Machine vision assembly appearance defect identification method for empty-cup Hall brushless motor

Abstract

The invention discloses a machine vision assembly appearance defect identification method for an empty-cup Hall brushless motor, and relates to the technical field of machine vision detection, and the method comprises the following steps: in a motor assembly operation process, according to the synchronization parameters of a reflection angle, illumination intensity and a rotation rhythm, carrying out the recognition of the appearance defect of the machine vision assembly; the method comprises the following steps: performing multi-angle image acquisition on the metal surface of the empty-cup Hall brushless motor in each direction to obtain continuous image fragments, recombining the continuous image fragments into an image distribution diagram, extracting double-image initial features and generating a double-image feature list; by constructing a synchronous acquisition mechanism of a reflection angle, illumination intensity and rotation rhythm and combining a double-image dial gauge and a phase deviation fingerprint, time matching of exposure and magnetic pole switching is realized, and a double-image error caused by magnetic noise interference is eliminated; and through cooperative control of a reverse light screen, graded skip shooting exposure and a magnetic noise isolation window, illumination and imaging rhythm are stabilized, and image purity and appearance detection precision are improved.

Description

Machine vision assembly appearance defect identification method for empty cup Hall brushless motors Technical Field

[0001] This invention relates to the field of machine vision inspection technology, specifically to a machine vision assembly appearance defect identification method for empty cup Hall brushless motors. Background Technology

[0002] Machine vision assembly appearance defect recognition for empty-cup Hall effect brushless motors refers to the introduction of machine vision technology into the assembly process of empty-cup Hall effect brushless motors. This involves using an industrial camera to capture real-time images of the appearance of each assembly component (such as the stator, rotor, magnets, Hall elements, end caps, etc.) and combining this with image processing techniques such as illumination equalization, image segmentation, edge detection, and feature extraction to automatically identify and determine defects such as scratches, misalignments, missing parts, deformation, and contamination on the surface of the assembled components. This method uses a vision system to replace manual visual inspection, enabling multi-angle detection and defect classification within the assembly cycle. This achieves intelligent monitoring of the motor's appearance quality and online control of assembly accuracy, ensuring the consistency and reliability of the entire machine.

[0003] The existing technology has the following shortcomings: In the existing technology, when the empty cup Hall brushless motor performs Hall positioning, the rotor magnetic poles dynamically switch magnetic directions at a high frequency, which easily generates transient magnetic noise in local areas. This magnetic noise can interfere with the photosensitive element of the machine vision system, causing distortion of the light signal reflected from the metal surface, and thus forming slightly misaligned or duplicated edge images in the image. Due to the strong reflective properties of the metal surface, this double image phenomenon is easily identified as cracks or fracture defects during image analysis, leading to misjudgment of defect identification results. Especially when the assembly cycle changes rapidly, the asynchronous switching of magnetic poles and exposure timing will exacerbate imaging drift, significantly reducing the accuracy of defect detection, thereby affecting the reliability of appearance inspection and the consistency of assembly quality.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a machine vision assembly appearance defect identification method for empty cup Hall brushless motors, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, this invention provides the following technical solution: a machine vision assembly appearance defect identification method for empty-cup Hall brushless motors, comprising the following steps: during motor assembly and operation, based on synchronization parameters of reflection angle, illumination intensity, and rotation rhythm, multi-angle image acquisition is performed on the metal surface of the empty-cup Hall brushless motor from various directions to obtain continuous image segments, and the continuous image segments are reconstructed into an image distribution map, extracting initial double-image features and generating a double-image feature list; using the double-image feature list, the continuous image segments are compared frame by frame, and the magnetic noise interference area and exposure offset position are determined through differences in grayscale changes, reflection density, and edge sharpness; the interference area and offset position are converted into time scales to generate a double-image scale table, providing data for subsequent beat analysis. Time reference: Based on the double-image scale, the assembly acquisition cycle is traced back to analyze the grayscale jump pattern during rotor magnetic pole switching, extract the phase deviation fingerprint, establish the distribution relationship of imaging synchronization error in the time series, and use the phase deviation fingerprint output as the basis for exposure synchronization adjustment; the time sequence of exposure and framing is determined according to the phase deviation fingerprint, the framing stop point, polarization switching sequence and light source trigger delay are replanned, and a complete exposure synchronization adjustment scheme is generated to provide time instructions for dynamic imaging control; dynamic control is implemented according to the exposure synchronization adjustment scheme, the interference of metal reflection is canceled by projecting a reverse light screen, and a graded skip exposure and magnetic noise isolation window are used in conjunction to correct the imaging rhythm in real time, eliminate double-image error and restore the image purity for assembly appearance defect identification.

[0007] Preferably, the steps for generating the double-image feature list are as follows: Based on the rotation rhythm signal of the motor assembly and operation, establish a data acquisition schedule that matches the rotation cycle; determine the framing angle interval by monitoring the change in the rotation angle of the rotating shaft; and adjust the light source brightness and illumination distance according to the reflective characteristics of the metal surface to keep the reflection angle, light intensity, and rotation rhythm synchronized; perform multi-angle synchronous framing according to the preset angle step, so that the metal surface in all directions of the motor rotor is continuously acquired within the complete rotation cycle, and record the timestamp, angle number, and light intensity information; based on the recorded data of reflection angle and light intensity, spatially reconstruct the continuous image segments to form an image distribution map, while maintaining the smooth and continuous distribution of light intensity; analyze the light reflection differences at different angles based on the image distribution map, identify double-image regions and extract the initial double-image features, record the spatial position, angle, time, and light reflection intensity information, and generate a double-image feature list.

[0008] Preferably, in the process of forming the image distribution map, the circumferential coordinates of the motor housing are used as a reference, the continuous image segments are stitched together in the order of shooting angle, the brightness difference between adjacent images is processed for light intensity balance, and the combined data of reflection angle and light intensity is used as weight to integrate the multi-angle images, so that the image distribution map reflects the real reflection distribution state of the metal surface and maintains spatial continuity.

