an electronic device, the electronic device comprising a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the above-mentioned positioning method for the oil intake port of the converter station transformer.

[0015] Compared with the prior art, the present invention has the following advantages: This invention achieves a two-level pose correction, from coarse to fine, by constructing a positioning architecture that integrates artificial visual marker guidance with the geometric constraints of the oil intake port. This ensures that the final positioning result not only possesses the ability to quickly recognize visual markers but also iteratively constrains and corrects the initial pose using the geometric features of the oil intake port end face, center, and axis. This effectively suppresses the impact of factors such as marker installation deviation, image distortion, illumination changes, and edge detection errors on positioning accuracy. Simultaneously, by constructing a local region of interest using the initial pose, the range of geometric feature extraction is significantly reduced from the entire scene to a local area, significantly lowering the computational overhead of point cloud processing and feature fitting, and avoiding misidentification problems caused by similar structures in complex backgrounds. Furthermore, by establishing an iteration termination judgment mechanism, the positioning result can simultaneously meet the quantitative requirements of positional accuracy and attitude accuracy, and possess the identifiability of positioning failures. This provides accurate, reliable, and verifiable six-degree-of-freedom pose information for the robotic arm's automatic oil extraction, effectively improving the success rate and operational stability of autonomous docking between the oil intake connector and the oil intake port. Attached Figure Description

[0016] Figure 1 This is a flowchart of a positioning method for the oil intake port of a converter station transformer according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a positioning system for the oil intake port of a converter station transformer, according to an embodiment of the present invention.

[0017] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0019] like Figure 1 As shown, an embodiment of the present invention proposes a positioning method for the oil intake port of a converter station transformer, the method comprising steps S101 to S105, wherein: Step S101: Obtain a color image containing the artificial visual sign and the oil extraction port, and then perform grayscale conversion and binarization on the color image to obtain a binary image, and extract the sign corner points of the artificial visual sign from the binary image; It should be noted that in this step, the camera coordinate system C, the artificial visual marker coordinate system M, and the oil intake coordinate system O are first established. The origin of the oil intake coordinate system is located at the center of the oil intake. The shaft is aligned with the insertion direction of the oil tap. shaft and The shaft is located inside the oil inlet end face.

[0020] In addition, artificial visual markers are installed near the oil intake port, and the relative positions between the markers and the oil intake port are calibrated before use to obtain a fixed transformation matrix: ; in, This is the homogeneous transformation matrix of the oil inlet coordinate system relative to the manual visual marker coordinate system; The rotational relationship between the artificial visual marker coordinate system and the oil intake coordinate system; It is the translation vector of the oil intake center relative to the center of the artificial visual marker.

[0021] Furthermore, in some embodiments, the color image is converted to a grayscale image, the grayscale image is divided into multiple local image blocks, and a local threshold is calculated for each local image block. The grayscale image is then converted to a binary image based on the local threshold. ; in, The binarization threshold for the k-th local image patch; , These are the maximum and minimum gray values ​​of the k-th local image block, respectively.

[0022] The grayscale image is then converted into a binary image based on a local threshold. Regions where the difference between the maximum and minimum grayscale values ​​is less than a preset contrast threshold are marked as low-contrast regions and are not included in subsequent manual visual marker boundary detection.

[0023] The black and white pixels in the binary image are then divided into several black connected regions and white connected regions, respectively, and all boundary pixels between adjacent black connected regions and white connected regions are obtained. Calculate the geometric center of all boundary pixels between adjacent black and white connected regions, obtain the azimuth angle of each boundary pixel relative to the geometric center, and sort all boundary pixels according to the azimuth angle. A sliding window is set on the sorted boundary pixels. The principal component analysis method is used to fit the boundary pixels in the sliding window with a straight line and the mean square error of the straight line fitting is calculated. The center position of the window where the mean square error of the straight line fitting is a local maximum and the local maximum is greater than the preset corner response threshold is recorded as a candidate corner point. Enumerate four-corner point combinations from all candidate corner points, connect the four candidate corner points of each four-corner point combination in sequence to obtain a quadrilateral, and perform the following judgments in sequence: the total number of boundary pixels is greater than a preset number threshold, the fitting error of all four sides of the quadrilateral is less than a preset side fitting error threshold, the included angle between adjacent sides of the quadrilateral is within a preset angle range, the area of ​​the quadrilateral is greater than a preset minimum area threshold, the aspect ratio of the quadrilateral is less than a preset maximum aspect ratio threshold, and the four candidate corner points satisfy the perspective projection coplanarity constraint. Among all the combinations that pass the judgment, the combination with the smallest sum of the fitting errors of the four sides, the largest quadrilateral area, and the closest included angle between adjacent sides to a right angle is selected as the optimal quadrilateral, and the four candidate corner points of the optimal quadrilateral are used as the marker corner points.

