Visual monitoring system and method based on rotor magnetic pole

Abstract

The invention relates to the technical field of visual monitoring, and discloses a visual monitoring system and method based on a rotor magnetic pole, and the system comprises an image processing module which carries out the image processing of a rotor magnetic pole edge feature image sequence, and extracts a current frame magnetic pole contour feature vector based on a multi-scale edge detection algorithm; the visual monitoring module is used for generating a magnetic pole deformation parameter according to a curvature error of the current feature vector and the reference feature vector, extracting mass center offset time sequence data of continuous multi-frame feature vectors, generating a magnetic pole displacement parameter by combining the mass center offset time sequence data with a rotor rotation angular velocity value, identifying a surface defect in the current feature vector, and generating a surface defect parameter; and the monitoring evaluation module is used for carrying out weighted normalization on the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter to obtain a visual monitoring evaluation result. According to the invention, the precision and anti-interference capability of rotor magnetic pole state monitoring are significantly improved, precise identification of tiny changes of the magnetic pole can be realized, and high-precision and high-reliability real-time monitoring requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of visual monitoring, and in particular to a visual monitoring system and method based on rotor magnetic poles. Background Art

[0002] In modern motor systems, the state of the rotor magnetic poles directly affects the operating efficiency and reliability of the motor. Traditional magnetic pole monitoring methods usually rely on contact sensors or manual detection. These methods have the problems of insufficient accuracy, poor real-time performance, high cost and limited adaptability to complex working conditions. With the rapid development of industrial automation and intelligence, the demand for real-time monitoring of the rotor magnetic pole state is increasing. In the existing technology, although some systems use non-contact measurement methods, their monitoring accuracy is still greatly affected by environmental interference, and it is difficult to achieve accurate identification of small changes in the magnetic poles. In addition, the traditional method is not stable enough under complex working conditions (such as high temperature, high speed, and strong electromagnetic interference), and it is difficult to meet the requirements of high-precision and high-reliability monitoring.

[0003] Therefore, there is an urgent need to provide a rotor magnetic pole-based visual monitoring system and method to solve the above problems. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the above-mentioned defects of the prior art and provide a visual monitoring system and method based on rotor poles that significantly improves the accuracy and anti-interference ability of rotor pole state monitoring, can accurately identify tiny changes in the poles, and meet the real-time monitoring needs of high precision and high reliability.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows: a visual monitoring system based on rotor magnetic poles, comprising:

[0006] The image processing module is used to perform frame-by-frame decomposition and spatiotemporal registration of a dynamic image sequence containing rotor magnetic pole edge features, eliminate image distortion caused by high-speed rotor rotation and environmental electromagnetic interference, and extract the magnetic pole contour feature vector of the current frame based on a multi-scale edge detection algorithm;

[0007] The visual monitoring module is used to generate magnetic pole deformation parameters based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment, extract the center of mass offset time series data of multiple consecutive frames of magnetic pole profile feature vectors, combine the rotor rotation angular velocity feedback value to generate magnetic pole displacement parameters, and identify the geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector to generate surface defect parameters; wherein, the calculation formula of the magnetic pole deformation parameter is: The calculation formula of magnetic pole displacement parameter is: The calculation formula of surface defect parameters is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base，i represents the curvature of the i-th sampling point of the reference magnetic pole profile feature vector; C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the t-th frame, r represents the rotor magnetic pole radius, f represents the camera frame rate; D represents the surface defect parameter, Ω represents the pixel point set in the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the j-th pixel in the pixel set;

[0008] The monitoring and evaluation module is used to perform weighted normalization on the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter, and output a rotor magnetic pole visual monitoring and evaluation result.

[0009] Preferably, the rotor pole-based visual monitoring system further comprises:

[0010] The image acquisition module is used to use a high-speed industrial camera arranged outside the rotor rotation plane and maintaining a preset inclination angle with the rotor rotation axis to capture the dynamic image sequence containing the rotor pole edge features in real time.