[0009] Preferably, the steps for generating the double-image scale table are as follows: Based on the angle number, reflection direction, illumination intensity, and acquisition time information recorded in the double-image feature list, a correspondence is established for continuous image segments. The angle difference between adjacent frames is calculated using the angle recording information and rotation rhythm parameters to ensure that continuous images maintain consistent illumination. Based on the established correspondence, optical attribute difference analysis is performed on continuous image frames of the same surface area. The magnetic noise interference area and exposure offset position are identified based on changes in grayscale distribution, reflection density, and edge sharpness. The magnetic noise interference area and exposure offset position are time-series calibrated with the acquisition time, and a time correspondence is established using timestamps and rotation angle parameters. A double-image scale table is generated based on the time information, grayscale changes, reflection density, and edge sharpness changes of the interference event, so that the interference event forms a continuous distribution on the time axis, providing a time reference for subsequent beat analysis.

[0010] Preferably, the phase deviation fingerprint extraction process is as follows: Based on the time scale information recorded in the dual-image scale table and the time sequence of interference events, the assembly acquisition cycle is traced back to establish the time correspondence between the motor rotor magnetic pole switching and image acquisition; after obtaining the time frame, the gray-level jump pattern is analyzed for the image sequence within each magnetic pole switching interval to determine the start and end points of gray-level changes, forming a gray-level jump time distribution sequence; the gray-level change data is compared and analyzed with the magnetic pole switching cycle to extract the phase deviation fingerprint and form a continuous deviation time series; based on the phase deviation information, the distribution relationship of imaging synchronization error in the time series is established, and the phase deviation fingerprint output is used as the time basis for exposure synchronization adjustment to keep the imaging process and the magnetic pole switching process in sync.

[0011] Preferably, in the analysis of grayscale jump patterns, each magnetic pole switching cycle is divided into a start-up phase, a magnetic pole reversal phase, and a stabilization phase. The time interval of rapid grayscale change is used as the grayscale jump range. By comparing the grayscale change amplitude with the magnetic pole switching time difference, the temporal distribution of phase deviation is determined, so that the exposure trigger moment is synchronized with the magnetic pole switching cycle, thereby improving the temporal accuracy of the imaging synchronization error distribution relationship.

[0012] Preferably, the steps for generating the exposure synchronization adjustment scheme are as follows: Based on the time scale information and exposure offset data in the phase deviation fingerprint, the phase deviation curve is analyzed, and the magnetic pole switching cycle is divided into the pre-magnetic pole reversal zone, the middle magnetic pole reversal zone, and the post-magnetic pole stabilization zone to determine the executable imaging time interval; based on the time interval, the time sequence of exposure and framing is determined, with the pre-magnetic pole reversal zone set as the framing start window and the post-magnetic pole stabilization zone set as the exposure execution window, forming a continuous and coordinated time sequence; based on the phase deviation fingerprint, the framing stop point is reset so that the framing position corresponds to the magnetic pole stabilization stage, and the angle position, time parameters, and light intensity information are recorded; based on the optical change law, the polarization switching sequence is planned so that the polarization angle switching action corresponds to the framing stop point time node; based on the exposure offset information, the light source trigger delay is set, and the framing stop point time schedule, polarization switching sequence, and light source trigger time are integrated to generate a complete exposure synchronization adjustment scheme.

[0013] Preferably, dynamic control is implemented according to the exposure synchronization adjustment scheme. The imaging rhythm correction steps, using a projected backlight screen to counteract metal reflection interference and a coordinated control method of graded skip-shot exposure and magnetic noise isolation window, are as follows: Imaging parameters are loaded according to the exposure synchronization adjustment scheme. The viewfinder stop timetable, polarization switching sequence, light source trigger delay, and exposure duration are imported into the control flow. A time correspondence between exposure trigger and magnetic pole switching cycle is established based on the assembly beat signal. A backlight screen projection operation is performed, with the projection direction opposite to the camera's viewfinder direction. A reflection balance area is formed through symmetrical light sources, maintaining stable illumination intensity within the exposure cycle. Graded skip-shot exposure is performed according to the exposure synchronization scheme, dividing the exposure cycle into initial, transition, and stable stages, ensuring a smooth and continuous transition between illumination intensity and time. The exposure window of the photosensitive element is controlled, and a magnetic noise isolation window timing operation is implemented, ensuring exposure only occurs during the stable magnetic field stage. The exposure trigger time and light source illumination time are corrected in real-time according to the assembly beat signal, synchronizing the exposure process with magnetic pole switching, eliminating double image errors, and restoring image purity.

[0014] In the above technical solution, the technical effects and advantages provided by this invention are as follows: This invention introduces a synchronous acquisition mechanism for reflection angle, light intensity, and rotation rhythm during motor assembly and operation. Combined with the time-based construction of a double-image feature list and a double-image scale table, magnetic noise interference is quantified in the time dimension. Furthermore, precise matching of the exposure and magnetic pole switching processes is achieved through phase deviation fingerprinting. In this way, the imaging exposure process is synchronized with the motor assembly rhythm, avoiding exposure drift caused by magnetic field disturbances. This ensures that image acquisition maintains stable grayscale and clear edges in a high-frequency magnetic switching environment, fundamentally eliminating the misjudgment problem caused by double-image errors and improving the imaging accuracy and recognition reliability of appearance inspection.

[0015] This invention establishes an exposure synchronization adjustment scheme and implements dynamic control, coordinating the control of the backlight screen, graded skip-shooting exposure, and magnetic noise isolation window to ensure that the imaging illumination and reflection states remain continuous and consistent over time. Through illumination reversal compensation and exposure rhythm correction, real-time cancellation of metal reflection interference is achieved, restoring a balanced image grayscale distribution, preserving edge texture details, and effectively improving image purity. This method maintains stable visual imaging under high-speed assembly inspection conditions, providing a high-fidelity image foundation for appearance defect identification, thereby ensuring the consistency of motor assembly quality and the long-term reliable operation of visual inspection. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 is a flowchart of the machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to the present invention. Detailed Implementation