[0024] In summary, by dividing connected regions and grouping boundary points, the outer contour boundaries of artificial visual signs can be accurately extracted, avoiding the problem of incorrect connection of different boundaries when small-sized signs or adjacent structures are closely spaced. Based on this, a combination of azimuth sorting and sliding window PCA line fitting is used to detect candidate corner points. The local maxima of the mean square error of the line fitting are used as the criteria for corner point determination. Combined with error thresholds and minimum spacing constraints, false corner points caused by boundary noise, jagged edges, and local reflections can be effectively suppressed. Furthermore, by enumerating four-corner point combinations and sequentially performing six progressive determinations—boundary point count, edge fitting error, adjacent side angle, quadrilateral area, aspect ratio, and perspective coplanarity—quadrilaterals that conform to the projection characteristics of artificial visual signs can be accurately selected in complex equipment backgrounds, avoiding misidentification of similar-shaped structures such as valves, flanges, and connecting pipelines near the oil intake as signs. Finally, by comprehensively selecting the optimal quadrilateral, high-precision and high-reliability image feature input is provided for subsequent pose calculation, thereby improving the adaptability and stability of the entire positioning method in the complex environment of converter stations.

[0025] Step S102: Construct the pose matrix of the artificial visual sign based on the pixel coordinates and three-dimensional coordinates of the corner points of the sign, and calculate the initial pose of the oil extraction port based on the pose matrix; It should be noted that in this step, the three-dimensional coordinates of the four corner points of the artificial visual sign in the sign coordinate system are known: ; in, Let i be the three-dimensional coordinates of the i-th marker corner point. , , These are the X, Y, and Z coordinates of the i-th marker corner point, respectively.

[0026] Based on the refined image corner points, camera intrinsic parameters, and the actual size of the sign, the PnP method is used to solve for the position and orientation of the artificial visual sign relative to the camera coordinate system: ; in, K is the scale factor; K is the camera intrinsic parameter matrix; This is the rotation matrix of the artificial visual sign relative to the camera coordinate system; Let be the translation vector of the artificial visual sign relative to the camera coordinate system.

[0027] Constructing the pose matrix of artificial visual markers: ; Calculate the initial pose of the oil sampling port based on the pre-calibration relationship between the artificial visual marker and the oil sampling port: ; in, The initial pose of the oil intake port relative to the camera coordinate system.

[0028] This initial pose is used to determine the local search area of ​​the oil intake port and serves as the initial value for subsequent fusion localization iterations; it is not directly used as the final localization result.

[0029] In summary, by pre-defining a fixed transformation relationship between the artificial visual marker and the oil intake port, the real-time marker pose obtained by the PnP algorithm is directly converted into the initial pose of the oil intake port through homogeneous matrix multiplication. This allows the oil intake port position estimation to be dynamically updated following the marker observation results, eliminating the need to re-detect the oil intake port itself during each localization, thus improving the real-time performance and convenience of initial pose acquisition. Simultaneously, this initial pose provides a reasonable initial value for subsequent geometric constraint fusion localization and a local search benchmark, enabling subsequent geometric feature extraction to be performed within a reduced local area, avoiding large-scale blind searches and significantly reducing computational complexity. Furthermore, since this initial pose originates from real-time image observation of the marker, it can always be acquired when the marker is not occluded and detection is successful, improving the robustness of the localization method and providing a reliable guarantee for the convergence speed and final localization accuracy of subsequent fusion localization.

