Defect detection method for polaroid cutting

Abstract

The invention discloses a defect detection method for polaroid cutting, and relates to the technical field of polaroid processing, and the method comprises the following steps: before polaroid cutting is started, carrying out space calibration and time calibration on a cutting station, establishing a mapping relation between imaging pixel coordinates and cutting station plane coordinates, and obtaining a synchronous reference signal of cutting equipment; collecting a reference image of the polaroid and determining a reference pose of the polaroid under the condition that the cutting tool is not in contact with the polaroid and the pressing mechanism presses the polaroid; and when the cutting equipment executes a cutting action, determining a cutting event moment when the cutting tool is in contact with the polaroid based on the synchronous reference signal. Spatial calibration and time calibration are carried out on the cutting station before polaroid cutting is started, and the mapping relation between imaging pixel coordinates and cutting station plane coordinates is established, so that image acquisition data and cutting equipment motion data can be uniformly processed under the same spatial coordinate system and time reference.

Description

Technical Field

[0001] This invention relates to the field of polarizer processing technology, specifically to a defect detection method for polarizer cutting. Background Technology

[0002] As an important optical component of display panels, the cutting accuracy of polarizers directly affects the subsequent bonding quality and display performance. During the cutting process of polarizers, the relative positional relationship between the cutting tool and the polarizer is affected by various factors such as the motion accuracy of the equipment, the flexibility of the material, and changes in the cutting conditions. Therefore, inspecting the cutting process and judging cutting defects are important technical steps to improve the cutting quality of polarizers. Existing methods for detecting defects in polarizer film cutting typically rely on images of the polarizer film after cutting to detect and evaluate the cutting edges, or on analyzing the cutting path based on the motion trajectory of the cutting equipment. These methods can reflect the cutting results to some extent, but due to the lack of acquisition of the actual pose changes of the polarizer film during the cutting process, they are difficult to accurately reflect the deviations caused by material stress, elastic deformation, or dynamic response during the cutting process. As a result, in some cutting conditions, the detection results are not completely consistent with the actual cutting process. On the other hand, integrating image data and cutting equipment motion data into the cutting process for comprehensive analysis is a development direction to improve the accuracy of cutting defect judgment. However, due to the time reference difference between image acquisition and the cutting equipment control system, and the certain response delay of the polarizer to the cutting tool, how to accurately correlate the motion data of the cutting tool with the actual pose change of the polarizer under unified spatial coordinates and time reference, and further analyze the corresponding relationship between the cutting trajectory and the edge result after cutting, still needs further improvement. Summary of the Invention

[0003] The purpose of this invention is to provide a defect detection method for polarizer cutting, so as to solve the problems mentioned in the background art.

[0004] This invention can be achieved through the following technical solution: a defect detection method for polarizer film cutting, comprising the following steps: Step 1: Before the polarizer cutting begins, spatial and temporal calibrations are performed on the cutting station to establish the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, and the synchronization reference signal of the cutting equipment is obtained. Before the cutting tool contacts the polarizer and the holding mechanism applies pressure to the polarizer, a reference image of the polarizer is acquired and the reference pose of the polarizer is determined. Step 2: When the cutting equipment performs the cutting action, the cutting event time when the cutting tool comes into contact with the polarizer is determined based on the synchronization reference signal, and a cutting event window is formed with the cutting event time as the center. Within the cutting event window, the imaging device is triggered to perform continuous image acquisition, and a timestamp is assigned to each frame of the image. Step 3: Based on the continuous images acquired within the cropping event window, and combined with the polarizer image features extracted before cropping begins, establish cross-frame correspondence, calculate the pose data of the polarizer at each time point within the cropping event window, and calculate the transient displacement sequence of the polarizer relative to the reference pose. Step 4: Synchronously acquire the motion data of the cutting tool within the cutting event window, and perform time alignment and correlation calculations on the motion data of the cutting tool and the transient displacement sequence to obtain the cutting trajectory of the cutting tool relative to the polarizer; Step 5: After the cropping is completed, acquire the image of the cropped polarizer and extract the cropping edge contour. Map the cropping trajectory to the coordinate system corresponding to the cropped image, and generate analysis data for cropping defect judgment based on the spatial relationship between the cropping trajectory and the cropping edge contour.

[0005] A further technical improvement of the present invention is that: when performing spatial and temporal calibration on the cutting station in step one, it includes: A calibration reference with known geometric relationships is set at the cutting station, and the imaging device is controlled to acquire calibration images containing the calibration reference. The mapping parameters between the imaging pixel coordinates and the cutting station plane coordinates are calculated based on the calibration image and stored. Before the cropping device performs the cropping action, the synchronization reference signal output by the cropping device is acquired, and the time offset parameter between the synchronization reference signal and the image acquisition is calculated by repeatedly acquiring the image exposure start time corresponding to the synchronization reference signal.

[0006] A further technical improvement of the present invention is that: step one, determining the reference pose of the polarizer, includes: The control holding mechanism applies pressure to the polarizer and holds it for a preset time; Multiple frames of polarizer images were continuously acquired under pressure, and the inter-frame displacement of the polarizer outline in adjacent images was calculated. A reference image is acquired when the inter-frame displacement continuously meets the stability condition. The polarizer's outline is extracted from the reference image, and its planar position and orientation in the cutting station are calculated as the reference pose.

[0007] A further technical improvement of the present invention is that: when forming the clipping event window in step two, it includes: Acquire the cutting tool motion data of the cutting equipment before and after the cutting event, and calculate the instantaneous speed of the cutting tool; A preset time range is determined based on the instantaneous velocity, wherein the preset time range includes a first time length before the cutting event and a second time length after the cutting event, and at least one of the first time length and the second time length is adjusted as the instantaneous velocity changes; The start and end times of the clipping event window are determined by taking the clipping event moment as the center and combining the first time length and the second time length; Within the cropping event window, the imaging device is triggered to continuously acquire images, and a timestamp is assigned to each frame.

[0008] A further technical improvement of the present invention is that: in step three, when calculating the transient displacement sequence, the following is included: Based on the polarizer image features extracted before cropping begins, a cross-frame correspondence is established for consecutive images within the cropping event window; Based on the cross-frame correspondence and the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, the polarizer pose data corresponding to each time point is calculated; The pose change is calculated based on pose data at adjacent time points, and the pose change is compared with the preset allowable pose change range per unit time. When the pose change exceeds the allowable pose change range, the pose data at the corresponding time point is determined to be pose data that does not meet the allowable change range, and continuous compensation processing is performed on the pose data based on adjacent pose data. The displacement and attitude change of the pose data after continuous compensation are calculated relative to the reference pose, and a transient displacement sequence is generated in time sequence.

[0009] A further technical improvement of the present invention is that, when establishing cross-frame correspondence, it includes: Extract polarizer image features from two adjacent frames within the cropping event window and generate feature description information; Calculate candidate matching pairs based on feature description information; By utilizing the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, candidate matching pairs are converted into displacement vectors under the cutting station plane coordinates; Consistency estimation of the displacement vector yields the overall motion direction of the polarizer, and candidate matching pairs whose motion direction deviates from the preset angle range or whose displacement amplitude deviates from the preset range are eliminated. The retained candidate matching pairs are determined as cross-frame correspondences.