[0011] Preferably, the image processing module is specifically used to:

[0012] According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated; wherein, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity, r is the rotor pole radius, f is the camera focal length, d is the object distance, and Δt is the frame interval time;

[0013] When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation;

[0014] A fast Fourier transform is performed on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain. An adaptive notch filter is used to attenuate the energy of the characteristic frequency band, retaining the main frequency component of the magnetic pole contour, and obtaining a dynamic image sequence after dynamic compensation.

[0015] Preferably, the image processing module is specifically used to:

[0016] Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge;

[0017] For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points.

[0018] The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour;

[0019] Equal-angle key point sampling is performed along the fused magnetic pole contour, and the curvature, normal direction and local gradient amplitude of each point are calculated to form the current frame magnetic pole contour feature vector with unified dimension.

[0020] Preferably, the monitoring and evaluation module is specifically used to:

[0021] The weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are dynamically adjusted based on the rotor speed and weighted normalization is performed to output the rotor magnetic pole visual monitoring evaluation result.

[0022] The present invention further solves the technical problem by adopting the following technical solution: a visual monitoring method based on rotor magnetic poles, comprising:

[0023] The dynamic image sequence containing the rotor magnetic pole edge features is decomposed frame by frame and aligned in the spatiotemporal domain to eliminate image distortion caused by the high-speed rotation of the rotor and environmental electromagnetic interference. The magnetic pole contour feature vector of the current frame is extracted based on a multi-scale edge detection algorithm.

[0024] The magnetic pole deformation parameters are generated based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment. The centroid offset time series data of the continuous multi-frame magnetic pole profile feature vectors are extracted and combined with the rotor rotation angular velocity feedback value to generate the magnetic pole displacement parameters. The geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector is identified to generate the surface defect parameters. The calculation formula of the magnetic pole deformation parameters is as follows: The calculation formula of magnetic pole displacement parameter is: The calculation formula of surface defect parameters is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base，irepresents the curvature of the i-th sampling point of the reference magnetic pole profile feature vector; C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the t-th frame, r represents the rotor magnetic pole radius, f represents the camera frame rate; D represents the surface defect parameter, Ω represents the pixel point set in the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the j-th pixel in the pixel set;

[0025] The magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are weighted and normalized, and a rotor magnetic pole visual monitoring evaluation result is output.

[0026] Preferably, the visual monitoring method based on the rotor magnetic poles further comprises:

[0027] The dynamic image sequence containing the rotor magnetic pole edge features is collected in real time using a high-speed industrial camera that is arranged outside the rotor rotation plane and maintains a preset inclination angle with the rotor rotation axis.

[0028] Preferably, the step of performing frame-by-frame decomposition and spatiotemporal registration on a dynamic image sequence containing rotor magnetic pole edge features to eliminate image distortion caused by high-speed rotation of the rotor and environmental electromagnetic interference includes:

[0029] According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated; wherein, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity, r is the rotor pole radius, f is the camera focal length, d is the object distance, and Δt is the frame interval time;

[0030] When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation;

[0031] A fast Fourier transform is performed on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain. An adaptive notch filter is used to attenuate the energy of the characteristic frequency band, retaining the main frequency component of the magnetic pole contour, and obtaining a dynamic image sequence after dynamic compensation.

[0032] Preferably, the step of extracting the magnetic pole contour feature vector of the current frame based on a multi-scale edge detection algorithm includes:

[0033] Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge;

[0034] For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points.

[0035] The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour;

[0036] Equal-angle key point sampling is performed along the fused magnetic pole contour, and the curvature, normal direction and local gradient amplitude of each point are calculated to form the current frame magnetic pole contour feature vector with unified dimension.

[0037] Preferably, the step of weighted normalization of the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter and outputting the rotor magnetic pole visual monitoring evaluation result comprises:

[0038] The weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are dynamically adjusted based on the rotor speed and weighted normalization is performed to output the rotor magnetic pole visual monitoring evaluation result.

