A method and system for winding the rotor of a propulsion motor

Abstract

The present invention relates to the technical field of motor manufacturing, and specifically relates to a method and system for winding a rotor of a propulsion motor. The method includes: obtaining a spectral image of a winding grayscale image, selecting and filtering high-frequency information by a central circle to obtain a high-frequency image; determining a motion stability factor according to the position of a winding probe, and determining an image stability factor by edge detection and image recognition; screening to obtain a stable frame image; respectively performing gradient analysis on the stable frame image and the high-frequency image to determine a first histogram and a second histogram, and determining a similarity parameter between the first histogram and the second histogram according to the difference in the gradient distribution structures of the first histogram and the second histogram; screening to obtain an optimal radius according to the similarity parameter, converting the high-frequency image at the optimal radius into a spatial domain to obtain a deblurred image, and performing winding detection on the deblurred image to determine a winding grayscale image with winding anomalies. The present invention can improve the accuracy of rotor winding detection in the manufacturing process of a propulsion motor.

Description

Technical Field

[0001] The present invention relates to the technical field of motor manufacturing, and particularly relates to a method and system for winding a rotor of a propulsion motor. Background Art

[0002] The rotor of a motor generally includes a rotor core and a rotor winding. At present, most manufacturers in the market use high-speed stamping machines to process and produce the rotor core, and then wind copper wires on the rotor core through specialized production equipment to form the rotor winding. The quality of the rotor is closely related to the processing quality of the rotor core and the winding quality of the rotor winding. Uneven winding of the rotor winding is an important factor affecting the quality of the rotor winding.

[0003] In related technologies, by acquiring image information and checking whether there are abnormalities in rotor winding (such as uneven winding, wrong wiring, etc.) during the motor manufacturing process according to the image information, in this way, affected by the shutter speed of the camera, motion blur will be generated when the rotor rotates. This motion blur effect will make the rotor area blurred, losing the fine features on the rotor surface and the motion details during the winding process, affecting the winding detection in the manufacturing process of the propulsion motor, and resulting in poor accuracy of the rotor winding detection of the propulsion motor. Summary of the Invention

[0004] In order to solve the technical problem that in related technologies, due to the motion blur generated by the rotation of the rotor, the accuracy of the rotor winding detection of the propulsion motor is poor, the present invention provides a method and system for winding a rotor of a propulsion motor, and the specific technical solutions adopted are as follows:

[0005] The present invention proposes a method for winding a rotor of a propulsion motor, and the method includes:

[0006] Obtaining the winding grayscale images of the winding probe on the side of the motor during the rotor winding process at different sampling moments, converting the winding grayscale images into spectral images in the frequency domain, selecting a central circle with different radii, screening the high-frequency information of the central circle area from the spectral images, and obtaining high-frequency images with different radii;

[0007] Determining a motion stability factor according to the position of the winding probe in the winding grayscale image, performing edge detection and image recognition on the winding grayscale image to determine the image connected domain belonging to the rotor, and determining an image stability factor according to the size of the image connected domain and the number of edge pixel points; Combining the motion stability factor and the image stability factor, screening and obtaining stable frame images from different frames of winding grayscale images;

[0008] For high-frequency images of any radius, perform gradient analysis on the stable frame image and the high-frequency image of the same winding grayscale image respectively. Determine the gradient distribution histogram of the stable frame image as the first histogram, and determine the gradient distribution histogram of the high-frequency image as the second histogram. According to the difference in the gradient distribution structure between the first histogram and the second histogram, determine the similarity parameter between the first histogram and the second histogram;

[0009] Select the optimal radius according to the similarity parameter, convert the high-frequency image under the optimal radius into the spatial domain to obtain a deblurred image, and perform winding detection on the deblurred image to determine the winding grayscale image with winding anomalies.

