A method for detecting the winding spacing of a transformer

Abstract

The present invention provides an automatic measurement method for the wire winding pitch of a transformer winding based on vision detection, which is achieved through the following steps: 1. Image acquisition; 2. Image processing; 3. Wire contour recognition; 4. Establishing a contour model; 5. Finding extreme points; 6. Finding the intersection point of the perpendicular bisectors of the connecting lines of the extreme points; 7. Calibrating parameters and returning the wire pitch. The method of the present invention has the ability to non-contact measure key parameters. Machine vision is widely used in the field of image detection and is also the development direction of future unmanned technology. For the current high-speed automated equipment production line, the use of vision detection methods can effectively avoid missed or misjudged cases, ensure the finished product efficiency of products, and to a certain extent, save labor costs and improve inspection efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition, and particularly relates to a method for detecting the winding pitch of a transformer winding. Background Art

[0002] As a basic equipment for power transmission and transformation in the power industry, transformers are widely used in power plants, converter stations, substations and user terminals, and occupy a very important position in the power industry. Usually, a power transmission and distribution transformer mainly consists of high and low voltage windings, a magnetic core, an insulation device and a protection device. Among them, the high voltage winding is made of polyester enameled copper wire through a winding mechanism, and the quality of the winding is a key factor affecting the working reliability of the transformer.

[0003] Generally speaking, the winding pitch of the wire will affect the wire arrangement and wiring of the winding machine, indirectly affecting the laying of the next layer. Usually, the goal is to introduce as much copper wire as possible into the limited winding space. However, in an irregular winding, the wires of each layer are wound without any rules, adjacent and overlapping unevenly with each other, resulting in intersections and cavities in the winding structure. The intersections and cavities cause the wires of the next layer to be embedded into the gaps of the upper layer, changing the overall resistance of the winding and making the entire winding unable to work properly. The ideal winding structure should be a structure whose cross-section is similar to a honeycomb. Each turn of wire extends side by side and as parallel as possible to the flange of the coil body, and the number of turns in a layer extends on the cylindrical surface coaxial with the central axis of the coil frame. Therefore, it is necessary to synchronously check the winding state during winding to optimize the process parameters of the winding process. According to the investigation of domestic machinery processing enterprises at present, the quality inspection of most metal products still mainly relies on manual inspection. After winding, the operator measures whether the winding angle and wire spacing are within the specified range by observing or using a micrometer. However, the judgment criteria of different operators rely on subjective experience, and it is necessary to observe the wire for a long time, which is easy to cause visual fatigue, and it is impossible to timely check the internal situation of the turn-by-turn winding during rapid machine winding.

[0004] Machine vision is widely used in the field of image detection and is also the development direction of future unmanned technology. For the current high-speed automated equipment production line, using the vision detection method can effectively avoid missed or misjudged cases and ensure the finished product efficiency of the product. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for detecting the winding pitch of a transformer winding to solve the problem of low detection efficiency mentioned in the above background art.

[0006] The technical problems solved by the present invention are realized through the following technical solutions:

[0007] A method for detecting the winding pitch of a transformer winding includes the following steps:

[0008] Step 1: Extract the recognition area from the high-voltage winding image, establish the image coordinates, and perform preprocessing;

[0009] Step 2: Calculate the wire contour coordinates, expand the contour edges, and calculate the set of extreme points of the contour coordinates;

[0010] Step 3: Extract the optimal values from the set of extreme coordinates, and establish 4 contour models according to the characteristics of the optimal values;

[0011] Step 4: Connect adjacent extreme points according to the characteristics of different models, and find the perpendicular bisector of the connecting line;

[0012] Step 5: Calculate the intersection points of adjacent perpendiculars, classify the intersection points, calibrate the pixels, and convert the pixel distance between adjacent intersection points into the actual physical distance.