[0039] The present invention effectively eliminates image distortion caused by high-speed rotor rotation and environmental electromagnetic interference through frame-by-frame decomposition and spatiotemporal registration of dynamic image sequences. A multi-scale edge detection algorithm is used to precisely extract magnetic pole contour feature vectors. Combined with curvature matching error and center of mass offset time series data, magnetic pole deformation, displacement, and surface defect parameters are generated in real time, and a comprehensive evaluation result is output through weighted normalization. This invention significantly improves the accuracy and anti-interference capability of rotor pole state monitoring, particularly under complex operating conditions (such as high temperature, high speed, and strong electromagnetic interference). It can accurately identify subtle changes in the magnetic poles, meeting the needs of high-precision, high-reliability real-time monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 1 is a schematic structural diagram of Embodiment 1 of the visual monitoring system based on rotor magnetic poles of the present invention;

[0041] Figure 2 It is a flow chart of Example 1 of the visual monitoring method based on rotor magnetic poles of the present invention. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the embodiments and drawings.

[0043] Example 1 of a visual monitoring system based on rotor magnetic poles:

[0044] like Figure 1 As shown, an embodiment of the present invention provides a rotor magnetic pole-based visual monitoring system 100 , including: an image processing module 110 , a visual monitoring module 120 and a monitoring and evaluation module 130 .

[0045] The image processing module 110 is used to perform frame-by-frame decomposition and spatiotemporal registration of a dynamic image sequence containing rotor magnetic pole edge features, eliminate image distortion caused by high-speed rotor rotation and environmental electromagnetic interference, and extract the magnetic pole contour feature vector of the current frame based on a multi-scale edge detection algorithm;

[0046] The dynamic image sequence includes multiple frames of images containing the edge features of the rotor magnetic poles collected in real time. Spatiotemporal registration refers to the elimination of motion blur caused by high-speed rotor rotation and image distortion caused by electromagnetic interference by combining the time series (motion compensation between adjacent frames) with the spatial domain (distortion correction of single-frame images). Spatiotemporal registration specifically includes the coordinated operation of pixel displacement correction (time domain) based on rotational angular velocity feedback and frequency domain filtering (spatial domain);

[0047] The multi-scale edge detection algorithm generates a multi-scale image set through Gaussian kernel convolution with different standard deviations. It then combines morphological gradient operations with an edge fusion mechanism to extract sub-pixel-level magnetic pole contour feature vectors. The multi-scale edge detection algorithm includes multi-scale Gaussian filtering, adaptive threshold segmentation, morphological gradient operations, multi-scale edge fusion, and feature vectorization.

[0048] The visual monitoring module 120 is configured to generate magnetic pole deformation parameters based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment, extract the center of mass offset time series data of multiple consecutive frames of magnetic pole profile feature vectors, combine the rotor rotation angular velocity feedback value to generate magnetic pole displacement parameters, and identify the geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector to generate surface defect parameters;

[0049] The magnetic pole deformation parameter is a numerical indicator that quantifies the degree of bending, distortion, or local expansion / concavity of the magnetic pole structure by comparing the geometric curvature difference between the real-time magnetic pole profile and the reference profile. The calculation formula for the magnetic pole deformation parameter is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base，iThe curvature of the i-th sampling point of the reference magnetic pole profile feature vector is represented by the magnetic pole deformation parameter E, which is expressed in millimeters (mm) with an error resolution of ±0.01 mm.

[0050] The magnetic pole displacement parameter refers to the overall position offset of the magnetic pole in the radial or axial direction due to mechanical looseness, bearing wear, etc., and is a dynamic compensation value based on the center of mass offset time series data and rotor rotation angular velocity feedback. The calculation formula for the magnetic pole displacement parameter is: C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the tth frame, r represents the rotor magnetic pole radius, and f represents the camera frame rate; the unit of the magnetic pole displacement parameter Δd is millimeter (mm), and the resolution reaches 0.01mm at a rotation speed of 10000rpm;