[0010] Further, the step of using different radii to select the central circle, screening the high-frequency information in the central circle area from the spectral image, and obtaining high-frequency images with different radii includes:

[0011] Cut the spectral image with a straight line passing through the image center and parallel to both sides, and piece together the four corners belonging to the spectral image to a central point to obtain an image to be selected with low-frequency information in the center;

[0012] Taking the central point of the image to be selected as the center of the circle, use different radii to determine the central circle of the image to be selected, and delete the central circle part to obtain high-frequency images with different radii.

[0013] Further, the step of determining the motion stability factor according to the position of the winding probe in the winding grayscale image includes:

[0014] Determine the positions of the winding probe in all frames of the winding grayscale image to obtain the motion trajectory of the winding probe, and obtain two endpoints of the motion trajectory;

[0015] Obtain the Euclidean distances between the position of the winding probe in any frame of the winding grayscale image and the two endpoints respectively, and take the maximum value of the Euclidean distances as the motion stability factor of the corresponding frame of the winding grayscale image.

[0016] Further, the step of performing edge detection and image recognition on the winding grayscale image to determine the image connected domain belonging to the rotor includes:

[0017] Perform canny edge detection on the winding grayscale image to obtain edge pixel points, and the edge pixel points form an initial connected domain;

[0018] Perform image recognition on the initial connected domain, and recognize the initial connected domain belonging to the rotor as the image connected domain.

[0019] Further, the step of determining the image stability factor according to the size of the image connected domain and the number of edge pixel points includes:

[0020] Calculate the ratio of the number of pixel points within the connected region to the number of the edge pixel points obtained by edge detection as the image stability factor.

[0021] Further, combining the motion stability factor and the image stability factor to screen out stable frame images from different-frame winding grayscale images includes:

[0022] Calculate the product of the motion stability factor and the image stability factor, and perform maximum-minimum normalization processing to obtain a stability index;

[0023] Take the winding grayscale image with the stability index greater than the preset stability threshold as the stable frame image.

[0024] Further, determining the similarity parameter between the first histogram and the second histogram according to the gradient distribution structure difference between the first histogram and the second histogram includes:

[0025] Divide the gradient intensities in the first histogram and the second histogram into the same intervals to obtain corresponding gradient intervals in the first histogram and the second histogram;

[0026] Calculate the squared value of the difference in frequencies under the same gradient intervals in the first histogram and the second histogram to obtain a first index;

[0027] Calculate the sum value of the frequencies under the same gradient intervals in the first histogram and the second histogram as a second index;

[0028] Calculate the ratio of the first index to the second index, and perform maximum-minimum normalization processing on the reciprocal of this ratio as the interval similarity coefficient for the corresponding gradient interval;

[0029] Take the mean value of the interval similarity coefficients of all gradient intervals as the similarity parameter between the first histogram and the second histogram.

[0030] Further, screening out the optimal radius according to the similarity parameter includes:

[0031] Take the radius corresponding to the maximum value of the similarity parameter of all high-frequency images as the optimal radius.

[0032] Further, performing winding detection on the deblurred image to determine the winding grayscale image with winding anomalies includes:

[0033] Input the deblurred image into a pre-trained winding detection model to output the winding anomaly inspection result of the winding grayscale image.

[0034] On the other hand, the present invention also provides a winding system for a propulsion motor rotor. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of a winding method for a propulsion motor rotor as described in any one of the foregoing are implemented.

[0035] The present invention has the following beneficial effects:

[0036] In the present invention, the central circle is selected through the spectral image, the low-frequency information affected by motion blur is removed, and the high-frequency information is retained to obtain high-frequency images with different radii. Then, the stable frame image is determined according to features such as the position of the winding probe, the number of edge pixel points of the rotor, and the area size. Gradient analysis is performed on the stable frame image and the high-frequency image, the similarity parameter of the gradient feature is determined through the gradient distribution histogram, the optimal radius is selected according to the similarity parameter, the high-frequency image under the optimal radius is converted into the spatial domain to obtain the deblurred image, and the deblurred image is subjected to winding detection to determine the winding gray image with winding anomalies. The present invention can effectively analyze the winding process of the rotating rotor, and through the analysis of the characteristics of the image acquisition device itself and the dynamic blur characteristics generated by the rotation of the rotor, the stable frame image can be accurately selected, and the deblurring process can be performed using an adaptive and appropriate scale to obtain a clearer and more reliable deblurred image. Winding detection is performed through the deblurred image, avoiding the interference caused by dynamic blur and improving the accuracy of rotor winding detection in the manufacturing process of the propulsion motor. Description of the Drawings