[0030] Step S103: Obtain the initial center position of the oil extraction port according to the initial pose, project the initial center position onto the color image, obtain the pixel coordinates of the initial center position according to the projection result, and determine the local region of interest according to the pixel coordinates of the initial center position. It should be noted that, centered on the pixel coordinates of the initial center position, a local region of interest (ROI) is established based on the actual size of the oil tap, the installation distance between the marker and the oil tap, and the allowable initial positioning error. The ROI should cover the entire oil tap and its surrounding end face, while excluding most adjacent valves, flanges, bolts, and piping structures.

[0031] Step S104: Convert the pixels in the local region of interest into three-dimensional points in the camera coordinate system, and form a local point cloud of the oil intake port from all the three-dimensional points, and extract the center of the oil intake port, the end face normal vector and the axis direction from the local point cloud. It should be noted that in this step, firstly, the image pixels are converted into three-dimensional points in the camera coordinate system based on the depth values ​​in the local region of interest: ; in, These are the coordinates of a 3D point in the camera coordinate system. These are pixel coordinates; This represents the depth value of the corresponding 3D point; For the camera's horizontal and vertical focal lengths, , , where x and y are the coordinates of the camera's principal point, respectively.

[0032] During the fitting process, several 3D points are selected from the local point cloud to establish a candidate plane. The distances from the remaining 3D points to the candidate plane are calculated, and 3D points whose distances are less than the in-plane point threshold are taken as in-plane points.

[0033] Then, the inner points of the oil intake port end face are transformed into the local two-dimensional coordinate system of the end face, and the circular edge of the oil intake port is fitted to obtain a circle fitting model. The center coordinates of the circle are obtained according to the circle fitting model, and the center coordinates of the circle are transformed from the local two-dimensional coordinate system of the end face back to the camera coordinate system to obtain the center coordinates of the oil intake port.

[0034] Furthermore, the expression for the oil intake port axis is: ; in, Center of the oil intake port To obtain the unit direction vector of the oil port axis, For axis parameters, The parametric equation for the oil inlet axis; Construct two three-dimensional feature points on the oil intake axis: ; ; in, To preset the feature point spacing, , All are three-dimensional feature points.

[0035] In summary, by back-projecting depth pixels within a local region of interest into a 3D point cloud, the fitting of the oil intake end face and the extraction of circular edges are confined to a local area, avoiding interference from numerous irrelevant structures (such as valves, flanges, and pipelines) in the entire point cloud. This significantly reduces the difficulty and computational load of point cloud segmentation and feature extraction. Simultaneously, a step-by-step strategy of first fitting the end face plane and then transforming it to a local 2D coordinate system for circle fitting reduces the dimensionality of the circular edge fitting problem in 3D space to a circle fitting problem in a 2D plane, reducing fitting complexity and improving the computational accuracy and stability of the oil intake center coordinates. Furthermore, by determining the axial direction through fitting the end face normal vector and constructing two 3D feature points on the axial direction using the oil intake center, not only is the position information of the oil intake provided, but the complete axial direction is also characterized. This provides accurate geometric constraint observations for subsequent iterative corrections, enabling fusion positioning to improve from "correcting only the position" to "correcting both position and direction simultaneously," meeting the complete six-degree-of-freedom pose requirements of the robotic arm's insertion operation.

[0036] Step S105: Using the initial pose of the oil sampling port as the initial value for fusion positioning, the center of the oil sampling port, the end face normal vector, and the axial direction are used as geometric constraints to iteratively correct the initial pose of the oil sampling port until the preset termination condition is met, and the final pose of the oil sampling port is obtained.

[0037] It should be noted that three mutually orthogonal reference planes are first established, specifically including: First reference plane The coordinate system corresponding to the candidate oil intake port flat; Second reference plane The coordinate system corresponding to the candidate oil intake port flat; Third Reference Plane The coordinate system corresponding to the candidate oil intake port flat.

[0038] Furthermore, the intersection of the first reference plane and the second reference plane is the oil intake port axis, and the third reference plane is used to constrain the position of the oil intake port center in the axial direction.

[0039] Specifically, the j-th reference plane is represented as: ; in, Let j be the reference plane at the k-th iteration. Let be the unit normal vector of the j-th reference plane at the k-th iteration. It is the origin of the oil intake coordinate system during the k-th iteration.

