Transparent film abnormality detection multi-pose micro-step shooting method and device

Abstract

The application discloses a transparent film abnormality detection multi-pose micro-step shooting method and device. The method comprises the following steps: collecting a global vision image of a transparent film, identifying an abnormal candidate target area with defects in the global vision image, and extracting defect feature parameters of the abnormal candidate target area; taking the center point of the abnormal candidate target area as a reference, constructing a local pose search space according to the defect feature parameters, and the local pose search space contains multiple groups of shooting poses; performing priority ranking on the shooting poses in the local pose search space based on a preset pose value function, and obtaining a target pose sequence; and driving an execution mechanism to perform hierarchical micro-step shooting on the abnormal candidate target area according to the target pose sequence.

Description

Technical Field

[0001] This application relates to the field of automated shooting control technology, and more specifically, to a method and apparatus for multi-pose micro-stepping shooting for transparent film anomaly detection. Background Technology

[0002] In the quality inspection process of thin film products such as optically transparent films, polarizers, and functional coating substrates, it is necessary to identify minute anomalies such as scratches on the film surface, internal impurities, and dot-like particles. The light reflection and scattering imaging effects of thin film defects vary significantly under different observation angles. Only multi-angle imaging can fully restore the defect morphology and distinguish the location of the defect in the film layer.

[0003] In the existing technology, automated visual imaging solutions for transparent film inspection are mainly divided into two types: one uses fixed-angle combined imaging, and pre-sets a limited observation angle to complete the full-area scanning; the other uses a uniform and dense full-area scanning mode, and controls the robotic arm to evenly distribute a large number of observation postures across the entire area of ​​the film surface to complete the imaging.

[0004] However, the fixed-angle combination mode lacks targeted viewpoint selection, and some tiny defects cannot form effective optical features, making it difficult to accurately locate defects; the global uniform scanning mode will bring high time and mechanical movement costs. Summary of the Invention

[0005] The main objective of this application is to provide a multi-pose micro-stepping imaging method and device for transparent film anomaly detection, in order to solve the technical problems of insufficient defect recognition accuracy, excessive detection time and mechanical wear in existing imaging schemes. This achieves the technical effect of significantly reducing detection time and reducing equipment movement wear while ensuring the recognition of minute defects.

[0006] To achieve the above objectives, a first aspect of this application proposes a multi-pose microstepping imaging method for anomaly detection of transparent films, comprising: acquiring a global visual image of the transparent film, identifying anomaly candidate target regions with defects in the global visual image, and extracting defect feature parameters of the anomaly candidate target regions; constructing a local pose search space based on the center point of the anomaly candidate target regions and the defect feature parameters, the local pose search space containing multiple sets of shooting poses; prioritizing the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; driving an actuator to perform layered microstepping imaging on the anomaly candidate target regions according to the target pose sequence; wherein, the layered microstepping imaging refers to performing multi-angle deflection sampling based on a reference pose, and dynamically increasing or decreasing the sampling pose according to the imaging gain; the preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target positioning gain, and mechanical motion cost.

[0007] According to the multi-pose microstepping imaging method for anomaly detection of transparent film provided in this application, the defect feature parameters include scale parameters, target orientation, and anomaly category; the extraction of defect feature parameters of the candidate anomaly target region includes: locating the center coordinates of each candidate anomaly target region, determining the target bounding box size, and obtaining the scale parameters; extracting the extension direction of the strip-shaped defect to obtain the target orientation; and distinguishing the anomaly category according to the defect optical imaging characteristics, wherein the anomaly category includes dot-like particles, scratches, and internal impurities.

[0008] According to the multi-pose micro-stepping imaging method for transparent membrane anomaly detection provided in this application, the method constructs a local pose search space based on the center point of the candidate anomaly target region and the defect feature parameters, including: using the center point of the candidate anomaly target region as the imaging reference origin, and setting a safe observation distance range between the camera and the membrane surface in combination with the scale parameters; defining the micro-angle deflection boundary of the camera relative to the membrane surface normal by combining the joint motion limit of the actuator and the camera stable imaging constraint; setting a differentiated angle sampling density according to the anomaly category and the target orientation to generate candidate observation angles suitable for imaging and recognition of various defects; generating multiple sets of candidate shooting poses around the defect region within the preset local constraint space according to a preset step size, and summarizing all valid shooting poses after legality verification to obtain the local pose search space.