[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 Flowchart of a winding method for a propulsion motor rotor provided by an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of motor rotor winding provided by an embodiment of the present invention;

[0040] Figure 3 Schematic diagram of the high-frequency image acquisition process provided by an embodiment of the present invention;

[0041] Figure 4 Structural diagram of a winding system for a propulsion motor rotor provided by an embodiment of the present invention. Detailed Embodiments

[0042] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following specifically describes, with reference to the accompanying drawings and preferred embodiments, a method and system for winding the rotor of a propulsion motor according to the present invention, including its specific implementation manner, structure, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0044] It should be noted that to ensure the significance of the calculation results, in the fractional operations in the embodiments of the present invention, when encountering the situation where the denominator is 0, a tuning factor greater than 0 needs to be added to the denominator to prevent the denominator from being 0. The value of the tuning factor is set by the implementer according to the actual situation, and no special restrictions are imposed in this application.

[0045] The following specifically describes the specific solution of a method for winding the rotor of a propulsion motor provided by the present invention with reference to the accompanying drawings.

[0046] Please refer to Figure 1 , which shows a flowchart of a method for winding the rotor of a propulsion motor provided by an embodiment of the present invention. The method includes:

[0047] S101: Obtain the winding grayscale images of the winding probe on the side of the motor during the rotor winding process at different sampling moments, convert the winding grayscale images into spectral images in the frequency domain, select the central circle using different radii, and screen the high-frequency information in the central circle area from the spectral images to obtain high-frequency images at different radii.

[0048] The scenario corresponding to the embodiment of the present invention is the motor rotor winding scenario. Refer to Figure 2 , Figure 2 , which is a schematic diagram of the motor rotor winding provided by an embodiment of the present invention; during the winding process, the rotor rotates, and the winding probe realizes the winding of the copper wire on the motor through the up and down periodic movement.

[0049] In the embodiment of the present invention, a high-resolution industrial camera can be installed on the side of the rotor, and the winding process of the motor rotor can be continuously photographed at different sampling moments to obtain the original images. Then, the original images are grayscale processed to obtain the winding grayscale images.

[0050] In the embodiment of the present invention, the original images can be collected once every 0.1 second, and the grayscale processing can use average grayscale processing, and no restrictions are imposed on this.

[0051] It should be noted that in two adjacent frames of winding grayscale images, the rotor and the winding probe change. The rotor rotates at a high speed. At this time, if the shutter speed of the industrial camera is slow, motion blur will be caused by the rotation of the rotor. This motion blur effect will make the rotor area blurred, losing the fine features on the rotor surface and the motion details during the winding process, affecting the winding detection during the winding process. Therefore, deblurring can be performed.

[0052] In the image captured by the camera, the edge information of the rotor and the coil belongs to the detailed information and is the high-frequency component in the winding grayscale image; motion blur belongs to the interference information and is manifested as the low-frequency component, which needs to be removed. Fourier transform can convert the image to the frequency domain. The frequency domain image usually presents as a complex image, and its amplitude represents the intensity of the frequency. According to the frequency domain characteristics, the present invention designs a filter by selecting appropriate parameters (radius) to remove the interference of the low-frequency component and achieve the purpose of removing motion blur.

[0053] Further, in some embodiments of the present invention, different radii are used to select the central circle, and the high-frequency information in the central circle area is screened from the frequency spectrum image to obtain high-frequency images with different radii, including: using a straight line passing through the center of the image and parallel to both sides to cut the frequency spectrum image, and splicing the four corners of the cut frequency spectrum image to a central point position to obtain an image to be selected with low-frequency information in the center; taking the central point position of the image to be selected as the center of the circle, using different radii to determine the central circle of the image to be selected, and deleting the central circle part to obtain high-frequency images with different radii.