[0040] Then, the geometric error vector for the k-th iteration is constructed according to the following formula. : ; in, Let be the signed distance from the first 3D feature point to the first reference plane. The signed distance from the second three-dimensional feature point to the first reference plane; Let be the signed distance from the first 3D feature point to the second reference plane. Let be the signed distance from the second 3D feature point to the second reference plane. The signed distance from the first three-dimensional feature point to the third reference plane is denoted as .

[0041] in: ; When the geometric error vector approaches zero, the origin of the oil sampling port coordinate system coincides with the actual center of the oil sampling port, and the oil sampling port coordinate system... The shaft coincides with the actual oil intake port axis.

[0042] Next, the parameters to be corrected for the oil port pose are obtained at the k-th iteration. Represented as: ; in, These represent the x, y, and z translation corrections for the oil intake center in the camera coordinate system. , These are the rotation corrections for the oil inlet around the x-axis and y-axis of the camera coordinate system, respectively.

[0043] Since the oil intake port is approximately axisymmetric, its geometric features cannot stably determine its rotation angle around its own axis. Therefore, the rotation angle around the axis maintains the corresponding result in the initial pose of the artificial visual marker. When the oil intake port or docking joint has non-axisymmetric directional features, an additional directional feature can be added to further correct the rotation angle around the axis.

[0044] Then, calculate the Jacobian matrix of the geometric error vector relative to the pose parameters using the following formula: ; Calculate the pose correction for the k-th iteration based on the current geometric error. : ; in, Let be the Jacobian matrix of the k-th iteration. This is the pseudo-inverse of the Jacobian matrix. The iteration step size is greater than zero; Update the oil extraction port position according to the following formula: ; in, This represents the exponential mapping from Lie algebras to Lie groups. Let represent the antisymmetric increment matrix constructed from the pose correction vectors. Let be the pose increment transformation matrix for the k-th iteration. Let be the pose matrix of the oil outlet relative to the camera coordinate system at the k-th iteration. This is the updated pose matrix of the oil extraction port relative to the camera coordinate system at the (k+1)th iteration.

[0045] The three reference planes are re-established with the updated pose, and the geometric error is recalculated until the termination condition is met. This process allows the initial position and orientation provided by the artificial visual signage to be gradually corrected under the constraints of the geometric features of the oil intake port, ultimately forming a complete sign-guided and geometrically constrained positioning process.

[0046] In some embodiments, the geometric error norm is calculated after each pose update: ; And calculate the pose change between two adjacent iterations: ; in, Let be the geometric error norm of the k-th iteration. This represents the pose update amount in the k-th iteration.

[0047] Iteration stops when any of the following conditions are met: (1) The geometric error norm is less than the preset error threshold: (2) The pose update amount is less than the preset update threshold: .

[0048] in, To preset the error threshold, This is the preset update threshold.

[0049] When either condition (1) or condition (2) is met, the current candidate pose is taken as the final pose of the oil extraction port: ; When condition (1) or condition (2) is met, and the maximum number of iterations is reached but the positioning error is still greater than the preset threshold, a positioning failure flag is output, and the local region of interest is expanded again or the input image is reacquired.

[0050] In summary, this invention achieves a two-level pose correction, from coarse to fine, by constructing a positioning architecture that integrates artificial visual marker guidance with the geometric constraints of the oil intake port. This enables the final positioning result to not only rapidly recognize visual markers but also iteratively constrain and correct the initial pose using the geometric features of the oil intake port end face, center, and axis. This effectively suppresses the impact of factors such as marker installation deviation, image distortion, illumination changes, and edge detection errors on positioning accuracy. Simultaneously, by constructing a local region of interest using the initial pose, the range of geometric feature extraction is significantly reduced from the entire scene to a local area, significantly lowering the computational overhead of point cloud processing and feature fitting, and avoiding misidentification problems caused by similar structures in complex backgrounds. Furthermore, by establishing an iteration termination judgment mechanism, the positioning result can simultaneously meet the quantitative requirements of positional accuracy and attitude accuracy, and possess the identifiability of positioning failures. This provides accurate, reliable, and verifiable six-degree-of-freedom pose information for the robotic arm's automatic oil extraction, effectively improving the success rate and operational stability of autonomous docking between the oil intake connector and the oil intake port.