[0013] In the above method for detecting the winding pitch of a transformer, Step 1 is specifically: Take out the wire edge area facing the camera from the captured winding image, take the wire edge image of about 10 turns in front of the wire dropping point according to production experience, establish a Cartesian coordinate system, binarize this part of the image using the Otsu threshold segmentation method, and use the Canny edge detection algorithm to identify the edge contour.

[0014] In the above method for detecting the winding pitch of a transformer, Step 2 is specifically: Calculate the contour edge coordinates of the contour image in Step 1, horizontally expand the head and tail of the contour, recalculate the edge coordinates and store them in the coordinate container, and further find the extreme points in the coordinate container.

[0015] In the above method for detecting the winding pitch of a transformer, Step 3 is specifically: After the wire is flattened, the contour is horizontal, and the extreme points are not unique. It is necessary to select the most suitable points as the final extreme points. The extreme points include maximum points and minimum points, and 4 practical application models are established according to the distribution form.

[0016] In the above method for detecting the winding pitch of a transformer, Step 4 is specifically: Connect adjacent maximum and minimum points, find the slope and midpoint coordinates of the connecting line, and calculate the perpendicular bisector.

[0017] In the above method for detecting the winding pitch of a transformer, Step 5 is specifically: The perpendicular bisectors obtained in Step 4 will form intersection points. Sort and classify the intersection points according to different models. The Euclidean distance between a group of adjacent intersection points represents the pixel pitch of the wire. Perform camera calibration to obtain the conversion relationship between the pixel length and the physical length, and convert the pixel distance into the actual physical distance.

[0018] The beneficial effects of the present invention are as follows: (1) The imaging device used in this method occupies a small space and is easy to assemble; (2) Compared with traditional manual measurement, this detection method has the characteristics of accurate positioning, fast calculation, simple operation, and non-contact. Measuring parameters using machine vision technology saves human resources to a certain extent while ensuring the accuracy of measurement results, which is not available in existing manual detection methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 For deriving the calculation assistance for the wire winding pitch Figure 1

[0020] Figure 2 For deriving the calculation assistance for the wire winding pitch Figure 2

[0021] Figure 3 For deriving the calculation assistance for the wire winding pitch Figure 3

[0022] Figure 4 For deriving the calculation assistance for the wire winding pitch Figure 4

[0023] Figure 5 For deriving the calculation assistance for the wire winding pitch Figure 5

[0024] Figure 6 Steps for calculating the wire winding pitch DETAILED DESCRIPTION OF THE INVENTION

[0025] The present invention will be described in detail below with reference to the drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention and do not represent all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] A method for detecting the winding pitch of a transformer winding includes the following steps:

[0027] Step 1: Extract the recognition area from the high-voltage winding image, establish an image coordinate, and perform preprocessing;

[0028] Step 2: Calculate the wire contour coordinates, expand the contour edge, and calculate the set of extreme points of the contour coordinates;

[0029] Step 3: Extract the optimal values from the set of extreme coordinates and establish 4 contour models according to the characteristics of the optimal values;

[0030] Step 4: Connect adjacent extreme points according to the characteristics of different models and find the perpendicular bisector of the connection line;

[0031] Step Five: Calculate the intersection points of adjacent perpendicular lines, classify the intersection points, calibrate the pixels, and convert the pixel distance between adjacent intersection points into the actual physical distance.

[0032] The photo taken by the image acquisition device in Step One contains windings, wires, the main shaft, and other information. Set a Mask mask for the image, perform convolution on the mask and the original image to extract the contour area image of about 10 turns of the coil in front of the wire drop point, establish a Cartesian coordinate system with the upper left corner as the image origin, use the Otsu threshold segmentation algorithm to distinguish the foreground and background of the image to obtain a binary image, and use the Canny edge detection algorithm to obtain the contour curve of the wire.