[0054] See Figure 3 , Figure 3 is a schematic diagram of the high-frequency image acquisition process provided by an embodiment of the present invention; the image undergoes Fourier transform to obtain the frequency spectrum image. In the frequency spectrum image, the central area represents the high-frequency part, and the four corner areas represent the low-frequency part. Therefore, by transposing the four corner areas, the low-frequency information remains in the center of the image to be selected, and then the low-frequency information can be directly selected by using the central circle and deleted to obtain the high-frequency image. That is, the high-frequency image is an image that removes the low-frequency information and retains the high-frequency information.

[0055] Among them, the edge information of the rotor and the coil in the winding grayscale image belongs to the high-frequency information due to the intense change in brightness or grayscale; the structural information of the motion blur interference changes gently and belongs to the low-frequency information.

[0056] Specifically, a coordinate system is established with the frequency domain center of the image to be selected as the coordinate origin, and a parametric equation is established according to the shape characteristics of the image to be selected:

[0057]

[0058] Among them, r represents the radius, which is also the screening parameter for removing low-frequency components, x represents the abscissa in the image coordinate system to be selected, y represents the ordinate in the image coordinate system to be selected, and t represents the angle.

[0059] According to the parametric equation, determine the cut-off frequency of the high-pass filter:

[0060]

[0061] Among them is the cut-off frequency of the high-pass filter; r represents the radius, and different pixel lengths such as r = 10, 20, 30, 40, and 50 are used to remove low-frequency information from the image to be selected.

[0062] Thus, high-frequency images under different radii are obtained. Since only when the radius is selected appropriately can low-frequency information be effectively removed and high-frequency information be retained, it is necessary to conduct a specific analysis for each high-frequency image.

[0063] S102: According to the position of the winding probe in the winding grayscale image, determine the motion stability factor, perform edge detection and image recognition on the winding grayscale image to determine the image connected domain belonging to the rotor, and determine the image stability factor according to the size of the image connected domain and the number of edge pixel points; combine the motion stability factor and the image stability factor to screen and obtain stable frame images from different frames of winding grayscale images.

[0064] Among them, the motion stability factor represents the coefficient index of the motion law corresponding to the position of the winding probe. The larger the value of the motion stability factor, the more stable the motion of the winding probe at the corresponding sampling moment; since the winding probe performs a cyclic periodic motion, when the winding probe moves to the lowest and highest ends of the winding column, due to the small interference of the winding probe in the captured image, the motion blur interference in the image can be ignored, and the overall motion characteristics of the image are more stable at this time. According to the motion law of the winding probe, calculate the motion stability factor and the image stability factor of the winding probe, and screen out stable frame images.

[0065] Further, in some embodiments of the present invention, determining the motion stability factor according to the position of the winding probe in the winding grayscale image includes: determining the positions of the winding probe in all frames of winding grayscale images, obtaining the motion trajectory of the winding probe, and obtaining two endpoints of the motion trajectory; obtaining the Euclidean distances between the position of the winding probe in any frame of winding grayscale image and the two endpoints respectively, and taking the maximum value of the Euclidean distances as the motion stability factor of the corresponding frame of winding grayscale image.

[0066] Taking the up and down movement of the winding probe as an example, mark the movement position of the winding probe in the winding grayscale image, obtain the corresponding sampling moments, form a position-time coordinate, and calculate the motion stability factor of the winding probe:

[0067]

[0068] Among them, represents the motion stability factor of the winding probe for photographing the rotor at the i-th sampling moment, represents the position of the winding probe photographed at the i-th sampling moment, represents the highest point (the first end point) of the position of the winding probe; represents the lowest point (the second end point) of the position of the winding probe, represents the maximum value function.