[0051] like Figure 2 As shown, one embodiment of the present invention proposes a positioning system for the oil intake port of a converter station transformer, the system comprising: Image processing module 10 is used to acquire a color image containing artificial visual signs and an oil extraction port, and to convert the color image to grayscale and then to binarize it to obtain a binary image, and to extract the corner points of the artificial visual signs from the binary image. The initial pose calculation module 20 is used to construct the pose matrix of the artificial visual sign based on the pixel coordinates and three-dimensional coordinates of the sign corner points, and to calculate the initial pose of the oil extraction port based on the pose matrix. The region recognition module 30 is used to obtain the initial center position of the oil extraction port according to the initial pose, project the initial center position onto the color image, obtain the pixel coordinates of the initial center position according to the projection result, and determine the local region of interest according to the pixel coordinates of the initial center position. The feature extraction module 40 is used to convert the pixels in the local region of interest into three-dimensional points in the camera coordinate system, and to form a local point cloud of the oil intake port by all the three-dimensional points, and to extract the center of the oil intake port, the end face normal vector and the axis direction from the local point cloud. The pose correction module 50 is used to take the initial pose of the oil port as the initial value for fusion positioning, and use the center of the oil port, the end face normal vector and the axis direction as geometric constraints to iteratively correct the initial pose of the oil port until the preset termination condition is met, so as to obtain the final pose of the oil port.

[0052] In another aspect, the present invention also proposes a storage medium on which one or more programs are stored, which, when executed by a processor, implement the above-described positioning method for the oil tap of a converter station transformer.

[0053] In another aspect, the present invention also proposes an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to realize the above-mentioned positioning method for the oil tap of the converter station transformer.

[0054] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain stored, communicated, propagated, or transmitted programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0055] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0056] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0057] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A method for locating the oil intake port of a converter station transformer, characterized in that, The method includes: A color image containing the artificial visual sign and the oil extraction port is acquired, and the color image is sequentially converted to grayscale and then binarized to obtain a binary image. The corner points of the artificial visual sign are then extracted from the binary image. The pose matrix of the artificial visual sign is constructed based on the pixel coordinates and three-dimensional coordinates of the corner points of the sign, and the initial pose of the oil extraction port is calculated based on the pose matrix. The initial center position of the oil sampling port is obtained based on the initial pose, and the initial center position is projected onto the color image. The pixel coordinates of the initial center position are obtained based on the projection result, and the local region of interest is determined based on the pixel coordinates of the initial center position. The pixels in the local region of interest are converted into three-dimensional points in the camera coordinate system, and all the three-dimensional points are used to form a local point cloud of the oil intake port. The center of the oil intake port, the end face normal vector, and the axis direction are extracted from the local point cloud. Using the initial pose of the oil intake port as the initial value for fusion localization, the center of the oil intake port, the end face normal vector, and the axial direction are used as geometric constraints to iteratively correct the initial pose of the oil intake port until the preset termination condition is met, thus obtaining the final pose of the oil intake port.

2. The positioning method for the oil intake port of the converter station transformer according to claim 1, characterized in that, The steps of acquiring a color image containing artificial visual markers and an oil extraction port, and then sequentially converting the color image to grayscale and binarizing it to obtain a binary image, include: The color image is converted to a grayscale image, the grayscale image is divided into multiple local image blocks, and a local threshold is calculated for each local image block. Based on the local threshold, the grayscale image is then converted to a binary image. ； in, The binarization threshold for the k-th local image patch; , These are the maximum and minimum gray values ​​of the k-th local image block, respectively.