[0018] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0019] This invention provides a machine vision assembly appearance defect recognition method for empty cup Hall brushless motors, as shown in Figure 1, comprising the following steps: During the motor assembly operation, based on the synchronization parameters of reflection angle, illumination intensity, and rotation rhythm, multi-angle image acquisition is performed on the metal surface of the empty cup Hall brushless motor from all directions to obtain continuous image segments, and the continuous image segments are reconstructed into an image distribution map, extracting initial double-image features and generating a double-image feature list; During the assembly operation of the empty cup Hall brushless motor, in order to achieve high-precision recognition of metal surface appearance defects in dynamic production cycles, the entire image acquisition process uses the synchronization parameters of reflection angle, illumination intensity, and rotation rhythm as the control core, and performs continuous, complete, and multi-angle image acquisition on the metal surface from all directions during the motor assembly process. The specific implementation steps are as follows: Based on the actual rotation rhythm signal of the rotor during the motor assembly process, an acquisition time schedule matching the rotation cycle is established. By monitoring the change of the shaft rotation angle, the specific angle interval for imaging is determined, so that each shooting frame corresponds to a specific rotation angle. Based on the geometry of the motor housing, stator, and end cover, the matching relationship between the incident angle and reflection angle of the reflected light is calculated to determine the arrangement direction of the imaging light source. Then, based on the reflective characteristics of the metal surface, the brightness and illumination distance of the light source are adjusted to maintain relative consistency in light intensity at different acquisition angles. The light source is generally composed of a multi-point array structure, with each light point emitting light independently at different angles. Time-series control ensures that each light source is triggered sequentially during acquisition, guaranteeing precise matching between the reflection angle and the illumination direction. To synchronize with the assembly rhythm of the motor, the imaging device locks the acquisition sequence by controlling the feedback of the motor speed signal. Before each frame, the lens focal length and aperture are automatically adjusted according to the rotation angle, ensuring consistency in brightness, sharpness, and focus position in the acquired image. After this sub-step, acquisition parameters that synchronously match the reflection angle, light intensity, and rotation rhythm are obtained, providing a spatiotemporal reference for subsequent multi-angle image acquisition.

[0020] After setting the acquisition parameters, a multi-angle synchronous framing process is executed. During assembly operation, the imaging device continuously captures images according to preset angular steps, ensuring that all metal surfaces within a complete rotation cycle of the motor rotor are fully recorded. To ensure consistent illumination, the light source brightness, incident direction, and exposure time are automatically corrected based on the parameters of the previous frame for each image acquisition, resulting in a smooth transition in lighting conditions across the continuous images. Each image frame corresponds to a unique timestamp, angle number, and light intensity record, and all acquired data is stored continuously in a time-series format. During acquisition, to prevent overly bright images due to reflection saturation, the incident angle of the light source and the camera shutter time are adjusted in real time to ensure a stable distribution of reflected light from the metal surface within the photosensitive range. The assembly rotation rhythm is also strictly synchronized with the acquisition cycle to avoid image blurring caused by mechanical vibration or changes in rotational acceleration. This sub-step yields a continuous sequence of image segments covering the entire surface of the empty cup Hall effect brushless motor, with each segment recording complete angle information, light intensity values, and shooting time parameters.

[0021] Based on the obtained sequence of continuous image segments, all image segments are spatially reconstructed according to the recorded information of reflection angle and illumination intensity to form an image distribution map. First, using the circumferential coordinates of the motor housing as a reference, the image segments are stitched together in order of shooting angle, ensuring spatial alignment of the image content. To ensure the smoothness of the image stitching boundaries, the brightness differences between adjacent images are balanced to maintain a consistent brightness distribution in overlapping areas. During the stitching process, multi-angle images at the same location are weighted and integrated using the combined data of reflection angle and illumination intensity, enabling the image distribution map to accurately reflect the light reflection distribution of the metal surface. Each pixel in the image distribution map contains correlation information between the acquisition angle, shooting time, and reflection intensity, forming a spatiotemporal mapping of the metal surface's reflection characteristics. Through this reconstruction process, the originally scattered image segments are integrated into a complete image with spatial continuity and temporal consistency, providing a complete visual record of the metal surface's reflection behavior throughout the entire assembly cycle. After this sub-step, the image distribution map not only displays the optical state of the metal surface in various directions but also retains the temporal correspondence between illumination changes and reflection angles, providing a continuous image foundation for subsequent double-image analysis.

[0022] After the image distribution map is constructed, the initial features of double images are identified and a double image feature list is generated. This process is based on the differences in light reflection at different angles in the image distribution map. By analyzing the brightness distribution and edge morphology changes of the same location in images from different angles, possible double image regions are identified. When a region exhibits repeated edges, abnormally increased reflection intensity, or contour displacement in images from consecutive angles, it is determined that the region has double image features. Subsequently, information such as the spatial location, acquisition angle, shooting time, reflected light intensity value, and incident light direction of the corresponding region is recorded and arranged in angular order to form double image feature entries. All entries are uniformly arranged according to the metal surface location index, forming a complete double image feature list. The double image feature list not only records the spatial location and temporal distribution of double image phenomena but also reflects the changing patterns of reflection angle and light intensity, providing a temporal basis for subsequent magnetic noise interference localization and exposure synchronization adjustment. Through this step, the extraction and organization of double image feature data from the image distribution map is completed, enabling a quantitative description of the reflection state of the metal surface during motor assembly in both temporal and spatial dimensions.

[0023] A frame-by-frame comparison of consecutive image segments is performed using a dual-image feature list. By analyzing differences in grayscale changes, reflectance density, and edge sharpness, magnetic noise interference areas and exposure offset positions are identified. These interference areas and offset positions are then converted into time scales to generate a dual-image scale table, providing a time reference for subsequent beat analysis. After establishing the dual-image feature list, frame-by-frame comparison is performed on continuously acquired image segments during the assembly process. By analyzing grayscale changes, reflectance density differences, and edge sharpness changes, magnetic noise interference areas and exposure offset positions are identified. These positions are then converted into time scales to generate a dual-image scale table for subsequent beat analysis. The specific implementation steps are as follows: Based on the angle number, reflection direction, illumination intensity, and acquisition time information recorded in the dual-image feature list, a correspondence is established for consecutive image segments. During this process, each frame is precisely matched spatially with adjacent frames to ensure that the same metal surface areas maintain positional matching in consecutive frames. To eliminate the impact of minor mechanical vibrations during assembly on image framing, angle differences between adjacent frames are calculated using angle recording information and rotation rhythm parameters, ensuring alignment accuracy remains within the allowable range of rotor angle offset. During alignment, image brightness distribution is also corrected based on illumination intensity data, maintaining continuity in illumination across each frame and preventing edge blurring or reflective bright spots caused by inconsistent brightness. This step establishes a temporal correspondence between consecutive image frames, with each frame matched one-to-one with acquisition angle, acquisition time, and illumination intensity data, providing an accurate spatial basis for optical property difference analysis.