[0069] It can be understood that represents and the maximum value in, that is, the maximum value of the Euclidean distance from both end points. The larger the motion stability factor, the farther the winding probe is from one of the end points, which means it is closer to the other end point. Therefore, its motion state is more in line with the stable state.

[0070] Furthermore, in some embodiments of the present invention, edge detection and image recognition are performed on the winding grayscale image to determine the image connected domain belonging to the rotor, including: performing canny edge detection on the winding grayscale image to obtain edge pixel points, and the edge pixel points form an initial connected domain; performing image recognition on the initial connected domain, and identifying the initial connected domain belonging to the rotor as the image connected domain.

[0071] Among them, canny edge detection is an edge detection algorithm well-known to those skilled in the relevant art. By performing canny edge detection to obtain edge pixel points, the area surrounded by the edge pixel points is used as the initial connected domain. It should be noted that since the initial connected domain represents the connected domain information of an entire image and needs to be further screened, image recognition technology can be used to determine the initial connected domain belonging to the rotor as the image connected domain. Among them, the image recognition technology can specifically be, for example, image recognition based on a big data model, or, for example, image annotation based on an artificial intelligence algorithm, so as to label the initial connected domain belonging to the rotor as the image connected domain, and no limitation is imposed on this.

[0072] It can be understood that in a clear image, the more accurate the edge information of the rotor, the larger the connected domain area of the rotor. In the motion blur state, the blurred state of the rotor rotating at high speed will generate low-frequency information, increasing the number of edge pixel points and reducing the connected domain belonging to the rotor area. Based on this feature, another factor characterizing motion stability, that is, the image stability factor, can be determined.

[0073] Further, in some embodiments of the present invention, determining the image stability factor according to the size of the image connected region and the number of edge pixel points includes: calculating the ratio of the number of pixel points in the connected region to the number of edge pixel points obtained by edge detection as the image stability factor.

[0074] That is to say, by directly calculating the ratio of the number of pixel points in the connected region to the number of edge pixel points obtained by edge detection, the image stability factor is obtained. The image stability factor is positively correlated with the number of pixel points in the connected region and negatively correlated with the number of edge pixel points obtained by edge detection. The larger the value of the ratio of the number of pixel points in the connected region to the number of edge pixel points obtained by edge detection, the more stable the motion state in the winding grayscale image and the smaller the motion blur interference.

[0075] It should be noted that the positive correlation relationship means that there is a co-directional change relationship between the independent variable and the dependent variable, where the larger the independent variable, the larger the dependent variable; the negative correlation relationship means that there is an inverse change relationship between the independent variable and the dependent variable, where the smaller the independent variable, the larger the dependent variable instead. The specific manifestation forms of the positive correlation relationship and the negative correlation relationship are determined by the actual application, and no special restrictions are made in this application.

[0076] Further, in some embodiments of the present invention, combining the motion stability factor and the image stability factor to screen out the stable frame images from different frames of winding grayscale images includes: calculating the product of the motion stability factor and the image stability factor, and performing maximum-minimum normalization processing to obtain the stability index; using the winding grayscale image with the stability index greater than the preset stability threshold as the stable frame image.

[0077] Combined with the above analysis, since the larger the motion stability factor, it can be explained that the farther the winding probe is from one of the endpoints, which means it is closer to the other endpoint. Therefore, its motion state is more in line with the stable state; and the larger the value of the image stability factor, the more stable the motion state in the winding grayscale image and the smaller the motion blur interference. Therefore, the product of the motion stability factor and the image stability factor can be directly calculated and normalized by maximum-minimum to obtain the stability index.

[0078] Among them, the preset stability threshold is the threshold value of the stability index. In the embodiments of the present invention, the preset stability threshold can be specifically, for example, 0.75. That is to say, the winding grayscale image with the stability index greater than 0.75 is used as the stable frame image.