3. The positioning method for the oil intake port of the converter station transformer according to claim 2, characterized in that, The step of extracting the corner points of the artificial visual symbol from the binary image includes: The black and white pixels in the binary image are divided into several black connected regions and white connected regions, respectively, and all boundary pixels between adjacent black connected regions and white connected regions are obtained. Calculate the geometric center of all boundary pixels between adjacent black and white connected regions, obtain the azimuth angle of each boundary pixel relative to the geometric center, and sort all boundary pixels according to the azimuth angle. A sliding window is set on the sorted boundary pixels. The principal component analysis method is used to fit the boundary pixels in the sliding window with a straight line and the mean square error of the straight line fitting is calculated. The center position of the window where the mean square error of the straight line fitting is a local maximum and the local maximum is greater than the preset corner response threshold is recorded as a candidate corner point. Enumerate four-corner point combinations from all candidate corner points, connect the four candidate corner points of each four-corner point combination in sequence to obtain a quadrilateral, and perform the following judgments in sequence: the total number of boundary pixels is greater than a preset number threshold, the fitting error of all four sides of the quadrilateral is less than a preset side fitting error threshold, the included angle between adjacent sides of the quadrilateral is within a preset angle range, the area of ​​the quadrilateral is greater than a preset minimum area threshold, the aspect ratio of the quadrilateral is less than a preset maximum aspect ratio threshold, and the four candidate corner points satisfy the perspective projection coplanarity constraint. Among all the combinations that pass the judgment, the combination with the smallest sum of the fitting errors of the four sides, the largest quadrilateral area, and the closest included angle between adjacent sides to a right angle is selected as the optimal quadrilateral, and the four candidate corner points of the optimal quadrilateral are used as the marker corner points.

4. The positioning method for the oil intake port of the converter station transformer according to claim 3, characterized in that, The steps of constructing the pose matrix of the artificial visual sign based on the pixel coordinates and three-dimensional coordinates of the sign corner points, and calculating the initial pose of the oil extraction port based on the pose matrix, include: The initial pose of the oil sampling port is calculated using the following formula: ； ； ； in, The initial pose of the oil sampling port relative to the camera coordinate system. Let be the pose matrix of the artificial visual sign relative to the camera coordinate system. To obtain the homogeneous transformation matrix of the oil port coordinate system relative to the manual visual sign coordinate system, where T is the transpose, Let be the rotation matrix of the artificial visual sign relative to the camera coordinate system. Let be the translation vector of the artificial visual sign relative to the camera coordinate system. , Let x and y be the horizontal and vertical pixel coordinates of the i-th marker corner point, respectively, and K be the camera intrinsic parameter matrix. , , These are the X, Y, and Z coordinates of the i-th marker corner point, respectively.

5. The positioning method for the oil intake port of the converter station transformer according to claim 4, characterized in that, The steps of converting the pixels in the local region of interest into three-dimensional points in the camera coordinate system, constructing a local point cloud of the oil extraction port from all the three-dimensional points, and extracting the center, end face normal vector, and axis direction of the oil extraction port from the local point cloud include: Convert image pixels into 3D points in camera coordinates based on the depth values ​​of the local region of interest: ； in, These are the coordinates of a 3D point in the camera coordinate system. These are pixel coordinates; This represents the depth value of the corresponding 3D point; For the camera's horizontal and vertical focal lengths, , These are the x and y coordinates of the camera's principal point, respectively. The end face where the oil port is located is fitted from the local point cloud. Represented as: ； in, This is the unit normal vector of the oil inlet end face; These are three-dimensional points on the end face; These are end face parameters; During the fitting process, several 3D points are selected from the local point cloud to establish a candidate plane. The distances from the remaining 3D points to the candidate plane are calculated, and 3D points whose distances are less than the threshold of points in the plane are taken as inliers. The inner points of the oil intake port end face are transformed into the local two-dimensional coordinate system of the end face, and the circular edge of the oil intake port is fitted to obtain a circle fitting model. The center coordinates of the circle are obtained according to the circle fitting model, and the center coordinates of the circle are transformed from the local two-dimensional coordinate system of the end face back to the camera coordinate system to obtain the center coordinates of the oil intake port. The expression for the oil inlet axis is: ； in, Center of the oil intake port To obtain the unit direction vector of the oil port axis, For axis parameters, The parametric equation for the oil port axis; Construct two three-dimensional feature points on the oil intake axis: ； ； in, To preset the feature point spacing, For the first three-dimensional feature point, This is the second three-dimensional feature point.