[0010] A further technical improvement of the present invention is that: in step four, when performing time alignment and correlation calculation of motion data and transient displacement sequence, the following is included: Based on the timing of the clipping event, the clipping event window is divided into the entry phase, the stable clipping phase, and the exit phase. Set association calculation weights for different stages, so that the weight of the stable cutting stage is higher than that of the cutting-in stage and the cutting-out stage; Based on the flexible response characteristics of polarizer material, a preset time compensation amount is determined to characterize the response delay of polarizer to cutting tool action. The time compensation amount is used to characterize the time offset relationship between the cutting tool motion data time point and the corresponding physical response time point in the transient displacement sequence. Under the condition of combining stage weights and introducing time compensation, the following correlation calculation steps are performed: A1. Map the cutting tool motion data to a unified timeline within the cutting event window according to timestamps; A2. For each time point in the cutting tool motion data, determine the corresponding polarizer physical response time point according to the time compensation amount, and perform interpolation calculation between adjacent sampling time points in the transient displacement sequence based on the physical response time point to obtain the polarizer displacement data corresponding to the physical response time point. A3. In the plane coordinate system of the cutting station, the relative position of the cutting tool position data at the same time point and the polarizer displacement data obtained by interpolation are calculated to obtain the relative position of the cutting tool relative to the polarizer at each time point. A4. Based on the stage weights corresponding to each time point, the relative positions are weighted and connected in chronological order to generate the cutting trajectory of the cutting tool relative to the polarizer.

[0011] A further technical improvement of the present invention is that the determination of the preset time compensation amount includes the following steps: B1. When the cutting equipment is in the same cutting condition as actual production, continuous image acquisition is performed in the cutting event window, and the motion data of the cutting tool is acquired synchronously. B2. Based on the acquired continuous images, obtain the transient displacement sequence of the polarizer; Based on the acquired cutting tool motion data, a motion change sequence characterizing the change in the action of the cutting tool is calculated, and based on the transient displacement sequence, a displacement change sequence characterizing the change in the polarizer response is calculated. B3. Within the preset time delay search range, perform time delay correlation calculations on the motion change sequence and the displacement change sequence, and determine the time delay amount that maximizes the correlation as a candidate time compensation amount. B4. Under different cutting speed conditions or different holding conditions, repeat B1 to B3 to obtain multiple sets of time compensation candidate quantities, and establish the correspondence between the time compensation quantity and the corresponding working condition based on the cutting speed conditions or holding conditions. When performing the correlation calculation between the cutting tool motion data and the transient displacement sequence, the preset time compensation amount is selected or interpolated from the correspondence between the time compensation amount and the working condition according to the current cutting condition.

[0012] A further technical improvement of the present invention is that step five, when performing cutting defect analysis, includes: After the cutting is completed, the image of the cut polarizer is acquired, and the coordinates of the cut image are calibrated based on the mapping relationship between the plane coordinates of the cutting station and the pixel coordinates of the cut image. Extract the cropping edge contour from the cropped image after coordinate calibration, and establish the cropping edge curve according to the order of the contour points; The cropping trajectory is transformed into the coordinate system corresponding to the cropped image through a mapping relationship, so that the cropping trajectory and the cropping edge curve are in the same reference coordinates; Multiple sampling positions are set along the arc length direction of the cutting trajectory. The normal distance and tangential offset between the cutting trajectory and the cutting edge curve are calculated at each sampling position. The normal distance and tangential offset of each sampling position are combined into a spatial deviation sequence. The maximum deviation value, mean square deviation value, and number of deviation sign changes are calculated based on the spatial deviation sequence to generate analytical data that serves as input for judging cutting defects.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention establishes a mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates by performing spatial and temporal calibration of the cutting station before the polarizer cutting begins, and acquires the synchronous reference signal of the cutting equipment. This allows the image acquisition data and the cutting equipment motion data to be processed uniformly under the same spatial coordinate system and temporal reference. At the same time, the reference pose of the polarizer is determined in a stable state where the cutting tool has not yet contacted the polarizer and the holding mechanism applies pressure to the polarizer, providing a consistent reference basis for subsequent pose changes and trajectory calculations. Furthermore, this invention acquires continuous images within the cropping event window and establishes cross-frame correspondences, calculates the pose data of the polarizer at each time point during the cropping process, and generates a transient displacement sequence. This invention can truly reflect the dynamic displacement changes of the polarizer caused by force during the cropping process. On this basis, the invention introduces the judgment and continuity processing of pose changes, so that the transient displacement sequence has good stability and usability while maintaining temporal continuity, thereby providing a reliable input for the calculation of the cropping trajectory. On the other hand, this invention obtains the cutting trajectory of the cutting tool relative to the polarizer by performing time alignment and correlation calculation on the motion data and transient displacement sequence of the cutting tool, and maps the cutting trajectory to the coordinate system corresponding to the image after cutting after cutting, and performs spatial relationship analysis with the cutting edge contour; by calculating the spatial deviation sequence between the cutting trajectory and the cutting edge curve, and generating analysis data for cutting defect judgment, this invention can effectively correlate the dynamic behavior in the cutting process with the actual result after cutting, thereby improving the accuracy and objectivity of cutting defect judgment. Attached Figure Description

[0014] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 This is a schematic diagram of the method logic of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0017] Example 1 Please see Figure 1 As shown, the present invention provides a defect detection method for polarizer cutting, comprising: Step 1: Before the polarizer cutting begins, spatial and temporal calibrations are performed on the cutting station to establish the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, and the synchronization reference signal of the cutting equipment is obtained. Before the cutting tool contacts the polarizer and the holding mechanism applies pressure to the polarizer, a reference image of the polarizer is acquired and the reference pose of the polarizer is determined. By using spatial and temporal calibration, the image information acquired by the imaging device is unified with the actual physical position and time reference in the cutting station into the same reference system, providing a consistent spatial and temporal basis for all subsequent pose calculations, displacement calculations, and cutting trajectory calculations. Simultaneously, determining the reference pose of the polarizer in a stable state—before the cutting tool contacts the polarizer and the holding mechanism has applied pressure to it—avoids the influence of dynamic disturbances during the cutting process on the reference. This ensures that subsequent transient displacement sequences and cutting trajectories can all use this reference pose as a unified reference, thereby guaranteeing the comparability and consistency of the entire detection process.

[0018] When performing spatial and temporal calibration of the cutting station, the following are included: A calibration reference with known geometric relationships is set at the cutting station, and the imaging device is controlled to acquire calibration images containing the calibration reference. The mapping parameters between the imaging pixel coordinates and the cutting station plane coordinates are calculated based on the calibration image and stored. Before the cropping device performs the cropping action, the synchronization reference signal output by the cropping device is acquired, and the time offset parameter between the synchronization reference signal and the image acquisition is calculated by repeatedly acquiring the image exposure start time corresponding to the synchronization reference signal.