[0033] In Step Two, starting from the origin, first traverse the pixel columns and then the pixel rows to convert the contour coordinates into pixel coordinates. Since the width of the wire and the pixel calibration ratio at different shooting heights are known, the pixel width t of the wire can be calculated. Let m represent the number of extended pixels, and take 3 / 8 of the pixel width t of the wire as the value of m. Add background pixels with a width of m pixels to both sides of the contour image, and set all the pixels in the added image on the left that have the same ordinate as the first contour point in the original image to the contour line. Similarly, on the right, set all the pixels that have the same ordinate as the last contour point in the original image to the contour line. Establish a peak-valley data group of the wire boundary coordinates, traverse the extended edge coordinates, starting from the position of the original image contour coordinates, compare the ordinate of this edge point with the ordinates of a total of 2m contour points on both sides. If the ordinate of this point is less than or equal to the ordinates of these 2m contour points, it is considered a peak of the contour line (the ordinate direction is opposite to the Cartesian coordinate system during image processing). Store this contour point in the peak number container, and traverse all the points in the original image in turn to obtain all possible coordinate points of the peaks. Refer to the example Figure 1 When calculating the valleys, do the same, change the judgment symbol to less than or equal to, and traverse the contour pixels again to obtain all possible coordinate points of the valleys.

[0034] In Step Three, for a peak or valley at a certain position, it may not be a single extreme point but may be multiple extreme points. In this case, take the median of all the extreme points as the new extreme point. Refer to the example Figure 2 Using the pseudo-bubble sort method, the steps to simplify the valley data are as follows:

[0035] 1). Establish two empty sets, and store all the valley coordinate points obtained in Step Two into the first set;

[0036] 2). Select the first coordinate point in the set as the target value;

[0037] 3). Starting from the second coordinate, compare the pixel distances with the first coordinate point in sequence. As long as the distance is less than m pixels, these coordinate points are considered adjacent troughs. Put the coordinates of this series of troughs into another empty set, calculate the median point of these points, and record the length of this set as L;

[0038] 4). Delete L coordinates from the beginning of the first set and return to step 2;

[0039] 5). Repeat steps 2, 3, and 4 until the length of the first set is 0 and end.

[0040] Each cycle will leave one coordinate, and after simplification, a combination of trough coordinates is obtained. Simplifying the peak data is the same principle.

[0041] Furthermore, the contour line of the image generally does not start exactly from a certain peak or trough, nor does it end exactly at a certain peak or trough. In this way, two pseudo peak and trough coordinate points are formed on the image, and the existence of these two pseudo points will affect the subsequent result calculation. Eliminate these two points, take the first coordinate in each of the two containers respectively. Let the first coordinate of the peak container be (x1, y1), and the first coordinate of the trough container be (x2, y2). Compare their magnitudes. If x1 > x2, it is considered that the image starts from the peak; if x1 < x2, the image starts from the trough; similarly, take the last coordinate in each of the two containers respectively. Let the last coordinate of the peak container be (xn, yn), and the last coordinate of the trough container be (xm, ym). If xn > xm, it is considered that the image ends at the trough; if xn < xm, the image ends at the peak. Such pairwise combinations will form four types. Referring to the example Figure 3 They are respectively defined as "W type", "M type", "N type" and "inv-N type", and for these four different situations, the subsequent steps need to calculate the results separately for each type.

[0042] Step 4 refers to the example Figure 4 Connect the adjacent peak and trough coordinate points, find the perpendicular bisector of each connection line. Each adjacent pair of perpendicular bisectors will intersect to form an intersection point. For example, the perpendicular bisector of a connection line from a trough to a peak can be represented by the point-slope form for its position, and the coordinate point is (k1, x1, y1), where k1 is the slope. The point-slope form of the perpendicular bisector of the connection line from the adjacent peak to the trough is (k2, x2, y2).

[0043] The intersection point calculation formula in step 5:

[0044] x = (k1x1 - y1 - k2x2 + y2) ÷ (k1 - k2)

[0045] y = k1(x - x1) + y1,

[0046] Among them, x1, y1, and k1 represent a straight line passing through the point (x1, y1) with a slope of k1, and x2, y2, and k2 represent a straight line passing through the point (x2, y2) with a slope of k2. The obtained x and y are the intersection coordinates of these two lines.