6. The positioning method for the oil intake port of the converter station transformer according to claim 5, characterized in that, The steps of using the initial pose of the oil intake port as the initial value for fusion localization, and using the center of the oil intake port, the end face normal vector, and the axial direction as geometric constraints to iteratively correct the initial pose of the oil intake port until a preset termination condition is met, to obtain the final pose of the oil intake port, include: Establish three mutually orthogonal reference planes, the j-th reference plane being denoted as: ； in, Let j be the reference plane in the k-th iteration. Let be the unit normal vector of the j-th reference plane at the k-th iteration. The origin of the oil intake coordinate system is the point at the k-th iteration. Construct the geometric error vector for the k-th iteration according to the following formula. : ； in, , These are the signed distances from the first three-dimensional feature point and the second three-dimensional feature point to the first reference plane, respectively. , These are the signed distances from the first three-dimensional feature point and the second three-dimensional feature point to the second reference plane, respectively. The signed distance from the first three-dimensional feature point to the third reference plane; In the k-th iteration, the parameters to be corrected for the oil port pose are obtained. Represented as: ； in, These represent the x, y, and z translation corrections for the oil intake center in the camera coordinate system. , These are the rotation corrections for the oil inlet around the x-axis and y-axis of the camera coordinate system, respectively. The Jacobian matrix of the geometric error vector relative to the pose parameters is calculated using the following formula: ； Calculate the pose correction for the k-th iteration based on the current geometric error. : ； in, Let be the Jacobian matrix of the k-th iteration. This is the pseudo-inverse of the Jacobian matrix. The iteration step size is greater than zero; Update the oil extraction port position according to the following formula: ； in, This represents the exponential mapping from Lie algebras to Lie groups. Let represent the antisymmetric increment matrix constructed from the pose correction vectors. Let be the pose increment transformation matrix for the k-th iteration. Let be the pose matrix of the oil outlet relative to the camera coordinate system at the k-th iteration. This is the updated pose matrix of the oil extraction port relative to the camera coordinate system at the (k+1)th iteration.

7. The positioning method for the oil intake port of the converter station transformer according to claim 6, characterized in that, The step of using the initial pose of the oil intake port as the initial value for fusion localization, and using the center of the oil intake port, the end face normal vector, and the axial direction as geometric constraints to iteratively correct the initial pose of the oil intake port until a preset termination condition is met, to obtain the final pose of the oil intake port, further includes: After each pose update, calculate the geometric error norm: ； And calculate the pose change between two adjacent iterations: ； in, Let be the geometric error norm of the k-th iteration. Let $\mathbf{k}$ be the pose update amount in the k-th iteration. The iteration stops when any of the following conditions are met: the geometric error norm is less than the preset error threshold, or the pose update amount is less than the preset update threshold. The current candidate pose is then used as the final pose of the oil extraction port. If the conditions are not met and the number of iterations reaches the preset maximum number of iterations, output a positioning failure flag.

8. A positioning system for the oil inlet of a converter station transformer, characterized in that, The system includes: The image processing module is used to acquire a color image containing artificial visual signs and an oil extraction port, and to convert the color image to grayscale and then to binarize it to obtain a binary image, and to extract the corner points of the artificial visual signs from the binary image; The initial pose calculation module is used to construct the pose matrix of the artificial visual sign based on the pixel coordinates and three-dimensional coordinates of the sign corner points, and to calculate the initial pose of the oil extraction port based on the pose matrix. The region recognition module is used to obtain the initial center position of the oil sampling port according to the initial pose, project the initial center position onto the color image, obtain the pixel coordinates of the initial center position according to the projection result, and determine the local region of interest according to the pixel coordinates of the initial center position. The feature extraction module is used to convert the pixels in the local region of interest into three-dimensional points in the camera coordinate system, and to form a local point cloud of the oil intake port from all the three-dimensional points, and to extract the center of the oil intake port, the end face normal vector and the axis direction from the local point cloud; The pose correction module is used to use the initial pose of the oil tap as the initial value for fusion positioning, and uses the center of the oil tap, the end face normal vector, and the axis direction as geometric constraints to iteratively correct the initial pose of the oil tap until the preset termination condition is met, thus obtaining the final pose of the oil tap.

9. A storage medium, characterized in that, The storage medium stores one or more programs that, when executed by a processor, implement the positioning method for the oil intake port of the converter station transformer as described in any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the positioning method for the oil tap of the converter station transformer as described in any one of claims 1-7.

Images