[0009] According to the multi-pose microstepping shooting method for transparent membrane anomaly detection provided in this application, the step of prioritizing the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence includes: substituting each group of shooting poses into the preset pose value function to calculate the corresponding comprehensive evaluation value; arranging all shooting poses in descending order according to the comprehensive evaluation value from high to low; and removing shooting poses with a comprehensive evaluation value lower than a preset threshold to generate the target pose sequence.

[0010] According to the multi-pose microstepping imaging method for transparent membrane anomaly detection provided in this application, the step of driving the actuator to perform layered microstepping imaging on the anomaly candidate target region according to the target pose sequence includes: controlling the actuator to adjust to the reference imaging pose corresponding to the membrane surface normal, acquiring an initial reference image with the reference imaging pose, and extracting the corresponding first defect imaging gain; taking the reference imaging pose as the center, sequentially switching each effective imaging pose in the target pose sequence according to a preset step size for step-by-step imaging acquisition; calculating the second defect imaging gain of the image under each effective imaging pose in real time; if the second defect imaging gain is higher than the first defect imaging gain, adding a candidate imaging pose around the corresponding effective imaging pose.

[0011] According to the multi-pose microstepping imaging method for transparent film anomaly detection provided in this application, the method further includes: if the second defect imaging gain of the image under N consecutive effective shooting poses is lower than the first defect imaging gain, then after all effective shooting poses have been traversed, the method switches to the next anomaly candidate target area for shooting; wherein, N is a positive integer greater than 1.

[0012] This application also provides a multi-pose microstepping imaging device for anomaly detection of transparent films, comprising the following modules: an acquisition module and a processing module; the acquisition module is used to acquire a full-field visual image of the transparent film, identify anomaly candidate target regions with defects in the full-field visual image, and extract defect feature parameters of the anomaly candidate target regions; the processing module is used to construct a local pose search space based on the center point of the anomaly candidate target region and the defect feature parameters, the local pose search space containing multiple shooting poses; prioritize the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; drive the actuator to perform layered microstepping imaging on the anomaly candidate target region according to the target pose sequence; wherein, the layered microstepping imaging refers to multi-angle deflection sampling based on a reference pose, and dynamically increasing or decreasing the sampling pose according to the imaging gain; the preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target positioning gain, and mechanical motion cost.

[0013] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-pose microstepping imaging method for transparent film anomaly detection as described above.

[0014] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-pose microstepping imaging method for transparent film anomaly detection as described above.

[0015] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the multi-pose microstepping imaging method for transparent film anomaly detection as described above.