[0051] The surface defect parameter refers to the quantitative evaluation value of local geometric abnormal areas such as cracks, pits, and spalling on the magnetic pole surface caused by fatigue, corrosion, or mechanical damage. The surface defect parameter is calculated by morphological gradient residual analysis and is defined as: D represents the surface defect parameter, Ω represents the pixel set of the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the jth pixel in the pixel set. The surface defect parameter D is expressed as the defect area ratio (%) or the maximum defect depth (mm). The minimum detectable defect size is 0.1mm×0.1mm;

[0052] It should be noted that the step of identifying the geometric abnormal area of the surface defect in the magnetic pole contour feature vector of the current frame includes: in a defect-free state, extracting the reference magnetic pole contour feature vector through a multi-scale edge detection algorithm, and calculating its morphological gradient field as a reference for the ideal smooth contour. The gradient field contains the normal direction gradient amplitude and curvature continuity information of each sampling point. Morphological gradient operation is performed on the magnetic pole contour feature vector of the current frame, and 3×3 circular structure elements are used to perform dilation and erosion operations on the binary image to generate a real-time gradient field. The gradient residual is calculated pixel by pixel to generate a residual map. Based on the residual histogram distribution, a threshold is dynamically set and areas with a value greater than the dynamic threshold are marked as candidate abnormal areas. The candidate abnormal areas are then filtered for geometric features to eliminate noise interference and obtain geometric abnormal areas.

[0053] A monitoring and evaluation module 130 is configured to perform weighted normalization on the magnetic pole deformation parameter, the magnetic pole displacement parameter, and the surface defect parameter, and output a rotor magnetic pole visual monitoring and evaluation result;

[0054] Among them, weighted normalization processing refers to assigning different weight values (such as 0.4-0.5 for deformation parameters, 0.3-0.35 for displacement parameters, and 0.15-0.25 for defect parameters) according to the degree of influence of magnetic pole deformation parameters, magnetic pole displacement parameters and surface defect parameters on system reliability, and generating rotor magnetic pole visual monitoring assessment results through linear weighting and extreme value normalization;

[0055] The rotor magnetic pole visual monitoring assessment result is a comprehensive status score of the magnetic pole deformation parameters, magnetic pole displacement parameters, and surface defect parameters, which is used to determine the magnetic pole health status level (such as normal, warning, and fault). A rotor magnetic pole visual monitoring assessment result of less than or equal to 0.3 is considered normal, 0.3 to 0.6 is a warning, and greater than or equal to 0.6 is a fault.

[0056] Embodiment 2 of a visual monitoring system based on rotor magnetic poles, based on the above embodiment 1, further includes:

[0057] An image acquisition module is configured to acquire the dynamic image sequence containing the rotor magnetic pole edge features in real time using a high-speed industrial camera disposed outside the rotor rotation plane and maintaining a preset inclination angle with the rotor rotation axis;

[0058] Among them, the preset inclination angle defaults to 20°~45°. In this embodiment, the installation angle of the high-speed industrial camera is adjusted by a laser calibration device so that the edge feature of the magnetic pole is in the center area of the camera's field of view. The high-speed industrial camera includes: a global shutter sensor (using a CCD or CMOS global shutter chip, an exposure time ≤100μs, a frame rate ≥5000fps), an anti-electromagnetic interference optical lens (the lens is coated with an ITO conductive film, the inner wall of the lens barrel is provided with a Permalloy shielding layer, and filled with absorbing material) and a synchronous trigger unit (receives the pulse signal of the rotor encoder and controls the camera exposure timing according to the trigger frequency).

[0059] Embodiment 3 of a visual monitoring system based on rotor magnetic poles, based on the above embodiment 1, the image processing module 110 is specifically used for:

[0060] According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated;

[0061] Among them, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity (rad / s), r is the rotor pole radius (mm), f is the camera focal length (mm), d is the object distance (mm), and Δt is the frame interval time (s);

[0062] When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation;

[0063] The image is transformed from the Cartesian coordinate system to the polar coordinate system with the rotor rotation center as the origin. According to the theoretical pixel displacement Δp, the full image displacement field D(x, y) is constructed; for the current frame image I t , perform inverse deformation according to the full image displacement field to generate the compensated image I' t For non-integer pixel positions, bilinear interpolation is used to calculate the grayscale value. The SIFT feature points of the two compensated frames are extracted and the matching error is calculated. If the error is greater than 0.5 pixels, the displacement is recalibrated. The residual image of the two frames is calculated and the proportion of non-zero pixels is counted. If it is greater than 5%, it is determined that the registration has failed and re-compensation is required.