[0024] After establishing the inter-frame correspondence, optical attribute differences are analyzed for consecutive image frames of the same surface region. Using the illumination intensity and reflection direction in the double-image feature list as references, the grayscale distribution, reflection density, and edge sharpness of corresponding regions in consecutive frames are compared one by one. Grayscale distribution reflects the changes in light energy in the image. By analyzing the trend of grayscale changes at the same location in adjacent frames, it can be determined whether there are light reflection fluctuations in that region. Reflection density reflects the reflectivity of the metal surface. By comparing the reflection intensity distribution of adjacent frames, reflection anomalies caused by magnetic field disturbances or illumination angle shifts can be identified. Edge sharpness reflects the imaging stability of image details. By analyzing the edge sharpness differences of the same contour region in consecutive frames, it can be determined whether there is optical jitter or imaging drift during the imaging process. When the grayscale distribution shows short-term periodic fluctuations and the change in reflection density distribution is independent of illumination intensity, it can be determined that the region is affected by magnetic noise interference. When the edge sharpness undergoes abrupt changes in adjacent frames and is accompanied by an abnormal increase in reflection brightness, it can be determined that there is an exposure shift at that location. This analysis process allows for the accurate spatial location of specific areas affected by magnetic noise interference and exposure shift, providing a data foundation for the next step of time calibration.

[0025] After identifying the interference areas, the identified magnetic noise interference areas and exposure offset positions are time-series calibrated with the acquisition time. Calibration uses the timestamp of each frame as a reference, mapping the specific frame index of the interference area to the acquisition time, and establishing a connection between this time information and the rotation angle parameter in the assembly rhythm signal, ensuring that each interference event has a unique temporal position on the timeline. During the time calibration process, the rotation rhythm parameter is used as a time interpolation reference, maintaining a linear relationship between the time interval between adjacent frames and the change in rotation angle, thus ensuring that the calibration time is synchronized with the physical motion process. To ensure the accuracy of the temporal correspondence, the illumination intensity and reflection density values ​​are recorded simultaneously during calibration, so that the time scale can be verified using these optical parameters in subsequent time-series analysis. After this step, each interference area and exposure offset position is assigned a specific time scale value, forming an index relationship between interference features and time.

[0026] After obtaining the time scales for all interference locations, a dual-image scale is generated based on the time information, grayscale changes, reflection density differences, and edge sharpness changes of each interference event. The dual-image scale uses the time axis as its main axis, arranging all interference events in the order of acquisition, with each time scale corresponding to a complete set of optical parameter records. Each record includes the time of interference occurrence, the corresponding rotation angle, the spatial coordinates of the interference area, the intensity of reflected light, the amplitude of grayscale changes, and the edge sharpness difference. To facilitate subsequent assembly cycle analysis, the dual-image scale is grouped according to the rotation period, ensuring a continuous distribution of all interference events within the same rotation period on the time axis. In this way, the dual-image scale not only reflects the temporal sequence of magnetic noise interference and exposure shift but also demonstrates the correlation between interference intensity and assembly cycle. Through this step, the dual-image scale becomes the fundamental data carrier describing the temporal characteristics of magnetic noise interference and exposure shift during imaging, providing a complete time reference for subsequent cycle synchronization analysis, exposure time adjustment, and imaging timing optimization.

[0027] Based on the dual-image scale, the assembly acquisition cycle is traced back to analyze the grayscale jump pattern during rotor magnetic pole switching, extract the phase deviation fingerprint, establish the distribution relationship of imaging synchronization error in the time series, and use the phase deviation fingerprint output as the basis for exposure synchronization adjustment. To achieve precise time control of imaging synchronization, the assembly acquisition cycle is traced back based on the dual-image scale to analyze the grayscale jump pattern during rotor magnetic pole switching, extract the phase deviation fingerprint, and establish the distribution relationship of imaging synchronization error in the time series, thus outputting a time basis that can be used for exposure synchronization adjustment. The specific implementation steps are as follows: Based on the time scale information recorded in the dual-image scale and the time sequence of interference events, the acquisition cycle of the assembly process is traced back to restore the time correspondence between motor rotor magnetic pole switching and image acquisition. In the implementation process, all time scale data in the dual-image scale is first read and compared with the cycle signal provided by the assembly control terminal item by item, and each time scale is precisely associated with the motor rotor rotation angle, rotation speed, and Hall signal trigger time. The physical position of the rotor at each acquisition time point can be determined through the rotor rotation angle sensing data, thus clarifying the time interval between the start and end of magnetic pole switching. To ensure the continuity of time correspondence, each magnetic pole switching process was divided into three time periods: the start-up phase, the magnetic pole reversal phase, and the stabilization phase. The corresponding relationships were marked in the double-image scale table, ensuring a one-to-one correspondence between the magnetic field change period and the imaging acquisition time axis. Through this backtracking process, a time structure framework for motor assembly and operation was established, giving each image acquisition moment a clear magnetic pole switching position in the time series, providing a basic time reference for grayscale change analysis.