[0019] Specifically, a calibration reference is set within the imaging field of view of the cutting station. The calibration reference is a dot matrix structure with known geometric relationships and is fixedly installed on the plane of the cutting station, so that the geometric center of the dot matrix corresponds to the reference position of the cutting station. Second, the imaging device is controlled to acquire calibration images containing the calibration reference. During acquisition, the relative position of the imaging device and the cutting station remains unchanged, and multiple frames of calibration images are continuously acquired to reduce the impact of single-frame noise. Third, the pixel coordinates of each marker point of the dot matrix are extracted in each frame of calibration image, and the consistency of the pixel coordinates of the same marker point in multiple frames is checked. When the fluctuation of the pixel coordinates is less than a preset pixel threshold, the marker point is included in the calculation set. Finally, the set of pixel coordinates that have passed the consistency check is matched one-to-one with the station plane coordinates of each marker point in the calibration reference to form a spatial calibration dataset for solving the mapping parameters.

[0020] Based on the spatial calibration dataset, a correspondence between the imaging pixel coordinates and the cutting station plane coordinates is established for each marker point, and a mapping equation is established with the cutting station plane coordinates as the target coordinates. Second, the mapping parameters from the imaging pixel coordinates to the cutting station plane coordinates are solved using the least squares fitting method, ensuring that the mapping parameters include at least scale, rotation, and translation terms. Third, the reprojection error of the fitting results is evaluated, and the mapping parameters are deemed valid when the reprojection error is less than a preset spatial error threshold. Finally, the mapping parameters are associated with the working state parameters of the imaging device calibrated in this study and stored, so that subsequent pose calculation, trajectory mapping, and post-cropped image coordinate calibration can all use the same mapping parameters to complete the coordinate transformation.

[0021] After storing the mapping parameters, the synchronization reference signal output by the cropping device is acquired before the cropping device performs the cropping action, and the triggering method of the synchronization reference signal and the acquisition triggering method of the imaging device are kept fixed. Second, using the triggering time of the synchronization reference signal as a reference, the image exposure start time corresponding to each synchronization reference signal is acquired multiple times to form multiple sets of time difference samples between the triggering time of the synchronization reference signal and the image exposure start time. Third, the time difference samples are statistically calculated to obtain the time offset parameter, and the dispersion of the time difference samples is checked. When the dispersion is less than the preset time jitter threshold, the time offset parameter is confirmed to be stable. Finally, the time offset parameter and the aforementioned mapping parameters are bound to the same calibration batch parameter to ensure that the cropping event time determined based on the synchronization reference signal can establish a definite correspondence with the image acquisition time axis, thereby completing the unified configuration of spatial calibration and time calibration.

[0022] The steps for determining the reference pose of the polarizer include: The control holding mechanism applies pressure to the polarizer and holds it for a preset time; Multiple frames of polarizer images were continuously acquired under pressure, and the inter-frame displacement of the polarizer outline in adjacent images was calculated. A reference image is acquired when the inter-frame displacement continuously meets the stability condition. The polarizer's outline is extracted from the reference image, and its planar position and orientation in the cutting station are calculated as the reference pose.

[0023] Specifically, after completing the spatial calibration of the cutting station and establishing the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, the holding mechanism is controlled to apply pressure to the polarizer, so that the polarizer is in a state of pressure constraint in the cutting station, thereby limiting the free movement of the polarizer and reducing the impact of springback at the moment of establishing the pressure. Under the premise of keeping the pressure state unchanged, the fixed acquisition cycle of the imaging device is set and the continuous image acquisition process is started to continuously acquire multiple frames of polarizer images. The acquired images are numbered according to the acquisition sequence, and a one-to-one correspondence is established between the image number and the corresponding image acquisition timestamp to ensure that the time relationship between adjacent images is clear. Under the condition that the pressure state remains unchanged, the continuous image sequence is used as the input data for subsequent stability determination, so that the subsequent inter-frame displacement reflects the actual small motion state of the polarizer under pressure constraint.

[0024] In a continuous image sequence, the polarizer outline is extracted for each frame of the polarizer image. The outline is the identifiable boundary outline of the polarizer within the imaging field of view, used to characterize the positional change of the polarizer within the current field of view. Registration calculations are performed on the outlines in two adjacent frames to obtain the translation and rotation components that characterize the overall motion of the polarizer between the two adjacent frames. The translation and rotation components are then combined into an inter-frame displacement. The inter-frame displacements obtained in the acquisition order are used to construct an inter-frame displacement sequence, and a threshold range for stability determination is set. The threshold range includes a translation threshold and a rotation threshold, where the translation threshold is set to 0.05 mm and the rotation threshold is set to 0.01 degrees for example. It is clarified that the above inter-frame displacement calculation is only used for determining the stable state of the polarizer and not for the high-precision solution of the subsequent transient displacement sequence, thus functionally distinguishing it from the subsequent fine pose calculation process.

[0025] Stability determination is performed based on the inter-frame displacement sequence. The stability condition is set to the inter-frame displacement of multiple consecutive adjacent image pairs falling within the threshold range. For example, the number of consecutive pairs is set to 8 to avoid misjudgment caused by single noise or occasional disturbances. A sliding window method is used to continuously detect the inter-frame displacement sequence. When the inter-frame displacement within a certain sliding window continuously meets the stability condition, the polarizer image corresponding to the last frame of the sliding window is determined as the reference image. The validity check is performed on the outline corresponding to the determined reference image. The validity check includes determining whether the outline forms a closed boundary and whether the proportion of missing outline points is lower than a preset proportion. The preset proportion is set to 2% for example. After confirming the validity of the reference image, lock the reference image and record its corresponding image number and timestamp, which will be used to establish a connection with the time alignment operation in the subsequent cropping event window.

[0026] After locking the reference image, based on the polarizer outline extracted from the reference image and combined with the established mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, the planar position and orientation of the polarizer in the cutting station plane coordinates are calculated as the reference pose. The planar position includes the translation along the coordinate axis of the cutting station plane, and the orientation includes the rotation angle around the normal of the cutting station plane. The reference pose is associated with and stored with the image number and timestamp of the reference image, so that the pose data calculation in the subsequent cutting event window can be performed in the same spatial coordinate system and time reference. In subsequent steps, the displacement and orientation change of the polarizer pose relative to the reference pose at each time point are calculated based on the reference pose, thereby generating a transient displacement sequence. When the polarizer specifications change or the holding conditions are adjusted, the above reference pose determination process is repeated to ensure that the reference pose is always consistent with the current cutting conditions.

[0027] Step 2: When the cutting equipment performs the cutting action, the cutting event time when the cutting tool comes into contact with the polarizer is determined based on the synchronization reference signal, and a cutting event window is formed with the cutting event time as the center. Within the cutting event window, the imaging device is triggered to perform continuous image acquisition, and a timestamp is assigned to each frame of the image. By precisely determining the moment of contact between the cutting tool and the polarizer using a synchronous reference signal, the key physical behaviors during the cutting process are clearly calibrated on the timeline. A cutting event window is then formed centered on this moment, limiting image acquisition to a time range highly relevant to the cutting action. Continuous image acquisition within the cutting event window and assigning a timestamp to each frame ensures that subsequent polarizer pose changes, transient displacements, and cutting tool motion states can be correlated and linked on the same timeline, preventing interference from irrelevant images before and after the cutting process.