[0047] Further, after determining the starting order of the image in step three, if the image starts from the trough, the 1st, 3rd, 5th,... (2n + 1)th intersection points of the perpendicular line are stored in a new data container for the center coordinates of the wire; if the image starts from the peak, the 2nd, 4th, 6th,... 2nth intersection points are stored in a new data container for the wire gap.

[0048] Further, referring to the example Figure 5 , two sets of coordinate data are obtained. Among them, the data in the data container for the center coordinates of the wire represents the position coordinates of all wires in the corresponding contour image. To obtain the true adjacent distance between the wires, it is necessary to obtain the true physical length corresponding to a single pixel point, that is, the conversion ratio k between the pixel coordinate system and the world coordinate system, which is defined as:

[0049]

[0050] In the formula, l is the true physical size, with the unit of mm, and p is the pixel size, with the unit of pixel.

[0051] Further, the pixel distances between adjacent coordinates are calculated by taking the differences between the data groups representing the center coordinates of the wires in pairs, and the true physical winding spacing is obtained by subtracting the pixel width of a wire from the pixel distance.

[0052] In addition, it should be understood that although this specification is described according to the embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. However, any simple modification, equivalent change, or modification of the above embodiments in any form that does not depart from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A method for detecting the winding pitch of a transformer, characterized in that It is implemented by an image acquisition device. The steps of the winding wire winding pitch detection method include: Step 1: Extract the recognition area from the high-voltage winding image, establish an image coordinate, and perform preprocessing; Step 2: Calculate the wire contour coordinates, expand the contour edge, and calculate the set of extreme points of the contour coordinates; Step 3: Extract the optimal value from the set of extreme coordinates, and establish 4 contour models according to the characteristics of the optimal value; Step 4: Connect adjacent extreme points according to the characteristics of different models, and find the perpendicular bisector of the connection line; Step 5: Calculate the intersection points of adjacent perpendiculars, classify the intersection points, calibrate the pixels, and convert the pixel distance between adjacent intersection points into the actual physical distance.

2. The method for detecting the winding pitch of a transformer winding according to claim 1, wherein Specifically, Step 1 is: Take out the wire edge area facing the camera from the captured winding image, take the wire edge images of the wire dropping point and about 10 turns in front, establish a Cartesian coordinate system, binarize this part of the image using the Otsu threshold segmentation method, and use the Canny edge detection algorithm to identify the edge contour.

3. A method for detecting the winding pitch of a transformer winding according to claim 1, characterized in that, Specifically, Step 2 is: Calculate the contour edge coordinates of the contour image in Step 1, horizontally expand the head and tail of the contour, recalculate the edge coordinates and store them in the coordinate container, and further find the extreme points in the coordinate container.

4. A method for detecting the winding pitch of a transformer winding according to claim 1, characterized in that, Specifically, Step 3 is: The contour after the wire is flattened is horizontal, and the extreme points are not unique. The most suitable points need to be selected from them as the final extreme points. The extreme points include maximum points and minimum points. 4 practical application models are established according to the distribution form.

5. A method for detecting the winding spacing of a transformer winding according to claim 1, characterized in that, Specifically, Step 4 is: Connect adjacent maximum and minimum points, find the slope and midpoint coordinates of the connection line, and calculate the perpendicular bisector.

6. The method for detecting the winding pitch of a transformer winding according to claim 1, characterized in that Specifically, Step 5 is: The perpendicular bisectors obtained in Step 4 will form intersection points. The intersection points are sorted and classified according to different models. The Euclidean distance between a group of adjacent intersection points represents the pixel pitch of the wire. Perform camera calibration to obtain the conversion relationship between the pixel length and the physical length, and convert the pixel distance into the actual physical distance.

Images