[0028] After obtaining the timeframe following the beat retracing, grayscale jump patterns were analyzed for the image acquisition sequences within each magnetic pole switching interval to identify the temporal impact range of magnetic noise interference. During the analysis, the interference areas marked in the dual-image scale table and their corresponding image frames were extracted one by one. The grayscale distribution curve of the metal surface region in each frame was calculated, and the continuous change characteristics of grayscale values ​​over time were observed. By comparing the changing trends of the grayscale distribution curves, the time periods of rapid grayscale increase or decrease were identified, thus determining the start and end points of grayscale jumps. The grayscale jump phenomenon reflects the influence of magnetic field disturbances on image brightness during magnetic pole switching. When the magnetic field direction reverses, the reflectivity of the metal surface changes, causing instability in the distribution of light energy received by the photosensitive element, which manifests as a sudden change in grayscale in the image. By statistically analyzing the grayscale jumps of all acquisition frames within the same magnetic pole switching cycle, a grayscale jump time distribution sequence can be formed. This sequence records the fluctuation pattern of grayscale changes on the time axis, providing dynamic basic data for subsequent phase deviation extraction.

[0029] After obtaining the temporal distribution sequence of grayscale jumps, these grayscale change data are compared and analyzed with the magnetic pole switching time frame to extract the phase deviation fingerprint. During implementation, the time difference between the grayscale jump time point and the magnetic pole switching time point is calculated as a baseline, thus forming the phase deviation value corresponding to each imaging frame. Each phase deviation value represents the time offset between the exposure trigger moment and the magnetic pole switching occurrence moment. By recording all deviation values ​​within a continuous acquisition cycle, a complete time series can be formed. The formation process of the phase deviation fingerprint includes organizing the relative relationship between exposure trigger and magnetic pole switching within a continuous cycle, arranging the deviation time points in each cycle in chronological order to form a continuous deviation curve. This curve records the changing trend of the imaging exposure sequence relative to the magnetic pole switching sequence, reflecting the changes in imaging synchronization error at different stages during assembly. Each phase deviation fingerprint point includes a time scale, the corresponding grayscale jump amplitude, the start and end times of the magnetic pole switching, and the exposure trigger time information. Through this process, the transition from grayscale change characteristics to temporal phase deviation characteristics is completed, enabling a quantifiable description of the temporal relationship between exposure and magnetic field interference.

[0030] After extracting the phase deviation fingerprint, the distribution relationship of imaging synchronization error in the time series is established based on all deviation information, and the phase deviation fingerprint output is used as the basis for exposure synchronization adjustment. In practice, all phase deviation points are rearranged according to the time scale to form a complete synchronization error time distribution curve. This curve reflects the trajectory of the time offset between exposure triggering and magnetic pole switching throughout the assembly cycle. To ensure the continuity of the synchronization error distribution, the phase deviation curves from multiple acquisition cycles are stitched together on the time axis, ensuring continuous connection of deviation information between cycles. This method allows observation of the repetitive patterns of synchronization error during long-term operation, thereby identifying the periodic changes in imaging error. The distribution relationship of imaging synchronization error not only shows the time difference between exposure triggering delay and magnetic pole switching delay but also reflects the accumulation trend of error throughout the acquisition process. Finally, this time distribution relationship is aligned with the exposure control signal, and the offset segment is extracted as the time basis for exposure synchronization adjustment. In the subsequent exposure control stage, based on the time information of the phase deviation fingerprint output, the exposure trigger time, the light source illumination delay, and the viewfinder stop point are adjusted synchronously to make the imaging process consistent with the magnetic pole switching process in time, thereby eliminating the exposure error caused by magnetic field changes.

[0031] Based on the phase deviation fingerprint, the timing sequence of exposure and framing is determined. The framing stop point, polarization switching sequence, and light source trigger delay are then redesigned to generate a complete exposure synchronization adjustment scheme, providing timing commands for dynamic imaging control. To achieve precise coordination of the exposure and framing timing sequence, the framing stop point, polarization switching sequence, and light source trigger delay must be redesigned based on the phase deviation fingerprint, thus forming an exposure synchronization adjustment scheme that covers the entire assembly process and provides accurate timing commands for dynamic imaging control. The specific implementation steps are as follows: Based on the time scale information and exposure offset data in the phase deviation fingerprint, the phase deviation curve is analyzed, and the time intervals within the assembly cycle are divided. In practice, the exposure offset, grayscale change amplitude, and magnetic pole switching stage corresponding to each time scale in the phase deviation fingerprint are correlated and organized to form an offset distribution table arranged in chronological order. Through aggregated analysis of the data in the offset distribution table, each magnetic pole switching cycle is divided into three time periods: the pre-magnetic pole reversal zone, the mid-magnetic pole reversal zone, and the post-magnetic pole stabilization zone. The pre-magnetic pole reversal zone is the period when the magnetic field is about to switch but has not yet disturbed the image sensor; the mid-magnetic pole reversal zone is the period when the magnetic flux direction changes and magnetic noise interference is concentrated; the post-magnetic pole stabilization zone is the period when the magnetic field has returned to equilibrium. This division clarifies the temporal executable areas for framing and exposure actions, providing a basic time framework for determining the subsequent time sequence. Each time interval is labeled with a specific start time, end time, and corresponding magnetic pole state, ensuring that the entire imaging process is consistent with the changes in the motor's magnetic field.

[0032] After dividing the time intervals, the timing sequence of exposure and framing is determined based on the characteristics of each interval, ensuring a continuous and coordinated relationship between framing and exposure actions on the timeline. In practice, the area before magnetic pole reversal is set as the framing initiation window, the area after magnetic pole stabilization is set as the exposure execution window, and the middle area of ​​magnetic pole reversal is defined as the no-exposure zone. To ensure seamless transition between framing and exposure, the framing initiation time is set at the middle of the area before magnetic pole reversal, allowing the framing to complete composition adjustments before magnetic field disturbances occur. The exposure initiation time is set at the beginning of the area after magnetic pole stabilization, ensuring the sensor performs exposure operations in a stable magnetic field state, thus avoiding the impact of grayscale jumps on imaging. The time interval between framing and exposure remains within the continuous range of the magnetic pole reversal cycle, with a fixed time offset between them to ensure that the exposure stage begins immediately after framing. This timing sequence determination process allows for seamless transition between framing and exposure actions within the same magnetic field cycle, ensuring that exposure triggering avoids magnetic noise interference areas and maintaining temporal stability and spatial continuity in the imaging process.