[0016] The technical solutions provided by the embodiments of this application can include the following beneficial effects: Since a full-area visual image of the transparent film is acquired first, abnormal candidate target areas are identified, and corresponding defect feature parameters are extracted, the location of film surface defects can be accurately locked and the inherent properties of the defects themselves can be obtained. Therefore, accurate data support can be provided for subsequent pose space construction, avoiding blindly conducting full-area imaging; Since the center point of the abnormal candidate target area is used as a reference, and a local pose search space is constructed specifically based on the defect feature parameters, candidate shooting poses are only arranged around the defect area, eliminating the need for indiscriminate full-area pose scanning of the entire transparent film. Therefore, a large number of invalid shooting actions in the full-area uniform scanning mode are completely avoided, effectively reducing invalid movement of the actuator and lowering detection time costs and mechanical movement losses; Since the use of defect imaging... The system employs a preset attitude value function that prioritizes candidate imaging attitudes based on at least one of the following indicators: benefit from imaging, layer determination benefit, point target location benefit, and mechanical movement cost. This function prioritizes and selects the optimal imaging sequence, enabling targeted selection of observation angles with better imaging effects and lower movement costs. This overcomes the shortcomings of traditional fixed-angle combined imaging, which suffers from fixed and untargeted perspectives, and improves the clarity of micro-defect imaging, defect location accuracy, and film layer differentiation capabilities. By performing layered micro-step imaging according to the target attitude sequence and completing multi-angle deflection sampling based on the baseline attitude, the system can dynamically increase or decrease sampling attitudes in conjunction with real-time imaging benefits. Therefore, it can supplement sampling for high-value imaging angles to ensure detection accuracy and reduce the number of samples for attitudes with no imaging benefits, achieving a dynamic balance between detection accuracy and detection efficiency. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application. In the drawings: Figure 1 A flowchart illustrating the multi-pose microstepping imaging method for detecting anomalies in transparent films provided in this application; Figure 2 This is a schematic diagram of the structure of the multi-pose microstepping imaging device for detecting anomalies in transparent films provided in this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] In this application, the terms "upper," "lower," "left," "right," "front," "rear," "top," "bottom," "inner," "outer," "middle," "vertical," "horizontal," "lateral," and "longitudinal" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are primarily for the purpose of better describing this application and its embodiments, and are not intended to limit the indicated device, element, or component to having a specific orientation, or to be constructed and operated in a specific orientation.

[0021] Furthermore, in addition to indicating location or positional relationship, some of the aforementioned terms may also have other meanings. For example, the term "above" may also be used in some cases to indicate a certain dependency or connection relationship. Those skilled in the art can understand the specific meaning of these terms in this application based on the specific circumstances.

[0022] Furthermore, the terms "installation," "setup," "equipped with," "connection," "linked," and "socketing" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or an internal connection between two devices, components, or parts. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0023] This application describes some exemplary embodiments for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0024] like Figure 1 As shown, this application provides a method for multi-pose microstepping imaging to detect anomalies in transparent films. This method can be applied to a device for multi-pose microstepping imaging to detect anomalies in transparent films. The method may include steps S101-S104: S101. The multi-pose micro-stepping imaging device for transparent film anomaly detection acquires a full-field visual image of the transparent film, identifies anomaly candidate target regions with defects in the full-field visual image, and extracts the defect feature parameters of the anomaly candidate target regions.

[0025] Specifically, the transparent membrane anomaly detection multi-pose micro-stepping imaging device can acquire full-coverage images of the transparent membrane in a uniform conveying state using a high-definition industrial camera, obtaining a full-area visual image of the membrane surface without blind spots; then, it calls the trained visual defect detection model to perform pixel-level feature traversal on the full-area visual image, segmenting and marking defect areas such as abrupt grayscale changes and abnormal reflections on the membrane surface, in order to determine all candidate anomaly target areas; then, for each segmented candidate anomaly target area, it performs backend parameter calculation to extract the defect feature data required for matching subsequent pose planning.

[0026] It should be noted that, since defects can be pre-identified by full-area imaging and then the specific features of the defects can be extracted in a targeted manner, there is no need to set up shooting points on the entire transparent film in advance. Therefore, the precise location of the defects and their inherent attributes can be locked in advance, avoiding blind scanning of the entire actuator from the source. This provides reliable and one-to-one original input data for the subsequent adaptive construction of a dedicated local posture search space, ensuring that the subsequent posture generation fits the real defect shape.

[0027] Optionally, the defect feature parameters include scale parameters, target orientation, and anomaly category; the extraction of defect feature parameters of the anomaly candidate target region includes: locating the center coordinates of each anomaly candidate target region, determining the target bounding box size, and obtaining the scale parameters; extracting the extension direction of the strip-shaped defect to obtain the target orientation; and distinguishing the anomaly category based on the defect optical imaging characteristics, wherein the anomaly category includes dot-like particles, scratches, and internal impurities.