[0028] When the clipping event window is created, it includes: Acquire the cutting tool motion data of the cutting equipment before and after the cutting event, and calculate the instantaneous speed of the cutting tool; A preset time range is determined based on the instantaneous velocity, wherein the preset time range includes a first time length before the cutting event and a second time length after the cutting event, and at least one of the first time length and the second time length is adjusted as the instantaneous velocity changes; The start and end times of the clipping event window are determined by taking the clipping event moment as the center and combining the first time length and the second time length; Within the cropping event window, the imaging device is triggered to continuously acquire images, and a timestamp is assigned to each frame.

[0029] Specifically, during the cutting process, based on the cutting tool displacement data output by the cutting equipment, the motion data of the cutting tool during the cutting process is continuously acquired. The motion data covers the entire process of the cutting tool approaching the polarizer, contacting the polarizer, and completing the cutting. The motion data of the cutting tool is differentially calculated to obtain the instantaneous speed change of the cutting tool during the cutting process. The instantaneous speed is smoothed by a moving average method to reduce the impact of encoding jitter or instantaneous noise on speed judgment. During the cutting process, the moment of the cutting event when the cutting tool contacts the polarizer is determined based on the synchronization reference signal, and the smoothed instantaneous speed is time-aligned with the moment of the cutting event. The instantaneous speed corresponding to the moment of the cutting event is used as an important parameter characterizing the current cutting condition for adaptive determination of the time range of subsequent cutting event windows.

[0030] After the cutting event time is determined, the time range for covering the key cutting process is determined based on the instantaneous speed of the cutting tool near the cutting event time. This increases the time range corresponding to higher cutting tool speeds to avoid missing key transient processes under high-speed cutting conditions. The time range is divided into a time period before the cutting event time and a time period after the cutting event time, with at least one time period adjusted according to the instantaneous speed, thus giving the cutting event window asymmetrical coverage before and after the cutting event time. To prevent the time range from being too wide and introducing irrelevant data, upper and lower limits are set for the time range to ensure that the time range always remains within a stable time interval that the imaging device can continuously acquire. Based on the cutting event time and the determined time range, the start and end times of the cutting event window are calculated, thus completing the determination of the cutting event window on the time axis.

[0031] During the cropping operation, the imaging device continuously acquires images at a fixed acquisition cycle and assigns a corresponding timestamp to each acquired frame, giving each frame a clear position on the timeline. After the cropping event time and the start and end times of the cropping event window are determined, image frames whose timestamps fall within the cropping event window range are selected from the continuously acquired image sequence based on the timestamps of each frame. These selected image frames are then used as image data within the cropping event window. The selected image frames are then uniformly associated with the corresponding start and end times of the cropping event window, establishing a one-to-one correspondence between the cropping event window and the image data in the time dimension. The image data within the cropping event window is used as input for subsequent cross-frame correspondence establishment, pose data calculation, and transient displacement sequence generation, thereby ensuring that subsequent analysis processes are based on the cropping events that have already occurred for post-alignment and deterministic processing.

[0032] Step 3: Based on the continuous images acquired within the cropping event window, and combined with the polarizer image features extracted before cropping begins, establish cross-frame correspondence, calculate the pose data of the polarizer at each time point within the cropping event window, and calculate the transient displacement sequence of the polarizer relative to the reference pose. By utilizing the polarizer image features extracted before cropping begins as a stable matching basis, a reliable cross-frame correspondence is established between consecutive images within the cropping event window, enabling continuous tracking of the polarizer's spatial state at different time points during the cropping process. By calculating the pose data corresponding to each time point and comparing it with the reference pose determined in step one, a transient displacement sequence reflecting the polarizer's displacement due to force, stress release, or structural flexibility during cropping can be obtained. This provides a data foundation that truly reflects the transient behavior of cropping, supporting subsequent analysis of cropping deviations and trajectories.

[0033] Calculating the transient displacement sequence includes: Based on the polarizer image features extracted before cropping begins, a cross-frame correspondence is established for consecutive images within the cropping event window, including: Extract polarizer image features from two adjacent frames within the cropping event window and generate feature description information; Calculate candidate matching pairs based on feature description information; By utilizing the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, candidate matching pairs are converted into displacement vectors under the cutting station plane coordinates; Consistency estimation of the displacement vector yields the overall motion direction of the polarizer, and candidate matching pairs whose motion direction deviates from the preset angle range or whose displacement amplitude deviates from the preset range are eliminated. The retained candidate matching pairs are determined as cross-frame correspondences; Based on the cross-frame correspondence and the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, the polarizer pose data corresponding to each time point is calculated; The pose change is calculated based on pose data at adjacent time points, and the pose change is compared with the preset allowable pose change range per unit time. When the pose change exceeds the allowable pose change range, the pose data at the corresponding time point is determined to be pose data that does not meet the allowable change range, and continuous compensation processing is performed on the pose data based on adjacent pose data. The displacement and attitude change of the pose data after continuous compensation are calculated relative to the reference pose, and a transient displacement sequence is generated in time sequence.

[0034] Specifically, before cropping begins, polarizer image features are extracted and solidified as a matching benchmark. Preferred polarizer image features include edge structure features and texture point features. Edge structure features are enhanced and connected to obtain stable boundary segments, while texture point features are extracted using corner response to obtain candidate feature points. Within the cropping event window, polarizer image features of the same type as the matching benchmark are extracted frame by frame from consecutive images, and feature description information is generated for each feature to facilitate cross-frame matching. Based on the feature description information, candidate matching pairs are calculated between adjacent frames, and initial screening is performed in the imaging pixel coordinate domain to eliminate obviously unreasonable matching pairs. The candidate matching pairs that pass the initial screening are combined with the mapping relationship between the imaging pixel coordinates and the cropping station plane coordinates to convert them into a set of displacement vectors in the cropping station plane coordinates. Consistency estimation is performed on the displacement vector set to obtain the overall movement direction of the polarizer, thereby eliminating candidate matching pairs whose movement direction deviates from a preset angle range or whose displacement amplitude deviates from a preset range, thus obtaining the cross-frame correspondence for subsequent pose calculation.

[0035] Based on the cross-frame correspondence obtained in the previous section, at least four spatially distributed pairs of corresponding feature points within the same frame are selected as the solution input. The imaging pixel coordinates of the corresponding feature points are converted into a point set in the cutting station plane coordinates through the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates. Rigid body transformation fitting is performed on the point set in the cutting station plane coordinates to solve the polarizer pose data corresponding to each time point. The pose data includes at least the planar position and the in-plane attitude angle, where the planar position is expressed in millimeters and the in-plane attitude angle is expressed in degrees. The pose change is calculated based on the pose data of adjacent time points. The pose change includes the change in planar position between adjacent time points and the change in in-plane attitude angle between adjacent time points. The pose change is compared with the preset allowable pose change range per unit time. The unit time is consistent with the acquisition period of continuous images. For example, the acquisition period is 2 milliseconds. The allowable planar position change range is, for example, no more than 0.06 mm, and the allowable in-plane attitude angle change range is, for example, no more than 0.012 degrees. The pose data that meets the allowable pose change range is marked as valid pose data and enters the next continuous compensation process to ensure the continuity and usability of subsequent transient displacement sequences.