[0079] S103: For high-frequency images of any radius, perform gradient analysis on the stable frame image and the high-frequency image of the same winding grayscale image respectively. Determine the gradient distribution histogram of the stable frame image as the first histogram, and determine the gradient distribution histogram of the high-frequency image as the second histogram. Determine the similarity parameter between the first histogram and the second histogram according to the difference in the gradient distribution structure between the first histogram and the second histogram.

[0080] The stable frame image represents the winding grayscale image with less influence from motion obtained through dynamic blur analysis. That is, when the winding grayscale image has a stable frame image, it can preliminarily indicate that it has stable characteristics. Therefore, by combining the gradient analysis in the frequency domain and the spatial domain, the similarity situation can be determined. The higher the similarity, the more corresponding the high-frequency information is to the spatial domain gradient, that is, the stronger the stability.

[0081] Use the sobel operator to obtain the gradient magnitude of each pixel point in the stable frame image and the high-frequency image respectively, and generate a normalized gradient distribution histogram. The abscissa in the figure represents the gradient intensity, and the ordinate represents the frequency.

[0082] When the high-frequency image during the rotor winding process is deblurred from motion (that is, the low-frequency interference information is removed), it should be similar to the stable frame image in terms of edge information, image texture, and local features. Then, there will be several peaks with similar heights and positions in the gradient distribution histogram of the image. By calculating the similarity in the performance between the gradient distribution histograms of the stable frame image and the high-frequency image, the structural difference between the two gradient distribution histograms is obtained, and further the deblurring effect of the high-frequency image is judged.

[0083] Further, in some embodiments of the present invention, determining the similarity parameter between the first histogram and the second histogram according to the difference in the gradient distribution structure between the first histogram and the second histogram includes: dividing the gradient intensities in the first histogram and the second histogram into the same intervals to obtain the corresponding gradient intervals in the first histogram and the second histogram; calculating the squared value of the difference in the frequencies under the same gradient intervals in the first histogram and the second histogram to obtain the first index; calculating the sum value of the frequencies under the same gradient intervals in the first histogram and the second histogram as the second index; calculating the ratio of the first index to the second index, and performing maximum-minimum normalization processing on the reciprocal of this ratio as the interval similarity coefficient for the corresponding gradient interval; taking the mean value of the interval similarity coefficients of all gradient intervals as the similarity parameter between the first histogram and the second histogram.

[0084] Divide the gradient intensities in the two gradient distribution histograms into the same number of intervals. For example, the gradient intensity of 1 - 3 is taken as one interval, the gradient intensity of 4 - 6 is taken as one interval, and so on. Thus, the corresponding gradient intervals in the first histogram and the second histogram are determined respectively.

[0085] The first histogram and the second histogram are similar, mainly because they are more similar in the same gradient interval and have less structural difference. Therefore, calculate the square value of the difference in frequencies in the same gradient interval between the first histogram and the second histogram to obtain the first index, and calculate the sum value of the frequencies in the same gradient interval between the first histogram and the second histogram as the second index.

[0086] The larger the ratio of the first index to the second index, the greater the difference in the frequency distribution within the same gradient interval, indicating that the similarity between the first histogram and the second histogram in the corresponding gradient interval is lower. Therefore, perform maximum-minimum normalization on the reciprocal of the ratio to obtain the interval similarity coefficient of the gradient interval.

[0087] In the embodiments of the present invention, since multiple gradient intervals are obtained by division, integrate all gradient intervals and calculate the mean value of all interval similarity coefficients, thereby obtaining the similarity parameter between the first histogram and the second histogram.

[0088] It should be noted that each radius corresponds to a high-frequency image, and the high-frequency image of each radius also corresponds to a similarity parameter.

[0089] S104: Screen the optimal radius according to the similarity parameter, convert the high-frequency image under the optimal radius into the spatial domain to obtain a deblurred image, and perform wire winding detection on the deblurred image to determine a wire winding gray image with wire winding anomalies.

[0090] Further, in some embodiments of the present invention, screening the optimal radius according to the similarity parameter includes: taking the radius corresponding to the maximum value of the similarity parameters corresponding to all high-frequency images as the optimal radius.