[0028] Specifically, the multi-pose micro-stepping imaging device for transparent film anomaly detection can use an image contour detection algorithm to select the circumscribed rectangle of the candidate target area for anomalies, solve for the geometric center of the rectangle as the coordinates of the target center, use the length and width of the circumscribed rectangle to represent the actual size of the defect, and generate corresponding scale parameters. For long linear defects such as scratches, the Hough transform can be used to fit the extension axis of the defect to obtain the true direction angle of the defect. Combining the differences in specular reflection and diffuse reflection imaging of different defects in the transparent film, typical film surface anomalies such as point particles, linear scratches, and impurities inside the interlayer can be distinguished to complete the automatic classification of defects.

[0029] It should be noted that by accurately extracting three types of parameters—scale parameters, target orientation, and anomaly category—in different dimensions, differentiated adaptation for different defects can be achieved. This allows large-sized defects to be matched with a wider range of pose adjustment, strip-shaped defects to be matched with the deflection shooting angle of vertical textures, and dot-shaped particles to be matched with the observation pose of highlight suppression. This completely solves the problem that a unified pose space cannot adapt to the detection of multiple types of defects and improves the targeting of subsequent candidate pose generation.

[0030] S102. The transparent film anomaly detection multi-pose microstepping imaging device uses the center point of the anomaly candidate target area as a reference and constructs a local pose search space based on the defect feature parameters.

[0031] The aforementioned local pose search space includes multiple shooting poses.

[0032] Optionally, the step of constructing a local pose search space based on the center point of the candidate abnormal target region and the defect feature parameters includes: using the center point of the candidate abnormal target region as the imaging reference origin, and setting a safe observation distance range between the camera and the membrane surface in conjunction with the scale parameters; defining the micro-angle deflection boundary of the camera relative to the normal of the membrane surface in conjunction with the joint motion limit of the actuator and the camera stable imaging constraint; setting a differentiated angle sampling density according to the abnormality category and the target orientation to generate candidate observation angles suitable for imaging and recognition of various defects; generating multiple sets of candidate shooting postures around the defect region within the preset local constraint space according to a preset step size, and summarizing all valid shooting postures after legality verification to obtain the local pose search space.

[0033] Specifically, the multi-pose micro-stepping imaging device for transparent film anomaly detection can first read the identified candidate anomaly regions and their corresponding defect feature parameters. Using the center point of each candidate anomaly region as the exclusive imaging reference origin, and combining the acquired scale parameters, it comprehensively calculates the safe observation distance range between the camera lens and the transparent film surface, avoiding issues such as lens contact with the film material and image defocusing. Then, it retrieves its stored actuator joint motion limit parameters and camera imaging stabilization time threshold to limit the allowable micro-angle deflection boundary of the camera relative to the film surface normal, directly eliminating observation positions that exceed the hardware's motion range and cannot achieve stable imaging. Next, combining the current anomaly type and defect target orientation, it targets point particles, scratches, and internal impurities. Differentiated angle sampling densities are set according to the optical reflection and scattering imaging laws of each defect. For defects with indistinct imaging features and easy blind spots, the sampling angles are denser, while the observation points are sparsely arranged for defects with low angle sensitivity. This generates candidate observation angles suitable for accurate identification of various defects in batches. Within the preset local constraint space formed by the joint constraints of distance and angle, multiple sets of candidate shooting postures surrounding the defect area are generated in batches according to a unified preset step size. Finally, all candidate shooting postures are validated for legality, such as motion interference, field of view coverage, and joint accessibility. Invalid postures with safety hazards or imaging failures are screened out. All valid shooting postures that pass the validation are integrated and collected to finally generate the local posture search space corresponding to a single defect area.

[0034] It should be noted that by establishing a local imaging benchmark at the defect center point and limiting the safe observation distance range, and constructing a posture space only around a small area of ​​the defect, a large number of candidate postures with no shooting value can be reduced from the source, shortening the subsequent mechanical movement distance and posture switching waiting time. By pre-locking the camera deflection boundary based on its own hardware parameters, invalid postures such as robotic arm over-limit, singular configurations, and imaging jitter can be filtered out during the posture generation stage, avoiding the risk of equipment jamming, alarms, and collisions during operation and ensuring continuous detection. Adaptively adjusting the angle sampling density according to the defect morphology and type can maximize the defect display information with a limited number of shots. Completing hardware and imaging legality verification in advance during the posture generation stage and eliminating interference, over-limit, and blurry invalid postures in advance can prevent equipment stoppages and repeated back-and-forth movements during operation, reduce equipment wear and tear, and improve the long-term operational stability of the automated detection system.