[0036] When the pose change exceeds the allowable pose change range, the pose data at the corresponding time point is determined to be pose data that does not meet the allowable change range, and the position of this time point between adjacent valid pose data is located on the time axis. Based on adjacent pose data, continuous compensation processing is performed on the pose data. Preferably, continuous compensation processing uses linear interpolation between the previous and subsequent valid pose data according to the time ratio to obtain compensated pose data. When multiple pose data that do not meet the allowable change range appear consecutively, piecewise interpolation constrained by the valid pose data at both ends is preferred, and a restriction is imposed on the interpolation result that the pose change does not exceed the allowable pose change range, so that the compensated pose data... The process is continuous in time and the changes are controlled. The pose data after continuous compensation is used to calculate the displacement and attitude change relative to the reference pose. The displacement is the difference between the planar position and the reference pose, and the attitude change is the difference between the attitude angle in the plane and the reference pose. The displacement and attitude change at each time point are matched one-to-one with the timestamp of that time point. The displacement and attitude change at each time point are combined in chronological order to generate a transient displacement sequence. The time span of the transient displacement sequence is consistent with the cutting event window, and the sampling density of the transient displacement sequence is consistent with the continuous image acquisition period. This provides definite input data for the time alignment and correlation calculation of the subsequent cutting tool motion data and the transient displacement sequence.

[0037] Step 4: Synchronously acquire the motion data of the cutting tool within the cutting event window, and perform time alignment and correlation calculations on the motion data of the cutting tool and the transient displacement sequence to obtain the cutting trajectory of the cutting tool relative to the polarizer; By synchronously acquiring the motion data of the cutting tool within the cutting event window and aligning it with the transient displacement sequence obtained in step three, a correspondence can be established between the motion state of the cutting tool and the pose change of the polarizer at the same point in time. Based on this, correlation calculations can eliminate the influence of the polarizer's own transient displacement on the cutting result, obtaining the true cutting trajectory of the cutting tool relative to the polarizer. This avoids misjudgments of the cutting position caused by relying solely on the tool's own motion path or the static result after cutting.

[0038] Specifically, during the cutting action of the cutting equipment, the motion data of the cutting tool in the cutting event window is acquired synchronously. The motion data includes at least the position information of the cutting tool in the plane coordinate system of the cutting station and the corresponding timestamp. At the same time, in the transient displacement sequence obtained in step three, the displacement and attitude change of the polarizer at each time point are also associated with the corresponding timestamp, so that the cutting tool motion data and the transient displacement sequence are under the same time reference.

[0039] During time alignment, the time axis of the cutting event window is used as a unified reference time axis. Based on the timestamps carried by the cutting tool motion data and the transient displacement sequence, time correspondence is performed between the two. When a certain time point of the cutting tool motion data is not completely consistent with the sampling time point in the transient displacement sequence, two adjacent sampling time points before and after that time point are selected in the transient displacement sequence, and the corresponding polarizer displacement and attitude change are interpolated according to the time ratio to obtain the polarizer displacement data corresponding to that cutting tool motion data time point.

[0040] After time alignment is completed, in the cutting station plane coordinate system, the position information of the cutting tool corresponding to the same time point is correlated with the polarizer displacement data obtained by interpolation. The correlation calculation is to combine the position data of the cutting tool in the cutting station plane coordinate system with the displacement of the polarizer relative to the reference pose to obtain the relative position of the cutting tool relative to the polarizer at that time point. The relative positions obtained at each time point in the cutting event window are arranged and connected in chronological order to form the cutting trajectory of the cutting tool relative to the polarizer.

[0041] Step 5: After the cropping is completed, acquire the image of the cropped polarizer and extract the cropping edge contour. Map the cropping trajectory to the coordinate system corresponding to the cropped image, and generate analysis data for cropping defect judgment based on the spatial relationship between the cropping trajectory and the cropping edge contour. By acquiring images of the cropped polarizer and extracting the cropping edge contours, the actual cropping result can be obtained. The cropping trajectory obtained in step four is mapped to the coordinate system corresponding to the cropped image, placing the cropping trajectory and the cropping edge contours under the same spatial reference, thus enabling direct analysis of their spatial relationship. Analysis data generated based on the spatial relationship between the cropping trajectory and the cropping edge contours allows for the mapping of dynamic behavior during the cropping process to the actual cropped result, providing a quantitative basis for the objective judgment of cropping defects.

[0042] Step five, when performing cutting defect analysis, includes: After the cutting is completed, the image of the cut polarizer is acquired, and the coordinates of the cut image are calibrated based on the mapping relationship between the plane coordinates of the cutting station and the pixel coordinates of the cut image. Extract the cropping edge contour from the cropped image after coordinate calibration, and establish the cropping edge curve according to the order of the contour points; The cropping trajectory is transformed into the coordinate system corresponding to the cropped image through a mapping relationship, so that the cropping trajectory and the cropping edge curve are in the same reference coordinates; Multiple sampling positions are set along the arc length direction of the cutting trajectory. The normal distance and tangential offset between the cutting trajectory and the cutting edge curve are calculated at each sampling position. The normal distance and tangential offset of each sampling position are combined into a spatial deviation sequence. The maximum deviation value, mean square deviation value, and number of deviation sign changes are calculated based on the spatial deviation sequence to generate analytical data that serves as input for judging cutting defects.

[0043] Specifically, after trimming, the polarizer is imaged to obtain an image of the trimmed polarizer. The clamping reference at the image acquisition position is consistent with that of the trimming station to reduce the impact of posture changes on coordinate correspondence. The coordinates of the trimmed polarizer image are calibrated using the mapping relationship between the plane coordinates of the trimming station and the pixel coordinates of the trimmed image. The mapping relationship is preferably represented by plane homography transformation parameters, and the mapping parameters are solved by at least four sets of corresponding point pairs distributed on the boundary of the trimming area, so that the plane coordinates of the trimming station can be mapped to the pixel coordinates of the trimmed image. The mapping parameters are applied to the trimmed polarizer image so that any plane coordinate point of the trimming station can obtain the corresponding image pixel coordinate position, thereby completing the coordinate calibration of the trimmed image under the same mapping relationship. The coordinate-calibrated trimmed image is used as a unified reference for subsequent trimming edge contour extraction and trajectory mapping.

[0044] In the cropped image calibrated by coordinates, the search region where the cropping edge is located is determined. The search region is expanded outward by a preset pixel width along the expected cropping boundary to cover possible edge offsets. Edge detection is performed within the search region to extract the cropping edge contour. Edge detection preferably uses gradient magnitude threshold and edge connection method to obtain a continuous contour point set, and the contour point set is smoothed at a small scale to reduce burr noise. The contour point set is sorted according to the order of the contour points. The sorting rule preferably uses the minimum Euclidean distance between adjacent points as the continuity criterion and combines curvature constraints to avoid jumping points, thereby forming an ordered cropping edge contour. A cropping edge curve is established based on the ordered contour points. The cropping edge curve preferably uses piecewise spline fitting to allow for subsequent interpolation at arbitrary arc length positions. For example, the curve interpolation step size is set to the pixel resolution conversion value corresponding to 0.01 mm.