[0064] Performing a fast Fourier transform on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain, and using an adaptive notch filter to attenuate the energy of the characteristic frequency band to retain the main frequency component of the magnetic pole contour to obtain a dynamic image sequence after dynamic compensation;

[0065] Among them, the adaptive notch filter automatically identifies the characteristic frequency band (center frequency f c ±Δf), and suppress the energy of the characteristic frequency band through Gaussian attenuation function, with an attenuation amplitude of ≥20dB.

[0066] Embodiment 4 of a visual monitoring system based on rotor magnetic poles, based on the above embodiment 3, the image processing module 110 is specifically used for:

[0067] Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge;

[0068] For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points.

[0069] The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour;

[0070] Sampling key points at equal angles along the fused magnetic pole contour, calculating the curvature, normal direction, and local gradient amplitude of each point, and forming a dimensional unified feature vector of the magnetic pole contour of the current frame;

[0071] The Gaussian kernel standard deviation σ1 of the coarse-grained scale feature is 1.5 to 2.5 pixels, which is used to suppress high-frequency noise; the Gaussian kernel standard deviation σ2 of the fine-grained scale feature is 0.5 to 1.0 pixels, which is used to preserve sub-pixel edge details.

[0072] In the process of sampling equal-angle key points, key points are sampled at equal-angle intervals of 5.625° (64 points) or 2.8125° (128 points) along the magnetic pole contour to ensure that the feature vector contains the full-circumferential magnetic pole state information.

[0073] Embodiment 5 of a visual monitoring system based on rotor magnetic poles, based on the above embodiment 4, the monitoring and evaluation module 130 is specifically used for:

[0074] Dynamically adjusting the weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter, and the surface defect parameter based on the rotor speed and performing weighted normalization, and outputting the rotor magnetic pole visual monitoring evaluation result;

[0075] The dynamic weight adjustment rule is as follows: when the rotor speed n ≥ 8000 rpm, the weight of the magnetic pole displacement parameter increases to 0.35-0.4, and the weight of the magnetic pole deformation parameter decreases to 0.35-0.4. When surface defects are detected, the weight of the surface defect parameter increases to 0.25-0.3. The weight values are linked to the speed and number of defects through a piecewise linear function.

[0076] This embodiment of the present invention effectively eliminates image distortion caused by high-speed rotor rotation and environmental electromagnetic interference through frame-by-frame decomposition and spatiotemporal registration of dynamic image sequences. It utilizes a multi-scale edge detection algorithm to accurately extract magnetic pole contour feature vectors. Combined with curvature matching error and center of mass offset time series data, it generates magnetic pole deformation, displacement, and surface defect parameters in real time, and outputs comprehensive evaluation results through weighted normalization. This embodiment significantly improves the accuracy and anti-interference capability of rotor pole state monitoring, especially under complex operating conditions (such as high temperature, high speed, and strong electromagnetic interference). It can accurately identify subtle changes in the magnetic poles, meeting the needs of high-precision, high-reliability real-time monitoring.