[0035] S103. The transparent membrane anomaly detection multi-pose microstepping shooting device prioritizes the shooting poses in the local pose search space based on a preset pose value function to obtain the target pose sequence.

[0036] The aforementioned preset posture value function includes at least one of the following: defect imaging benefit, layer determination benefit, point target localization benefit, and mechanical motion cost. Defect imaging benefit refers to the quantified score of the integrity of the membrane defect edge, defect contrast, and reflection suppression effect under the current shooting posture; a higher score indicates clearer defect morphology imaging from that perspective. Layer determination benefit refers to the score of the ability to distinguish different membrane layer positions of surface defects and membrane interlayer defects based on parallax differences under the current shooting posture. Point target localization benefit refers to the score of the ability to accurately capture tiny point particle defects, suppress noise interference, and complete sub-pixel localization under the current shooting posture. Mechanical motion cost refers to the quantified cost value of the movement stroke, adjustment time, and joint wear required for the actuator to switch to the current shooting posture; a smaller value indicates higher posture switching efficiency.

[0037] Optionally, the step of prioritizing the shooting postures in the local posture search space based on a preset posture value function to obtain a target posture sequence includes: substituting each group of shooting postures into the preset posture value function to calculate the corresponding comprehensive evaluation value; arranging all shooting postures in descending order according to the comprehensive evaluation value from high to low; and removing shooting postures with a comprehensive evaluation value lower than a preset threshold to generate the target posture sequence.

[0038] Specifically, the multi-pose micro-stepping imaging device for transparent film anomaly detection can retrieve all effective shooting poses within the local pose search space one by one, and score them separately according to four evaluation indicators: defect imaging benefit, layer determination benefit, point target positioning benefit, and mechanical motion cost. The scores of each sub-item are weighted and fused together with preset weight coefficients to solve the comprehensive evaluation value corresponding to each shooting pose. Based on the comprehensive evaluation value, all shooting poses are sorted from high to low. Low-scoring poses with poor imaging effect, high motion loss, and no detection value are filtered out, and high-quality effective shooting poses are retained. Finally, an ordered sequence of target poses is output.

[0039] It should be noted that by pre-setting the attitude value function and coupling the dual indicators of imaging detection benefits and mechanical motion loss in multiple dimensions, we can prioritize the retention of high imaging angles that are beneficial to defect identification and layer resolution, while avoiding inefficient attitudes with excessive motion travel and long time consumption. By automatically selecting the optimal shooting order through quantitative scoring, we can eliminate the blindness of random shooting and fixed-order shooting, and prioritize the execution of high-value shooting attitudes. While ensuring the accuracy of defect detection, we can further reduce the mechanical motion time and improve the overall detection operation efficiency.

[0040] S104. The transparent membrane anomaly detection multi-pose microstepping imaging device drives the actuator to perform layered microstepping imaging on the anomaly candidate target area according to the target pose sequence.

[0041] Among them, the aforementioned layered micro-stepping shooting refers to multi-angle deflection sampling based on a reference posture, and dynamically increasing or decreasing the sampling posture according to the imaging benefits.

[0042] Optionally, the step of driving the actuator to perform layered micro-step imaging of the abnormal candidate target region according to the target posture sequence includes: controlling the actuator to adjust to the reference imaging posture corresponding to the membrane surface normal, acquiring an initial reference image with the reference imaging posture, and extracting the corresponding first defect imaging gain; taking the reference imaging posture as the center, sequentially switching each effective imaging posture in the target posture sequence according to a preset step size for step-by-step imaging acquisition; calculating the second defect imaging gain of the image under each effective imaging posture in real time; if the second defect imaging gain is higher than the first defect imaging gain, then adding a candidate imaging posture around the corresponding effective imaging posture.