[0045] Obtain the sequence of trajectory points in the plane coordinates of the cropping station, and organize the trajectory point sequence in chronological or trajectory order to form a continuous trajectory. Transform the cropping trajectory into the coordinate system corresponding to the cropped image through a mapping relationship. Specifically, apply mapping parameters to the plane coordinates of the cropping station for each trajectory point to obtain the corresponding image pixel coordinates, thereby obtaining the cropping trajectory in the image coordinate domain. Ensure that the transformed cropping trajectory and the cropping edge curve are in the same reference coordinate system. The same reference coordinate system is either the pixel coordinate domain in the coordinate system of the cropped image or a plane coordinate domain proportionally converted from the pixel coordinate domain, and ensure that the cropping trajectory and the cropping edge curve use the same mapping parameter system. Establish a comparable relationship between the cropping trajectory and the cropping edge curve under the same reference coordinate system, providing a prerequisite for subsequent sampling position setting along the arc length direction and spatial deviation sequence calculation.

[0046] Multiple sampling positions are set along the arc length of the cutting trajectory. The sampling position interval is preferably a fixed arc length interval, such as 0.1 mm to 0.5 mm, to ensure the sampling positions cover the entire cutting trajectory and maintain uniform sampling density. At each sampling position, the tangential and normal directions of the cutting trajectory are calculated. The tangential direction is preferably obtained by differentiating and normalizing the trajectory points of adjacent sampling positions, and the normal direction is obtained by rotating the tangential direction by 90 degrees, thereby establishing a local coordinate base for each sampling position. Corresponding points on the cutting edge curve are determined at each sampling position. The preferred method is to search for the curve point with the smallest Euclidean distance to the sampling position in the cutting edge curve, and interpolate between adjacent curve points to obtain a more accurate corresponding point. Then, the position from the sampling position to the corresponding point is used as the reference point. The shift vector is projected onto the normal and tangential directions to obtain the normal distance and tangential offset. The normal distance and tangential offset of each sampling position are then arranged in arc length order to form a spatial deviation sequence. Based on the spatial deviation sequence, the maximum deviation value, the mean square deviation value, and the number of deviation sign changes are calculated. The maximum deviation value is the maximum value of the absolute value of the normal distance, the mean square deviation value is the root mean square of the normal distance, and the number of deviation sign changes is obtained by counting the number of times the normal distance sign changes between adjacent sampling positions. For example, when the normal distance sequence has three zero-crossing changes from positive to negative or from negative to positive, the number of deviation sign changes is 3. The maximum deviation value, the mean square deviation value, and the number of deviation sign changes are combined to generate analysis data as input for judging cutting defects.

[0047] Example 2 A defect detection method for polarizer film cutting, comprising: Step 1: Before the polarizer cutting begins, spatial and temporal calibrations are performed on the cutting station to establish the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, and the synchronization reference signal of the cutting equipment is obtained. Before the cutting tool contacts the polarizer and the holding mechanism applies pressure to the polarizer, a reference image of the polarizer is acquired and the reference pose of the polarizer is determined. Step 2: When the cutting equipment performs the cutting action, the cutting event time when the cutting tool comes into contact with the polarizer is determined based on the synchronization reference signal, and a cutting event window is formed with the cutting event time as the center. Within the cutting event window, the imaging device is triggered to perform continuous image acquisition, and a timestamp is assigned to each frame of the image. Step 3: Based on the continuous images acquired within the cropping event window, and combined with the polarizer image features extracted before cropping begins, establish cross-frame correspondence, calculate the pose data of the polarizer at each time point within the cropping event window, and calculate the transient displacement sequence of the polarizer relative to the reference pose. Step 4: Synchronously acquire the motion data of the cutting tool within the cutting event window, and perform time alignment and correlation calculations on the motion data of the cutting tool and the transient displacement sequence to obtain the cutting trajectory of the cutting tool relative to the polarizer; Compared to Example 1, Example 2 includes the following steps when performing time alignment and correlation calculations between motion data and transient displacement sequences: Based on the timing of the clipping event, the clipping event window is divided into the entry phase, the stable clipping phase, and the exit phase. Set association calculation weights for different stages, so that the weight of the stable cutting stage is higher than that of the cutting-in stage and the cutting-out stage; Based on the flexible response characteristics of polarizer material, a preset time compensation amount is determined to characterize the response delay of polarizer to cutting tool action. The time compensation amount is used to characterize the time offset relationship between the cutting tool motion data time point and the corresponding physical response time point in the transient displacement sequence. Determining the preset time compensation amount includes the following steps: B1. When the cutting equipment is in the same cutting condition as actual production, continuous image acquisition is performed in the cutting event window, and the motion data of the cutting tool is acquired synchronously. B2. Based on the acquired continuous images, obtain the transient displacement sequence of the polarizer; Based on the acquired cutting tool motion data, a motion change sequence characterizing the change in the action of the cutting tool is calculated, and based on the transient displacement sequence, a displacement change sequence characterizing the change in the polarizer response is calculated. B3. Within the preset time delay search range, perform time delay correlation calculations on the motion change sequence and the displacement change sequence, and determine the time delay amount that maximizes the correlation as a candidate time compensation amount. B4. Under different cutting speed conditions or different holding conditions, repeat B1 to B3 to obtain multiple sets of time compensation candidate quantities, and establish the correspondence between the time compensation quantity and the corresponding working condition based on the cutting speed conditions or holding conditions. When performing the correlation calculation between the cutting tool motion data and the transient displacement sequence, the preset time compensation amount is selected or interpolated from the correspondence between the time compensation amount and the working condition according to the current cutting condition.

[0048] Under the condition of combining stage weights and introducing time compensation, the following correlation calculation steps are performed: A1. Map the cutting tool motion data to a unified timeline within the cutting event window according to timestamps; A2. For each time point in the cutting tool motion data, determine the corresponding polarizer physical response time point according to the time compensation amount, and perform interpolation calculation between adjacent sampling time points in the transient displacement sequence based on the physical response time point to obtain the polarizer displacement data corresponding to the physical response time point. A3. In the plane coordinate system of the cutting station, the relative position of the cutting tool position data at the same time point and the polarizer displacement data obtained by interpolation are calculated to obtain the relative position of the cutting tool relative to the polarizer at each time point. A4. Based on the stage weights corresponding to each time point, the relative positions are weighted and connected in chronological order to generate the cutting trajectory of the cutting tool relative to the polarizer.

[0049] Specifically, each time point within the cutting event window has a clear stage attribute. The entry stage corresponds to the approach and initial action range before the cutting tool contacts the polarizer; the stable cutting stage corresponds to the range where the cutting tool and polarizer continue to interact with each other and the relative interaction relationship changes little; and the exit stage corresponds to the exit range after the cutting tool leaves the polarizer. To reflect the higher contribution of the stable cutting stage to the reliability of the cutting trajectory, for example, the weight of the entry stage is set to 0.6, the weight of the stable cutting stage to 1.0, and the weight of the exit stage to 0.7, and each weight is bound to a time point within the cutting event window. After the weights are bound, the correlation calculation of each subsequent time point can directly reference the stage weight corresponding to that time point, thereby ensuring that the relative position sequence has stage-differentiated constraints in the time dimension. The stage division results and stage weights are used as the pre-input for introducing time compensation and performing correlation calculations to ensure that the subsequent cutting trajectory generation process is consistent with the physical process at the moment of the cutting event.