[0091] Since the larger the value of the similarity parameter, the less the corresponding high-frequency image is affected by motion blur, that is, the stronger the deblurring effect of the corresponding high-frequency image, therefore, take the radius corresponding to the high-frequency image with the maximum value of the similarity parameter as the optimal radius.

[0092] Among them, converting the high-frequency image under the optimal radius into the spatial domain to obtain a deblurred image, the specific conversion method is inverse Fourier transform, which is a well-known technique to those skilled in the art and will not be elaborated here.

[0093] It should be noted that deblurring processing can be combined with the high-frequency image corresponding to the optimal radius, so as to further screen out a clearer and more reliable deblurred image. That is, under the optimal radius, the motion blur effect can be effectively removed to obtain a clearer and more reliable deblurred image.

[0094] Further, in some embodiments of the present invention, a wire winding detection is performed on the deblurred image to determine a wire winding grayscale image with wire winding anomalies, including: inputting the deblurred image into a pre-trained wire winding detection model to output a wire winding anomaly inspection result of the wire winding grayscale image.

[0095] The purpose of this solution is to determine a clearer and more reliable deblurred image. Therefore, a wire winding detection can be performed on the deblurred image, specifically for wire winding anomalies such as wire misalignment anomalies and uneven winding of coils. The wire winding detection model can be trained with multiple pre-annotated images to obtain a more reliable pre-trained wire winding detection model, and the deblurred image is input into the pre-trained wire winding detection model to output a wire winding anomaly inspection result of the wire winding grayscale image.

[0096] Of course, in some other embodiments of the present invention, a wire winding detection can also be achieved by means of image analysis, and there is no limitation in this regard.

[0097] The present invention selects a central circle through a spectral image, removes low-frequency information affected by motion blur, retains high-frequency information, and obtains high-frequency images with different radii; then, a stable frame image is determined according to features such as the position of the wire winding probe, the number of edge pixels of the rotor, and the area size; gradient analysis is performed on the stable frame image and the high-frequency image, the similarity parameter of the gradient feature is determined through a gradient distribution histogram, the optimal radius is selected according to the similarity parameter, the high-frequency image under the optimal radius is converted into the spatial domain to obtain a deblurred image, and a wire winding detection is performed on the deblurred image to determine a wire winding grayscale image with wire winding anomalies. The present invention can effectively analyze the wire winding process of a rotating rotor, through the analysis of the characteristics of the image acquisition device itself and the dynamic blur characteristics generated by the rotation of the rotor, so as to accurately select a stable frame image and perform a deblurring process using an adaptive and appropriate scale to obtain a clearer and more reliable deblurred image. By performing a wire winding detection on the deblurred image, the interference caused by dynamic blur is avoided, and the accuracy of the wire winding detection of the propulsion motor rotor is improved.

[0098] The present invention also provides a wire winding system for a propulsion motor rotor. Refer to Figure 4 , Figure 4 which is a structural diagram of a wire winding system for a propulsion motor rotor provided by an embodiment of the present invention. The system 600 includes a memory 601, a processor 602, and a computer program 603 stored in the memory 601 and executable on the processor 602. When the processor 602 executes the computer program 603, the steps of the aforementioned wire winding method for a propulsion motor rotor are implemented.