[0077] A visual monitoring method based on rotor magnetic poles Example 1:

[0078] like Figure 2 As shown, a visual monitoring method based on rotor magnetic poles includes the following steps:

[0079] S1. Decompose the dynamic image sequence containing the rotor magnetic pole edge features frame by frame and align them in the spatiotemporal domain to eliminate image distortion caused by the high-speed rotation of the rotor and environmental electromagnetic interference, and extract the magnetic pole contour feature vector of the current frame based on a multi-scale edge detection algorithm;

[0080] S2. Generate magnetic pole deformation parameters based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment, extract the center of mass offset time series data of multiple consecutive frames of magnetic pole profile feature vectors, combine them with the rotor rotation angular velocity feedback value, generate magnetic pole displacement parameters, and identify the geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector to generate surface defect parameters; wherein, the calculation formula of the magnetic pole deformation parameters is: The calculation formula of magnetic pole displacement parameter is: The calculation formula of surface defect parameters is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base，i represents the curvature of the i-th sampling point of the reference magnetic pole profile feature vector; C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the t-th frame, r represents the rotor magnetic pole radius, f represents the camera frame rate; D represents the surface defect parameter, Ω represents the pixel point set in the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the j-th pixel in the pixel set;

[0081] S3. Perform weighted normalization on the magnetic pole deformation parameter, the magnetic pole displacement parameter, and the surface defect parameter, and output a rotor magnetic pole visual monitoring evaluation result.

[0082] A second embodiment of a method for visually monitoring rotor magnetic poles, based on the above-mentioned first embodiment, further includes:

[0083] The dynamic image sequence containing the rotor magnetic pole edge features is collected in real time using a high-speed industrial camera that is arranged outside the rotor rotation plane and maintains a preset inclination angle with the rotor rotation axis.

[0084] A third embodiment of a rotor magnetic pole-based visual monitoring method, based on the above-mentioned first embodiment, performs frame-by-frame decomposition and spatiotemporal registration on a dynamic image sequence containing rotor magnetic pole edge features to eliminate image distortion caused by high-speed rotor rotation and environmental electromagnetic interference, including the following steps:

[0085] According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated; wherein, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity, r is the rotor pole radius, f is the camera focal length, d is the object distance, and Δt is the frame interval time;

[0086] When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation;

[0087] A fast Fourier transform is performed on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain. An adaptive notch filter is used to attenuate the energy of the characteristic frequency band, retaining the main frequency component of the magnetic pole contour, and obtaining a dynamic image sequence after dynamic compensation.

[0088] A fourth embodiment of a method for visual monitoring of rotor magnetic poles, based on the above-mentioned third embodiment, includes the following steps of extracting a magnetic pole contour feature vector of a current frame based on a multi-scale edge detection algorithm:

[0089] Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge;

[0090] For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points.

[0091] The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour;

[0092] Equal-angle key point sampling is performed along the fused magnetic pole contour, and the curvature, normal direction and local gradient amplitude of each point are calculated to form the current frame magnetic pole contour feature vector with unified dimension.

[0093] A fifth embodiment of a method for visual monitoring of rotor magnetic poles, based on the above-mentioned fourth embodiment, includes the steps of weighted normalization of the magnetic pole deformation parameter, the magnetic pole displacement parameter, and the surface defect parameter, and outputting a rotor magnetic pole visual monitoring evaluation result, including:

[0094] The weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are dynamically adjusted based on the rotor speed and weighted normalization is performed to output the rotor magnetic pole visual monitoring evaluation result.

[0095] This embodiment of the present invention effectively eliminates image distortion caused by high-speed rotor rotation and environmental electromagnetic interference through frame-by-frame decomposition and spatiotemporal registration of dynamic image sequences. It utilizes a multi-scale edge detection algorithm to accurately extract magnetic pole contour feature vectors. Combined with curvature matching error and center of mass offset time series data, it generates magnetic pole deformation, displacement, and surface defect parameters in real time, and outputs comprehensive evaluation results through weighted normalization. This embodiment significantly improves the accuracy and anti-interference capability of rotor pole state monitoring, especially under complex operating conditions (such as high temperature, high speed, and strong electromagnetic interference). It can accurately identify subtle changes in the magnetic poles, meeting the needs of high-precision, high-reliability real-time monitoring.

[0096] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, any of the above-mentioned visual monitoring methods based on rotor magnetic poles is implemented. That is to say, an electronic device according to an embodiment of the present invention may include but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the visual monitoring method based on rotor magnetic poles shown in any embodiment of the present invention by calling the computer program.