[0050] Under cutting conditions consistent with actual production, the motion data of the cutting tool is acquired within the cutting event window, and continuous image data for forming the transient displacement sequence of the polarizer is acquired simultaneously, ensuring that the changes in the cutting tool's action and the changes in the polarizer's response are within the same cutting condition context. Based on the acquired cutting tool motion data, a motion change sequence characterizing the changes in the cutting tool's action is calculated, and based on the transient displacement sequence, a displacement change sequence characterizing the changes in the polarizer's response is calculated. The motion change sequence and displacement change sequence reflect the changing trends of both over time during the cutting process. Within a preset time delay search range, time delay correlation calculations are performed on the motion change sequence and displacement change sequence, for example, within the range of 0 milliseconds to 20 milliseconds. Within a certain range, different time delays are iterated over in 0.5 millisecond steps, and the time delay that maximizes the correlation is determined as the candidate time compensation for the corresponding cutting condition. The above process is repeated under different cutting speed conditions or different holding conditions to obtain multiple sets of candidate time compensations. Based on the correspondence between the cutting condition and the candidate time compensations, when performing the correlation calculation between the cutting tool motion data and the transient displacement sequence, a preset time compensation is selected or interpolated according to the current cutting condition. The preset time compensation is used to characterize the time offset relationship between the cutting tool motion data time point and the corresponding physical response time point in the transient displacement sequence, thereby achieving causal alignment of the cutting tool action change and the polarizer response change on the time axis.

[0051] By incorporating stage weights and introducing a preset time compensation, the cutting tool motion data is arranged into a time-indexable tool position sequence within the cutting event window. For example, the time resolution of the unified time axis is set to 1 millisecond to cover short-term changes in the cutting process. For any time point t on the unified time axis, the physical response time point t′ of the polarizer is determined according to the preset time compensation Δt. The polarizer displacement data corresponding to t′ is obtained by interpolation between adjacent sampling time points t1 and t2 of the transient displacement sequence according to the time ratio. For example, when the sampling period of the transient displacement sequence is 2 milliseconds and t′ falls within t1... Between t1=10 milliseconds and t2=12 milliseconds, the displacement data at both ends are weighted according to (12−t′) / (12−10) and (t′−10) / (12−10) to obtain the interpolation result; the tool position sequence corresponding to each time point is paired with the interpolated polarizer displacement data under a unified time axis, so that the same time point has two types of inputs that can be directly used for relative position calculation; by interpolating the data, the risk of insufficient public disclosure of data that cannot be selected at the precise corresponding time point due to discrete sampling is avoided, thereby ensuring that the correlation calculation can still be stably reproduced under discrete time conditions.

[0052] In the plane coordinate system of the cutting station, the relative positions of the cutting tool position data and the polarizer displacement data obtained by interpolation at the same time point are calculated. The relative position is obtained by subtracting the polarizer displacement data from the tool position data. For example, if the tool position data is represented as (xt, yt) and the polarizer displacement data is represented as (Δx, Δy), then the relative position is (xt−Δx, yt−Δy), and a relative position sequence is formed according to a unified time axis. The relative position of each time point in the relative position sequence is weighted according to the stage to which that time point belongs, so that the relative position of the stable cutting stage has a higher proportion in the trajectory formation. The weighted relative positions are connected in time order to obtain the cutting trajectory, so that the cutting trajectory is consistent with the cutting event window on the time axis and can be mapped to the same reference space with the subsequent cut edge contour. The cutting trajectory of the cutting tool relative to the polarizer is output as the data input for subsequent cutting defect analysis, thus completing the time alignment and correlation calculation of motion data and transient displacement sequence under the joint constraints of stage weight and preset time compensation amount.

[0053] Compared to Example 1, Example 2 further specifies the time alignment and correlation calculation method between the cutting tool motion data and the transient displacement sequence in step four.

[0054] In Example 2, before performing the correlation calculation, the cutting event window is first divided into the cutting entry stage, the stable cutting stage, and the cutting exit stage based on the cutting event time. Correlation calculation weights are set for different stages, so that the stable cutting stage has a higher weight in the cutting trajectory formation process, thereby making the cutting trajectory more concentrated in reflecting the stable interaction relationship between the polarizer and the cutting tool during the cutting process.

[0055] Meanwhile, in Example 2, when performing the correlation calculation between the cutting tool motion data and the transient displacement sequence, a preset time compensation amount is introduced to characterize the response delay of the polarizer to the cutting tool. The preset time compensation amount characterizes the time offset relationship between the cutting tool motion data time point and the corresponding physical response time point in the transient displacement sequence. In Example 2, the correspondence between the cutting tool motion data and the transient displacement sequence on the time axis is no longer limited to data points with the same timestamp. Instead, the physical response time point of the polarizer is determined based on the preset time compensation amount, and interpolation calculations are performed between adjacent sampling time points in the transient displacement sequence to obtain the polarizer displacement data corresponding to the causal relationship with the cutting tool's action.

[0056] Furthermore, the introduction of the preset time compensation amount in Embodiment 2 does not change the timestamps corresponding to each pose data in the transient displacement sequence, nor does it regenerate the transient displacement sequence itself. Instead, in the process of correlation calculation between the cutting tool motion data and the transient displacement sequence, the adjustment of the time point selection method establishes a more realistic correspondence between the change in the action of the cutting tool and the change in the pose response of the polarizer on the time axis.

[0057] Furthermore, compared to the method in Example 1 that directly generates the cutting trajectory based on the time alignment result, Example 2, after obtaining the relative position of the cutting tool relative to the polarizer at each time point, further weights the relative positions according to the correlation calculation weight corresponding to the stage to which each time point belongs, and connects the weighted relative positions in chronological order to generate the cutting trajectory of the cutting tool relative to the polarizer. Through this method, Example 2 can more accurately reflect the actual force and displacement state of the polarizer in the stable cutting stage while maintaining the continuity and temporal consistency of the transient displacement sequence.

[0058] Step 5: After the cropping is completed, acquire the image of the cropped polarizer and extract the cropping edge contour. Map the cropping trajectory to the coordinate system corresponding to the cropped image, and generate analysis data for cropping defect judgment based on the spatial relationship between the cropping trajectory and the cropping edge contour.

[0059] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A defect detection method for cutting polarizing film, characterized in that, include: Step 1: Before the polarizer cutting begins, spatial and temporal calibrations are performed on the cutting station to establish the mapping relationship between the imaging pixel coordinates and the cutting station plane coordinates, and the synchronization reference signal of the cutting equipment is obtained. Before the cutting tool contacts the polarizer and the holding mechanism applies pressure to the polarizer, a reference image of the polarizer is acquired and the reference pose of the polarizer is determined. Step 2: When the cutting equipment performs the cutting action, the cutting event time when the cutting tool comes into contact with the polarizer is determined based on the synchronization reference signal, and a cutting event window is formed with the cutting event time as the center. Within the cutting event window, the imaging device is triggered to perform continuous image acquisition, and a timestamp is assigned to each frame of the image. Step 3: Based on the continuous images acquired within the cropping event window, and combined with the polarizer image features extracted before cropping begins, establish cross-frame correspondence, calculate the pose data of the polarizer at each time point within the cropping event window, and calculate the transient displacement sequence of the polarizer relative to the reference pose. Step 4: Synchronously acquire the motion data of the cutting tool within the cutting event window, and perform time alignment and correlation calculations on the motion data of the cutting tool and the transient displacement sequence to obtain the cutting trajectory of the cutting tool relative to the polarizer; Step 5: After the cropping is completed, acquire the image of the cropped polarizer and extract the cropping edge contour. Map the cropping trajectory to the coordinate system corresponding to the cropped image, and generate analysis data for cropping defect judgment based on the spatial relationship between the cropping trajectory and the cropping edge contour.