[0099] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0100] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A propulsion motor rotor winding method, characterized in that: The method comprises: At different sampling times, a winding grayscale image of the motor side winding probe during the rotor winding process is obtained, the winding grayscale image is converted into a spectrum image in the frequency domain, different radii are used to select the center circle, and high-frequency information of the center circle area is filtered from the spectrum image to obtain high-frequency images at different radii; According to the position of the winding probe in the winding grayscale image, a motion stability factor is determined, edge detection and image recognition are performed on the winding grayscale image to determine the image connected domain belonging to the rotor, and an image stability factor is determined according to the size of the image connected domain and the number of edge pixels; the motion stability factor and the image stability factor are combined to screen out stable frame images from different frames of winding grayscale images; For a high-frequency image of any radius, a gradient analysis is performed on the stable frame image and the high-frequency image of the same winding grayscale image, and the gradient distribution histogram of the stable frame image is determined as the first histogram, and the gradient distribution histogram of the high-frequency image is determined as the second histogram. According to the difference in the gradient distribution structures between the first histogram and the second histogram, the similarity parameters of the first histogram and the second histogram are determined; An optimal radius is obtained by screening according to the similarity parameters, a high-frequency image under the optimal radius is converted into a spatial domain to obtain a deblurred image, winding detection is performed on the deblurred image, and a winding grayscale image with winding abnormalities is determined; The method of selecting the central circle using different radii, filtering high-frequency information of the central circle area from the spectrum image, and obtaining high-frequency images under different radii includes: The spectrum image is cut using a straight line passing through the center of the image and parallel to both sides, and the four corners of the spectrum image after cutting are assembled to a central point to obtain an image to be selected with low-frequency information in the center; Taking the center point of the image to be selected as the center of the circle, different radii are used to determine the center circle of the image to be selected, and the center circle part is deleted to obtain high-frequency images under different radii.

2. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: The step of determining the motion stability factor according to the position of the wire winding probe in the wire winding grayscale image comprises: Determine the position of the wire winding probe in all frame wire winding grayscale images, obtain the motion trajectory of the wire winding probe, and obtain two endpoints of the motion trajectory; The Euclidean distances between the position of the winding probe and the two end points in any frame of the winding grayscale image are obtained, and the maximum value of the Euclidean distances is used as the motion stability factor of the corresponding frame of the winding grayscale image.

3. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: The performing edge detection and image recognition on the winding grayscale image to determine the image connected domain belonging to the rotor includes: Performing canny edge detection on the winding grayscale image to obtain edge pixel points, wherein the edge pixel points constitute an initial connected domain; Image recognition is performed on the initial connected domain, and the initial connected domain belonging to the rotor is identified as the image connected domain.

4. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: The determining of the image stabilization factor according to the size of the image connected domain and the number of edge pixels comprises: The ratio of the number of pixels in the connected domain to the number of edge pixels obtained by edge detection is calculated as an image stabilization factor.

5. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: The method combines the motion stabilization factor and the image stabilization factor to screen out the stable frame images from the winding grayscale images of different frames, including: Calculating the product of the motion stability factor and the image stability factor, and normalizing the maximum and minimum values ​​to obtain a stability index; The winding grayscale image whose stability index is greater than a preset stability threshold is used as a stable frame image.

6. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: Determining similar parameters of the first histogram and the second histogram according to the difference in the gradient distribution structures of the first histogram and the second histogram includes: Dividing the gradient strengths in the first histogram and the second histogram into the same intervals to obtain corresponding gradient intervals in the first histogram and the second histogram; Calculate the square value of the difference between the frequencies in the same gradient interval in the first histogram and the second histogram to obtain a first index; Calculate the sum of the frequencies in the same gradient interval in the first histogram and the second histogram as a second indicator; Calculate the ratio of the first index to the second index, and perform maximum and minimum normalization processing on the inverse of the ratio as the interval similarity coefficient of the corresponding gradient interval; The mean of the interval similarity coefficients of all gradient intervals is taken as the similarity parameter of the first histogram and the second histogram.

7. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: The optimal radius is obtained by screening according to the similar parameters, including: The radius at the maximum value of similar parameters corresponding to all high-frequency images is taken as the optimal radius.

8. A propulsion motor rotor winding method as claimed in claim 1, characterized in that: Performing winding detection on the deblurred image to determine a winding grayscale image with abnormal winding includes: The deblurred image is input into a pre-trained winding detection model, and a winding abnormality inspection result of the winding grayscale image is output.

9. A propulsion motor rotor winding system, the system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of a propulsion motor rotor winding method as described in any one of claims 1 to 8 are implemented.

Images