[0097] A computer-readable storage medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements any of the above-mentioned visual monitoring methods based on rotor magnetic poles.

[0098] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0099] In an exemplary embodiment, a computer program product or computer program is also provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the above-described rotor magnetic pole-based visual monitoring method.

[0100] The computer-readable storage medium provided in the embodiments of the present invention may be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or component.

[0101] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

Claims

1. A visual monitoring system based on rotor magnetic poles, characterized in that: include: The image processing module is used to perform frame-by-frame decomposition and spatiotemporal registration of a dynamic image sequence containing rotor magnetic pole edge features, eliminate image distortion caused by high-speed rotor rotation and environmental electromagnetic interference, and extract the magnetic pole contour feature vector of the current frame based on a multi-scale edge detection algorithm; The visual monitoring module is used to generate magnetic pole deformation parameters based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment, extract the center of mass offset time series data of multiple consecutive frames of magnetic pole profile feature vectors, combine the rotor rotation angular velocity feedback value to generate magnetic pole displacement parameters, and identify the geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector to generate surface defect parameters; wherein, the calculation formula of the magnetic pole deformation parameter is: The calculation formula of magnetic pole displacement parameter is: The calculation formula of surface defect parameters is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base,i represents the curvature of the i-th sampling point of the reference magnetic pole profile feature vector; C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the t-th frame, r represents the rotor magnetic pole radius, f represents the camera frame rate; D represents the surface defect parameter, Ω represents the pixel point set in the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the j-th pixel in the pixel set; The monitoring and evaluation module is used to perform weighted normalization on the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter, and output a rotor magnetic pole visual monitoring and evaluation result.

2. The rotor magnetic pole-based visual monitoring system according to claim 1, characterized in that: Also includes: The image acquisition module is used to use a high-speed industrial camera arranged outside the rotor rotation plane and maintaining a preset inclination angle with the rotor rotation axis to capture the dynamic image sequence containing the rotor pole edge features in real time.

3. The rotor magnetic pole-based visual monitoring system according to claim 1, characterized in that: The image processing module is specifically used for: According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated; wherein, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity, r is the rotor pole radius, f is the camera focal length, d is the object distance, and Δt is the frame interval time; When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation; A fast Fourier transform is performed on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain. An adaptive notch filter is used to attenuate the energy of the characteristic frequency band, retaining the main frequency component of the magnetic pole contour, and obtaining a dynamic image sequence after dynamic compensation.

4. The rotor magnetic pole-based visual monitoring system according to claim 3, characterized in that: The image processing module is specifically used for: Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge; For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points. The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour; Equal-angle key point sampling is performed along the fused magnetic pole contour, and the curvature, normal direction and local gradient amplitude of each point are calculated to form the current frame magnetic pole contour feature vector with unified dimension.

5. The rotor magnetic pole-based visual monitoring system according to any one of claims 1 to 4, characterized in that: The monitoring and evaluation module is specifically used for: The weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are dynamically adjusted based on the rotor speed and weighted normalization is performed to output the rotor magnetic pole visual monitoring evaluation result.

6. A visual monitoring method based on rotor magnetic poles, characterized in that: include: The dynamic image sequence containing the rotor magnetic pole edge features is decomposed frame by frame and aligned in the spatiotemporal domain to eliminate image distortion caused by the high-speed rotation of the rotor and environmental electromagnetic interference. The magnetic pole contour feature vector of the current frame is extracted based on a multi-scale edge detection algorithm. The magnetic pole deformation parameters are generated based on the curvature matching error between the current frame magnetic pole profile feature vector and the reference magnetic pole profile feature vector after spatial domain alignment. The centroid offset time series data of the continuous multi-frame magnetic pole profile feature vectors are extracted and combined with the rotor rotation angular velocity feedback value to generate the magnetic pole displacement parameters. The geometric abnormal area of the surface defect in the current frame magnetic pole profile feature vector is identified to generate the surface defect parameters. The calculation formula of the magnetic pole deformation parameters is as follows: The calculation formula of magnetic pole displacement parameter is: The calculation formula of surface defect parameters is: E represents the magnetic pole deformation parameter, N represents the total number of sampling points, k i represents the curvature of the i-th sampling point of the magnetic pole contour feature vector of the current frame, k base,i represents the curvature of the i-th sampling point of the reference magnetic pole profile feature vector; C t represents the centroid coordinates of the magnetic pole contour feature vector of the tth frame, C base represents the coordinates of the reference center of mass, w t represents the rotor rotation angular velocity feedback value corresponding to the t-th frame, r represents the rotor magnetic pole radius, f represents the camera frame rate; D represents the surface defect parameter, Ω represents the pixel point set in the geometric abnormal area, Represents the actual contour gradient of the jth pixel in the pixel set, Represents the ideal smooth contour gradient of the j-th pixel in the pixel set; The magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are weighted and normalized, and a rotor magnetic pole visual monitoring evaluation result is output.