2. The defect detection method for polarizer cutting according to claim 1, characterized in that, Step 1 involves spatial and temporal calibration of the cutting station, including: A calibration reference with known geometric relationships is set at the cutting station, and the imaging device is controlled to acquire calibration images containing the calibration reference. The mapping parameters between the imaging pixel coordinates and the cutting station plane coordinates are calculated based on the calibration image and stored. Before the cropping device performs the cropping action, the synchronization reference signal output by the cropping device is acquired, and the time offset parameter between the synchronization reference signal and the image acquisition is calculated by repeatedly acquiring the image exposure start time corresponding to the synchronization reference signal.

3. The defect detection method for polarizer cutting according to claim 2, characterized in that, Step one, determining the reference pose of the polarizer, includes: The control holding mechanism applies pressure to the polarizer and holds it for a preset time; Multiple frames of polarizer images were continuously acquired under pressure, and the inter-frame displacement of the polarizer outline in adjacent images was calculated. A reference image is acquired when the inter-frame displacement continuously meets the stability condition. The polarizer's outline is extracted from the reference image, and its planar position and orientation in the cutting station are calculated as the reference pose.

4. The defect detection method for polarizer cutting according to claim 1, characterized in that, Step two, when creating the clipping event window, includes: Acquire the cutting tool motion data of the cutting equipment before and after the cutting event, and calculate the instantaneous speed of the cutting tool; A preset time range is determined based on the instantaneous velocity, wherein the preset time range includes a first time length before the cutting event and a second time length after the cutting event, and at least one of the first time length and the second time length is adjusted as the instantaneous velocity changes; The start and end times of the clipping event window are determined by taking the clipping event moment as the center and combining the first time length and the second time length; Within the cropping event window, the imaging device is triggered to continuously acquire images, and a timestamp is assigned to each frame.

5. The defect detection method for polarizer cutting according to claim 1, characterized in that, Step 3, calculating the transient displacement sequence, includes: Based on the polarizer image features extracted before cropping begins, a cross-frame correspondence is established for consecutive images within the cropping event window; Based on the cross-frame correspondence and the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, the polarizer pose data corresponding to each time point is calculated; The pose change is calculated based on pose data at adjacent time points, and the pose change is compared with the preset allowable pose change range per unit time. When the pose change exceeds the allowable pose change range, the pose data at the corresponding time point is determined to be pose data that does not meet the allowable change range, and continuous compensation processing is performed on the pose data based on adjacent pose data. The displacement and attitude change of the pose data after continuous compensation are calculated relative to the reference pose, and a transient displacement sequence is generated in time sequence.

6. The defect detection method for polarizer cutting according to claim 5, characterized in that, When establishing cross-frame correspondence, the following is included: Extract polarizer image features from two adjacent frames within the cropping event window and generate feature description information; Calculate candidate matching pairs based on feature description information; By utilizing the mapping relationship between imaging pixel coordinates and cutting station plane coordinates, candidate matching pairs are converted into displacement vectors under the cutting station plane coordinates; Consistency estimation of the displacement vector yields the overall motion direction of the polarizer, and candidate matching pairs whose motion direction deviates from the preset angle range or whose displacement amplitude deviates from the preset range are eliminated. The retained candidate matching pairs are determined as cross-frame correspondences.

7. A defect detection method for polarizer cutting according to claim 5, characterized in that, Step four, which involves time alignment and correlation calculations between motion data and transient displacement sequences, includes: Based on the timing of the clipping event, the clipping event window is divided into the entry phase, the stable clipping phase, and the exit phase. Set association calculation weights for different stages, so that the weight of the stable cutting stage is higher than that of the cutting-in stage and the cutting-out stage; Based on the flexible response characteristics of polarizer material, a preset time compensation amount is determined to characterize the response delay of polarizer to cutting tool action. The time compensation amount is used to characterize the time offset relationship between the cutting tool motion data time point and the corresponding physical response time point in the transient displacement sequence. Under the condition of combining stage weights and introducing time compensation, the following correlation calculation steps are performed: A1. Map the cutting tool motion data to a unified timeline within the cutting event window according to timestamps; A2. For each time point in the cutting tool motion data, determine the corresponding polarizer physical response time point according to the time compensation amount, and perform interpolation calculation between adjacent sampling time points in the transient displacement sequence based on the physical response time point to obtain the polarizer displacement data corresponding to the physical response time point. A3. In the plane coordinate system of the cutting station, the relative position of the cutting tool position data at the same time point and the polarizer displacement data obtained by interpolation are calculated to obtain the relative position of the cutting tool relative to the polarizer at each time point. A4. Based on the stage weights corresponding to each time point, the relative positions are weighted and connected in chronological order to generate the cutting trajectory of the cutting tool relative to the polarizer.

8. A defect detection method for polarizer cutting according to claim 7, characterized in that, Determining the preset time compensation amount includes the following steps: B1. When the cutting equipment is in the same cutting condition as actual production, continuous image acquisition is performed in the cutting event window, and the motion data of the cutting tool is acquired synchronously. B2. Based on the acquired continuous images, obtain the transient displacement sequence of the polarizer; Based on the acquired cutting tool motion data, a motion change sequence characterizing the change in the action of the cutting tool is calculated, and based on the transient displacement sequence, a displacement change sequence characterizing the change in the polarizer response is calculated. B3. Within the preset time delay search range, perform time delay correlation calculations on the motion change sequence and the displacement change sequence, and determine the time delay amount that maximizes the correlation as a candidate time compensation amount. B4. Under different cutting speed conditions or different holding conditions, repeat B1 to B3 to obtain multiple sets of time compensation candidate quantities, and establish the correspondence between the time compensation quantity and the corresponding working condition based on the cutting speed conditions or holding conditions. When performing the correlation calculation between the cutting tool motion data and the transient displacement sequence, the preset time compensation amount is selected or interpolated from the correspondence between the time compensation amount and the working condition according to the current cutting condition.

9. A defect detection method for polarizer cutting according to claim 1, characterized in that, Step five, when performing cutting defect analysis, includes: After the cutting is completed, the image of the cut polarizer is acquired, and the coordinates of the cut image are calibrated based on the mapping relationship between the plane coordinates of the cutting station and the pixel coordinates of the cut image. Extract the cropping edge contour from the cropped image after coordinate calibration, and establish the cropping edge curve according to the order of the contour points; The cropping trajectory is transformed into the coordinate system corresponding to the cropped image through a mapping relationship, so that the cropping trajectory and the cropping edge curve are in the same reference coordinates; Multiple sampling positions are set along the arc length direction of the cutting trajectory. The normal distance and tangential offset between the cutting trajectory and the cutting edge curve are calculated at each sampling position. The normal distance and tangential offset of each sampling position are combined into a spatial deviation sequence. The maximum deviation value, mean square deviation value, and number of deviation sign changes are calculated based on the spatial deviation sequence to generate analytical data that serves as input for judging cutting defects.

Images