CN-121391998-B - Coil positioning method, device and computer equipment

CN121391998BCN 121391998 BCN121391998 BCN 121391998BCN-121391998-B

Abstract

The application relates to a coil positioning method, a coil positioning device and computer equipment. The method comprises the steps of obtaining a top view image of a target coil to be positioned, extracting edge features of the top view image according to prior parameters of the top view image to obtain a target edge feature set of the top view image, carrying out key region identification on the input top view image through a trained region positioning model to obtain a key region in the top view image, carrying out coil fitting according to the key region and the target edge feature set to obtain fitting data of the target coil, and determining position information corresponding to the target coil according to the fitting data. By adopting the method, the coil positioning precision can be improved.

Inventors

Yao Hekai
YUE XIAOFENG
SHI PAN
YANG XIAO
YANG LETIAN
DAI HENG
WANG FABIN
JIANG JING
ZHANG XUETAO

Assignees

武汉华工赛百数据系统有限公司
大连理工大学

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1. A coil positioning method, the method comprising: acquiring a top view image of a target coil to be positioned; Extracting edge pixels in the overlook image by using an edge drawing algorithm, and connecting all the extracted edge pixels to obtain an edge pixel chain; Performing forward tracking along the edge pixel chain, performing straight line fitting on a plurality of currently tracked edge pixels by adopting a least square method, calculating accumulated residual errors, and continuously tracking subsequent edge pixels until the accumulated residual errors are larger than a preset residual error threshold value under the condition that the accumulated residual errors are smaller than or equal to the preset residual error threshold value, and outputting an initial fitting line segment of the currently tracked pixels; Traversing the residual edge pixels, updating the initial fitting line segment based on the traversed edge pixels under the condition that the residual edge pixels which are traversed currently meet the preset increment condition until the residual edge pixels which are traversed currently do not meet the preset increment condition, obtaining edge straight line segments corresponding to all edge pixels which are tracked currently and traversed currently, and returning to execute the step of forward tracking along the edge pixel chain; according to prior parameters, carrying out data cleaning on an initial edge feature set to obtain a target edge feature set of the overlook image, wherein the initial edge feature set comprises an initial edge contour set and an initial edge line segment set, the initial edge contour set comprises the edge pixel chain, the initial edge line segment set comprises a plurality of edge straight line segments, and the target edge feature set comprises a target edge contour set and a target edge line segment set; carrying out key region identification on the input overlook image through a trained region positioning model to obtain a key region in the overlook image; Performing coil fitting according to the key region and the target edge feature set to obtain fitting data of the target coil; and determining the position information corresponding to the target coil according to the fitting data.
2. The method of claim 1, wherein the target coil comprises an elliptical structure and a boss structure connected to each other, the boss structure being located on a side of the elliptical structure, the critical region comprising at least an endpoint region, wherein the performing coil fitting based on the critical region and the target edge feature set, obtaining fitting data of the target coil comprises: performing ellipse fitting according to the target edge feature set to obtain ellipse fitting data of an ellipse shape, wherein the ellipse shape is the imaging of the ellipse structure in the overlook image; and performing line segment fitting according to the endpoint region and the target edge feature set to obtain line segment fitting data of a target edge of a boss shape, wherein the boss shape is formed by imaging the boss structure in the overlook image, and the target edge is positioned in the endpoint region.
3. The method of claim 2, wherein the critical region further comprises at least one vertex region, the method further comprising: And performing line segment fitting according to at least one vertex region and the target edge feature set to obtain line segment fitting data of at least one vertex tangent line, wherein the vertex tangent line is a tangent line of the vertex of the elliptical shape.
4. The method according to claim 2, wherein said performing ellipse fitting based on the target edge feature set to obtain ellipse fitting data of an ellipse shape comprises: and fitting the elliptical shape of the target coil in the overlook image by utilizing a random sampling consistency elliptical fitting technology according to the target edge contour set in the target edge feature set, and obtaining corresponding elliptical fitting data.
5. The method of claim 1, wherein the position information includes a target rotation angle of the target coil and clamping position information of a clamping device, and wherein the determining the position information corresponding to the target coil based on the fitting data includes: determining a target rotation angle of the target coil according to the fitting data; and determining the clamping position information of the clamping device for clamping the target coil according to the target rotation angle and the initial position information of the clamping device.
6. The method of claim 5, wherein the fitting data comprises line segment fitting data and ellipse fitting data, and wherein the determining the target rotation angle of the target coil from the fitting data comprises: and determining the target rotation angle by utilizing a multipath algorithm according to the line segment fitting data and the ellipse fitting data.
7. The method of claim 1, wherein the predetermined incremental condition comprises a geometric deviation less than or equal to a predetermined deviation threshold and an absolute value of an angular deviation between the edge pixel gradient direction and the current initial fit line segment less than or equal to a predetermined angle threshold.
8. The method according to claim 1, wherein the method further comprises: Constructing a training set, a testing set and a verification set based on an image dataset of the coil and a marked key area, wherein the image dataset comprises overlooking images under various rotation angles; and training the pre-constructed initial area positioning model according to the training set, the testing set and the verification set to obtain the trained area positioning model.
9. A coil positioning device, the device comprising: the acquisition module is used for acquiring a top view image of the target coil to be positioned; The extraction module is used for extracting edge pixels in the overlook image by utilizing an edge drawing algorithm and connecting all the extracted edge pixels to obtain an edge pixel chain; performing straight line fitting on a plurality of currently tracked edge pixels along the edge pixel chain in a forward tracking manner by adopting a least square method, calculating accumulated residual errors, continuously tracking subsequent edge pixels until the accumulated residual errors are smaller than or equal to a preset residual error threshold value and outputting initial fitting line segments of the currently tracked pixels, traversing residual edge pixels, updating the initial fitting line segments based on the traversed edge pixels until the residual edge pixels do not meet preset increment conditions under the condition that the residual edge pixels currently traversed meet the preset increment conditions, obtaining edge straight line segments corresponding to all the currently tracked and currently traversed edge pixels, and returning to the step of executing forward tracking along the edge pixel chain; the identification module is used for carrying out key area identification on the input overlook image through the trained area positioning model to obtain a key area in the overlook image; The fitting module is used for performing coil fitting according to the key region and the target edge feature set to obtain fitting data of the target coil; and the determining module is used for determining the position information corresponding to the target coil according to the fitting data.
10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 8 when the computer program is executed.