7. The visual monitoring method based on rotor magnetic poles according to claim 6, characterized in that: Also includes: The dynamic image sequence containing the rotor magnetic pole edge features is collected in real time using a high-speed industrial camera that is arranged outside the rotor rotation plane and maintains a preset inclination angle with the rotor rotation axis.

8. The visual monitoring method based on rotor magnetic poles according to claim 6, characterized in that: The steps of performing frame-by-frame decomposition and spatiotemporal registration of a dynamic image sequence containing rotor pole edge features to eliminate image distortion caused by high-speed rotor rotation and environmental electromagnetic interference include: According to the real-time angular velocity signal of the rotor encoder, the theoretical pixel displacement of the magnetic pole between adjacent frames is calculated; wherein, the calculation formula of the theoretical pixel displacement is: ω is the rotor angular velocity, r is the rotor pole radius, f is the camera focal length, d is the object distance, and Δt is the frame interval time; When decomposing the dynamic image sequence frame by frame, reverse displacement compensation is performed on the magnetic pole edge features of each frame based on the theoretical pixel displacement to eliminate motion blur and inter-frame misalignment caused by high-speed rotation; A fast Fourier transform is performed on each frame of the image to extract the characteristic frequency band related to electromagnetic interference in the frequency domain. An adaptive notch filter is used to attenuate the energy of the characteristic frequency band, retaining the main frequency component of the magnetic pole contour, and obtaining a dynamic image sequence after dynamic compensation.

9. The visual monitoring method based on rotor magnetic poles according to claim 8, characterized in that: The steps of extracting the magnetic pole contour feature vector of the current frame based on the multi-scale edge detection algorithm include: Applying Gaussian kernel convolution with different standard deviations to each frame of the dynamic image sequence after the dynamic compensation to generate a multi-scale image set including coarse-grained scale features and fine-grained scale features of the magnetic pole edge; For each scale image, the binarization threshold is dynamically adjusted based on the local contrast to generate a binarized image containing the magnetic pole edge. Dilation and erosion operations are then performed on the binarized image in sequence to calculate the pixel difference between the dilation and erosion results, thereby enhancing edge continuity and suppressing isolated noise points. The edge contour of the coarse-grained scale feature is used as the reference skeleton, and the sub-pixel edge offset of the fine-grained scale feature is superimposed to generate the fused magnetic pole contour; Equal-angle key point sampling is performed along the fused magnetic pole contour, and the curvature, normal direction and local gradient amplitude of each point are calculated to form the current frame magnetic pole contour feature vector with unified dimension.

10. The visual monitoring method based on rotor magnetic poles according to any one of claims 6 to 9, characterized in that: The step of weighted normalizing the magnetic pole deformation parameter, the magnetic pole displacement parameter, and the surface defect parameter to output a rotor magnetic pole visual monitoring evaluation result comprises: The weight values corresponding to the magnetic pole deformation parameter, the magnetic pole displacement parameter and the surface defect parameter are dynamically adjusted based on the rotor speed and weighted normalization is performed to output the rotor magnetic pole visual monitoring evaluation result.

Images