[0033] After determining the timing sequence of framing and exposure, the framing stop points are reset so that the framing positions correspond both spatially and temporally to the stable phase of the magnetic pole switching. The reset of the framing stop points is based on the time scale recorded in the phase deviation fingerprint, corresponding the execution time of each framing stop point to the stable phase of the magnetic pole switching cycle. During implementation, firstly, based on the correspondence between the rotor rotation angle and the time scale, the distribution interval of the framing stop points on the time axis is calculated, ensuring that the framing stop points coincide with the center time point of the magnetic pole stabilization zone. Then, the angular position, time parameters, and illumination intensity information of the framing stop points are recorded as a set of time node data. To ensure framing continuity, a constant time difference is maintained between adjacent stop points, ensuring that different orientations of the metal surface are continuously recorded within the assembly cycle. Each framing stop point is also associated with illumination intensity calibration data to ensure uniform illumination distribution at different angles. By replanning the framing stop points, the imaging device can perform framing within the time interval of magnetic field stability, reducing image blurring and light reflection distortion caused by magnetic field disturbances, and providing an accurate time reference for subsequent polarization adjustments.

[0034] After the viewfinder stop point is reset, the polarization switching sequence is planned based on the optical change patterns reflected in the phase deviation fingerprint to ensure that the incident light direction matches the reflection state of the metal surface. During implementation, the polarizer angle switching action is aligned with the time node of the viewfinder stop point, and each polarization angle adjustment is completed before the viewfinder begins. Based on the grayscale changes and reflection characteristics within different time intervals, the polarization angle adjustment range is divided into an initial angle zone, a transition angle zone, and a stable angle zone. The polarization angle in the initial angle zone is set before the magnetic pole reversal, ensuring the polarization direction is consistent with the incident light; the transition angle zone corresponds to the middle of the magnetic pole reversal, where the polarization angle switches rapidly to cope with changes in light reflection caused by magnetic field disturbances; the stable angle zone corresponds to the post-magnetic pole stabilization zone, where the polarization angle remains constant to ensure uniform distribution of reflected light. Each angle adjustment action in the polarization switching sequence is referenced to the time node of the viewfinder stop point, and the exposure stage begins immediately after the polarization adjustment is completed, thus forming a dynamic matching relationship between the optical path and the viewfinder time. By precisely planning the polarization timing, the illumination direction, reflection angle, and imaging angle are kept consistent over time, ensuring optical stability during image acquisition.

[0035] After completing the polarization switching timing plan, the light source trigger delay is set based on the exposure offset information in the phase deviation fingerprint, and a complete exposure synchronization adjustment scheme is generated. The light source trigger delay is set based on the exposure trigger time, combined with the time offset value recorded in the phase deviation fingerprint, to determine the precise times for light source illumination, holding, and extinguishing. The start time of the light source illumination phase is set before the viewfinder stop is initiated, ensuring the light reaches a stable intensity before exposure; the light source holding phase duration is consistent with the exposure duration, ensuring constant light intensity during exposure; the light source extinguishing phase delays the end of the exposure, allowing the light decay process to be gradual and avoiding sudden brightness changes that cause image ghosting. The light source trigger delay time is synchronized with the polarization switching completion time, ensuring that the polarization direction, light intensity, and exposure trigger holding time are consistent. After completing the light source trigger delay setting, the viewfinder stop timetable, polarization switching timing, and light source trigger time are sequentially integrated to form a complete exposure synchronization adjustment scheme. This scheme records the start and end times, duration, and time offsets between each action in detail, and outputs them in timetable form, providing executable time commands for dynamic imaging control.

[0036] Dynamic control is implemented based on the exposure synchronization adjustment scheme. Metal reflection interference is counteracted by projecting a reverse light screen. A coordinated approach of graded skip-shot exposure and magnetic noise isolation windows is used to correct the imaging rhythm in real time, eliminating double-image errors and restoring image purity for identifying assembly defects. To ensure stable and consistent exposure and imaging processes under magnetic field disturbances, dynamic control based on the exposure synchronization adjustment scheme enables multi-level coordination of imaging illumination, framing timing, and exposure actions. This process counteracts metal reflection interference by projecting a reverse light screen and combines graded skip-shot exposure with time-coordinated control of the magnetic noise isolation window to achieve real-time correction of the imaging rhythm, eliminate double-image errors, and restore image purity for identifying appearance defects. The specific implementation steps are as follows: Imaging parameters are loaded according to the exposure synchronization adjustment scheme, and time and illumination variables in the dynamic control process are initialized. Specifically, the framing stop time schedule, polarization switching timing, light source trigger delay, and exposure duration recorded in the synchronization adjustment scheme are sequentially imported into the imaging control flow. Based on the real-time beat signal of the motor assembly, a time correspondence is established between the exposure trigger time, framing hold time, light source illumination duration, and magnetic pole switching cycle. During initialization, the output power of the light source is configured in zones, and the brightness rise, plateau, and decay phases of each illumination cycle are each set with fixed durations to ensure that the light intensity reaches a stable state before exposure begins. Considering the reflectivity of different metal components, the illumination angle and incident direction of the light source are fine-tuned according to a preset reflectivity coefficient, ensuring that the light incident path aligns with the normal direction of the metal surface. Through this process, all time variables, illumination parameters, and imaging cycles are synchronized, providing a unified time reference for subsequent illumination adjustments and exposure execution.

[0037] After the imaging parameters are initialized, the projection operation of the backlight screen is executed to counteract optical interference caused by reflections from the metal surface. During implementation, the projection direction of the backlight screen is set opposite to the camera's framing direction, causing the light to overlap in reverse with the main reflection path in space. By placing symmetrical light sources on both sides of the motor assembly area, light from different directions converges on the metal surface, thus forming a reflection balance area. The intensity of the backlight is adjusted in layers according to the metal surface's reflection characteristics: areas with higher reflection intensity are illuminated with stronger backlight to reduce overly bright areas; areas with lower reflection intensity are illuminated with relatively weaker backlight to maintain overall brightness uniformity. The projection time of the backlight is synchronized with the exposure time, turning on five milliseconds before exposure begins and turning off five milliseconds after exposure ends, maintaining a balanced light distribution throughout the exposure cycle. To prevent excessive reflection caused by light overlap, the brightness of the backlight screen automatically compensates for the light source illuminance during exposure, creating an inverse light intensity curve between the two. This light cancellation method balances the reflected light from the metal surface, reducing image distortion caused by bright areas and specular reflection, and providing an optically balanced environment for subsequent graded exposures.