Description

Coil positioning method, device and computer equipment Technical Field The present application relates to the field of computer vision, and in particular, to a coil positioning method, apparatus, computer device, computer readable storage medium, and computer program product. Background In the manufacture of core components of power equipment, special-shaped coils with side boss structures are common key components in power equipment such as transformers and reactors due to adaptation to special working condition requirements. At present, most enterprises in the industry generally adopt a mode of 'clamping equipment matched with manual operation' when processing and manufacturing the coils, and complete feeding and discharging of the coils and transfer among different working procedures. In the process, an operator needs to concentrate on the height, and the position and the angle of the clamping device are precisely controlled, so that high requirements are placed on the accuracy of the operation of the operator. Meanwhile, the whole process is highly dependent on manual control equipment, so that the labor intensity of staff is obviously increased, and the labor cost of enterprises is increased. In addition, the operator needs to perform continuous operation for a long time under the high precision requirement, fatigue is very easy to cause, further, the efficiency of manual transportation is gradually reduced, and the precision of the clamping operation is also reduced, so that unexpected collision between the clamp and the coil structure is easy to occur. The coil is often physically deformed due to collision, so that the final quality of the coil is seriously affected, and potential safety hazards are hidden in the actual production environment. With the development of artificial intelligence technology, computer vision technology has made breakthrough progress in various fields. In the technical field of industrial automation, the position and the direction of a workpiece can be efficiently and accurately calculated by utilizing object positioning and grabbing of a computer vision technology, and the mechanical arm or the truss is guided to finish tasks such as automatic assembly and carrying, so that the teaching programming time can be reduced, randomly placed workpieces are adapted, and the production line efficiency is improved. The technology provides a solid foundation for solving the problems of accurate positioning and automatic grabbing of coils such as special-shaped coils. However, the existing computer vision positioning and grabbing technology still has the problem of inaccurate positioning when aiming at the special-shaped coil with the side boss structure. Disclosure of Invention In view of the foregoing, it is desirable to provide a coil positioning method, apparatus, computer device, computer-readable storage medium, and computer program product that are capable of improving coil positioning accuracy. In a first aspect, the present application provides a coil positioning method, the method comprising: acquiring a top view image of a target coil to be positioned; according to the prior parameters of the overlook image, extracting edge features of the overlook image to obtain a target edge feature set of the overlook image; carrying out key region identification on the input overlook image through a trained region positioning model to obtain a key region in the overlook image; Performing coil fitting according to the key region and the target edge feature set to obtain fitting data of the target coil; and determining the position information corresponding to the target coil according to the fitting data. In one embodiment, the extracting edge features of the top view image according to the prior parameters of the top view image to obtain the target edge feature set of the top view image includes: Extracting edge features of the overlook image to obtain an initial edge feature set of the overlook image; and carrying out data cleaning on the initial edge feature set according to the prior parameters to obtain the target edge feature set. In one embodiment, the initial edge feature set includes an initial edge contour set and an initial edge line segment set, the target edge feature set includes a target edge contour set and a target edge line segment set, and the extracting the edge feature of the top view image to obtain the initial edge feature set in the top view image includes: Extracting edge pixels in the overlook image by using an edge drawing algorithm, and connecting all the extracted edge pixels to obtain an edge pixel chain; and performing linear line segment fitting according to edge pixels in the initial edge contour set to obtain a plurality of edge linear line segments, wherein the initial edge line segment set comprises a plurality of edge linear line segments. In one embodiment, the performing straight line segment fitting according to the edge pixels in the initial edge