[0038] After the backlight projection is completed, a graded skip-exposure process is executed according to the time sequence set in the exposure synchronization scheme to dynamically balance the imaging brightness across different time periods. Specifically, a complete exposure cycle is divided into three consecutive stages: an initial exposure stage, a transition exposure stage, and a stable exposure stage. The initial exposure stage is executed immediately after exposure begins, with an exposure time of one-third of the entire cycle. The light source illuminance is set to the base brightness level to capture the initial reflection state of the metal surface. The transition exposure stage starts immediately after the initial exposure stage ends, with the exposure time and light intensity increasing simultaneously to record the reflection changes under gradually increasing illumination. The stable exposure stage starts after the transition stage ends, with the exposure duration and light intensity remaining constant to acquire a clear image during the stable magnetic field phase. The switching time of each exposure stage precisely corresponds to the time scale in the phase deviation fingerprint, ensuring that exposure triggering avoids the magnetic pole switching range. To prevent sudden changes in image grayscale caused by abrupt changes in illumination, the transition time between exposures remains continuous, forming a smooth connection in the exposure curve. The execution result of graded skip-exposure allows for the simultaneous preservation of details in both bright and dark areas of the image, resulting in a more balanced grayscale distribution and providing an optical basis for magnetic noise isolation control.

[0039] While performing stepped exposure, the exposure window of the image sensor is time-isolated and controlled synchronously with a magnetic noise isolation window to reduce interference from magnetic field disturbances on the image signal. The magnetic noise isolation window is precisely time-corresponding to the exposure trigger, with the opening and closing of the image sensor controlled precisely according to the magnetic pole switching cycle. The opening time of the isolation window is set within a delay after the magnetic pole reversal ends, allowing the image sensor to begin receiving light signals during the magnetic field stabilization phase; the closing time is set immediately after the exposure ends to prevent magnetic noise from entering the photosensitive process when the magnetic pole switching restarts. The duration of the isolation window is consistent with the stabilization phase of the stepped exposure, ensuring complete overlap between the exposure and magnetic field stabilization phases. To ensure optical continuity, the opening and closing of the isolation window employs a smooth transition, gradually increasing and decreasing the photosensitive sensitivity to prevent abrupt changes in light between exposure start and stop. Through the timing control of the magnetic noise isolation window, the image sensor only operates during periods free from magnetic field interference, thus completely isolating photosensitive errors caused by magnetic flux changes and maintaining a pure imaging signal.

[0040] Based on the coordinated operation of graded skip-shot exposure and magnetic noise isolation window, the imaging rhythm is corrected in real time, allowing exposure and framing actions to dynamically adapt to changes in the assembly rhythm. During implementation, the assembly rhythm signal is compared with the imaging feedback time. When a change in motor speed or rhythm is detected, the next exposure trigger time and light source illumination time are immediately adjusted to re-align the exposure event with the assembly rhythm. The amount of time adjustment for exposure correction is determined based on the synchronization error information recorded in the phase deviation fingerprint, with adjustment precision controlled within the microsecond range. The projection time of the light source and the opening time of the backlight screen are synchronously corrected to ensure consistency on the time axis. During the imaging rhythm correction process, the light source illuminance, polarization direction, and exposure duration are dynamically compensated simultaneously to ensure the stability of the imaging illumination distribution. The corrected exposure rhythm remains synchronized with the assembly movement, double image error is completely eliminated, the image grayscale distribution is restored to uniformity, edge contours are clear, and the details of metal surface reflections are fully presented. Through continuous periodic dynamic correction, the imaging system can automatically adapt to rhythm fluctuations during the assembly process, achieving continuous synchronization of exposure and magnetic pole switching.

[0041] This invention introduces a synchronous acquisition mechanism for reflection angle, illumination intensity, and rotation rhythm during motor assembly and operation. Combined with the time-based construction of a double-image feature list and double-image scale, magnetic noise interference is quantified in the time dimension. Furthermore, precise matching of the exposure and magnetic pole switching processes is achieved through phase deviation fingerprinting. This method synchronizes the imaging exposure process with the motor assembly rhythm, avoiding exposure drift caused by magnetic field disturbances. It ensures stable grayscale and clear edges in high-frequency magnetic switching environments, fundamentally eliminating misjudgment problems caused by double-image errors and improving the imaging accuracy and recognition reliability of appearance inspection.

[0042] This invention establishes an exposure synchronization adjustment scheme and implements dynamic control, coordinating the control of the backlight screen, graded skip-shooting exposure, and magnetic noise isolation window to ensure that the imaging illumination and reflection states remain continuous and consistent over time. Through illumination reversal compensation and exposure rhythm correction, real-time cancellation of metal reflection interference is achieved, restoring a balanced image grayscale distribution, preserving edge texture details, and effectively improving image purity. This method maintains stable visual imaging under high-speed assembly inspection conditions, providing a high-fidelity image foundation for appearance defect identification, thereby ensuring the consistency of motor assembly quality and the long-term reliable operation of visual inspection.

[0043] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A machine vision assembly appearance defect identification method for empty cup Hall brushless motors, characterized in that, Includes the following steps: During the assembly and operation of the motor, multi-angle image acquisition is performed on the metal surface of the empty cup Hall brushless motor from various directions based on the synchronization parameters of reflection angle, light intensity, and rotation rhythm. Continuous image segments are obtained and reconstructed into an image distribution map. Initial features of the double image are extracted and a double image feature list is generated. The double image feature list is used to compare the continuous image segments frame by frame. By analyzing the differences in grayscale changes, reflection density, and edge sharpness, the magnetic noise interference area and exposure offset position are determined. The interference area and offset position are converted into a time scale to generate a double image scale table. Based on the double image scale table, the assembly acquisition cycle is traced back to analyze the grayscale jump pattern during rotor magnetic pole switching, extract the phase deviation fingerprint, and establish the distribution relationship of imaging synchronization error in the time series. The exposure and framing time sequence is determined according to the phase deviation fingerprint. The framing stop point, polarization switching sequence, and light source trigger delay are replanned to generate a complete exposure synchronization adjustment scheme. Dynamic control is implemented according to the exposure synchronization adjustment scheme. Metal reflection interference is canceled by projecting a reverse light screen. A graded skip exposure and magnetic noise isolation window are used in conjunction to correct the imaging rhythm in real time.

2. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 1, characterized in that, The steps for generating the double-image feature list are as follows: Based on the rotation rhythm signal of the motor assembly and operation, establish a data acquisition schedule that matches the rotation cycle; determine the framing angle interval by monitoring the change in the rotation angle of the rotating shaft; and adjust the light source brightness and illumination distance according to the reflective characteristics of the metal surface to keep the reflection angle, light intensity, and rotation rhythm synchronized; perform multi-angle synchronous framing according to the preset angle step, so that the metal surface in all directions of the motor rotor is continuously acquired within the complete rotation cycle, and record the timestamp, angle number, and light intensity information; based on the recorded data of reflection angle and light intensity, spatially reconstruct the continuous image segments to form an image distribution map, while maintaining the smooth and continuous distribution of light intensity; analyze the light reflection differences at different angles based on the image distribution map, identify double-image regions and extract the initial double-image features, record the spatial position, angle, time, and light reflection intensity information, and generate the double-image feature list.

3. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 2, characterized in that, In the process of forming the image distribution map, the circumferential coordinates of the motor housing are used as the reference, and continuous image segments are stitched together in the order of shooting angle. The brightness difference between adjacent images is processed to balance the light intensity, and the combined data of reflection angle and light intensity is used as weight to integrate the multi-angle images, so that the image distribution map reflects the true reflection distribution state of the metal surface and maintains spatial continuity.

4. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 2, characterized in that, The steps for generating the double-image scale are as follows: Based on the angle number, reflection direction, illumination intensity, and acquisition time information recorded in the double-image feature list, a correspondence is established for continuous image segments. The angle difference between adjacent frames is calculated using the angle recording information and rotation rhythm parameters to ensure that continuous images maintain consistent illumination. Based on the established correspondence, optical attribute differences are analyzed for continuous image frames of the same surface area. The magnetic noise interference area and exposure offset position are identified based on changes in grayscale distribution, reflection density, and edge sharpness. The magnetic noise interference area and exposure offset position are time-series calibrated with the acquisition time, and a time correspondence is established using timestamps and rotation angle parameters. A double-image scale is generated based on the time information of the interference event, grayscale changes, reflection density changes, and edge sharpness changes.

5. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 4, characterized in that, The phase deviation fingerprint extraction process is as follows: Based on the time scale information recorded in the dual-image scale table and the time sequence of interference events, the assembly acquisition cycle is traced back to establish the time correspondence between the motor rotor magnetic pole switching and image acquisition. After obtaining the time frame, the gray-level jump pattern of the image sequence within each magnetic pole switching interval is analyzed to determine the start and end points of the gray-level change, forming a gray-level jump time distribution sequence. The gray-level change data is compared and analyzed with the magnetic pole switching beat to extract the phase deviation fingerprint and form a continuous deviation time series. Based on the phase deviation information, the distribution relationship of imaging synchronization error in the time series is established, and the phase deviation fingerprint output is used as the time basis for exposure synchronization adjustment.

6. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 5, characterized in that, In the analysis of the grayscale jump pattern, each magnetic pole switching cycle is divided into a start-up phase, a magnetic pole reversal phase, and a stable phase. The time interval of rapid grayscale change is used as the grayscale jump range. By comparing the grayscale change amplitude with the magnetic pole switching time difference, the time distribution of phase deviation is determined, so that the exposure trigger time is synchronized with the magnetic pole switching cycle.

7. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 5, characterized in that, The steps for generating the exposure synchronization adjustment scheme are as follows: Based on the time scale information and exposure offset data in the phase deviation fingerprint, the phase deviation curve is analyzed, and the magnetic pole switching cycle is divided into the pre-magnetic pole reversal zone, the middle magnetic pole reversal zone, and the post-magnetic pole stabilization zone to determine the executable imaging time interval. The timing sequence of exposure and framing is determined based on the time interval. The area before magnetic pole reversal is set as the framing start window, and the area after magnetic pole stabilization is set as the exposure execution window, forming a continuous and coordinated timing sequence. The framing stop point is reset based on the phase deviation fingerprint to make the framing position correspond to the magnetic pole stabilization stage, and the angle position, time parameters, and light intensity information are recorded. The polarization switching sequence is planned according to the optical change law to make the polarization angle switching action correspond to the framing stop point time node. The light source trigger delay is set according to the exposure offset information, and the framing stop point time table, polarization switching sequence, and light source trigger time are integrated to generate a complete exposure synchronization adjustment scheme.

8. The machine vision assembly appearance defect identification method for empty cup Hall brushless motors according to claim 7, characterized in that, The following steps are taken to correct the imaging rhythm by implementing dynamic control based on the exposure synchronization adjustment scheme, canceling metal reflection interference by projecting a backlight screen, and using a graded skip exposure and magnetic noise isolation window coordinated control method: The imaging parameters are loaded according to the exposure synchronization adjustment scheme, the viewfinder stop time schedule, polarization switching sequence, light source trigger delay and exposure duration are imported into the control process, and the time correspondence between exposure trigger and magnetic pole switching cycle is established according to the assembly beat signal. Perform a reverse light screen projection operation, with the projection direction opposite to the camera's framing direction, forming a reflection balance area through symmetrical light sources, and maintaining stable light intensity during the exposure cycle; The exposure synchronization scheme employs tiered skip exposure, dividing the exposure cycle into initial, transition, and stable phases to ensure a smooth and continuous transition between light intensity and time. The exposure window of the photosensitive element is controlled, and a magnetic noise isolation window timing operation is implemented to ensure that exposure occurs only during the stable phase of the magnetic field. The exposure trigger time and light source illumination time are corrected in real time based on the assembly cycle signal to keep the exposure process synchronized with the magnetic pole switching, eliminating double image errors and restoring image purity.

Images