[0043] Specifically, the multi-pose micro-stepping imaging device for transparent membrane anomaly detection can control the six-axis actuator to return to the standard normal reference pose perpendicular to the transparent membrane surface, ensuring that the initial shooting angle is perpendicular to the defect area, completing the initial image acquisition, and quantifying and calculating the defect clarity, layer identification, and point identification under the reference viewpoint, integrating them to obtain the first defect imaging benefit for fixed reference; then, using this reference shooting pose as the sampling center, the device sequentially and orderly switches the selected shooting poses within the sequence according to a uniform micro-step size, completing frame-by-frame step-by-step image acquisition; simultaneously, feature analysis is performed on each real-shot image under the deflection pose, and the corresponding second defect imaging benefit is solved in real time; when the imaging benefit of a certain deflection pose is better than that of the reference shooting pose, the viewpoint is determined to be a high-value effective viewpoint, and the multi-pose micro-stepping imaging device for transparent membrane anomaly detection can supplement the micro-deflection shooting poses with smaller step sizes around the pose for encrypted and refined sampling.

[0044] It should be noted that using the normal reference shooting posture of the film surface as a unified benchmark for benefit evaluation can achieve a fair comparison of the benefits of all deflected shooting angles and avoid confusion in benefit evaluation standards. Actively encrypted sampling for high-benefit angles can obtain more refined defect imaging data, further restore the true shape of defects, accurately distinguish the differences in internal layers of defects, effectively make up for the problem of insufficient sampling information from a single angle, and comprehensively improve the detection accuracy of small defects in transparent films.

[0045] Optionally, if the imaging gain of the second defect in the image is lower than the imaging gain of the first defect in the image under N consecutive valid shooting postures, then after all valid shooting postures have been traversed, the image is switched to the next abnormal candidate target area for shooting; where N is a positive integer greater than 1.

[0046] Specifically, the transparent film anomaly detection multi-pose microstepping imaging device can compare the imaging gains of the second defect with those of the first defect in real time for multiple consecutive sets of deflection poses, and count the number of times low-gain poses appear consecutively. When N consecutive sets of poses fail to achieve an imaging effect better than the reference viewpoint, it proves that further deflection of the viewpoint around the current defect cannot improve the imaging quality, and there is no need to continue redundant refinement sampling. After all the sequence poses corresponding to the current target have been traversed, the imaging process of the defect area can be directly terminated, and the device can quickly jump to the next candidate target area to be detected for imaging.

[0047] It should be noted that the continuous low-yield posture judgment mechanism can promptly cut off invalid fine sampling processes, avoiding continuous meaningless small-amplitude posture switching and image acquisition work of the actuator; without reducing the overall defect detection accuracy, it reduces the number of mechanical reciprocating movements and image processing time, thus balancing detection accuracy and overall machine detection cycle time.

[0048] In this embodiment, by first acquiring a full-area visual image of the transparent film, identifying abnormal candidate target areas, and extracting corresponding defect feature parameters, the location of film surface defects can be accurately located and the inherent properties of the defects themselves can be obtained. Therefore, it can provide accurate data support for subsequent pose space construction, avoiding blindly conducting full-area imaging. Since the center point of the abnormal candidate target area is used as a reference, and a local pose search space is constructed specifically based on the defect feature parameters, the candidate shooting poses are only arranged around the defect area. There is no need to perform indiscriminate full-area pose scanning on the entire transparent film. Therefore, a large number of invalid shooting actions in the full-area uniform scanning mode are completely avoided, effectively reducing invalid movement of the actuator and reducing detection time costs and mechanical movement losses. Since it adopts a method that includes defect imaging benefits and layer determination benefits, it can achieve the following results. The system employs a preset attitude value function that prioritizes candidate shooting attitudes based on at least one of the following indicators: point target localization benefit and mechanical motion cost. This function prioritizes and selects the optimal shooting sequence, enabling targeted selection of observation angles with better imaging effects and lower motion costs. This overcomes the shortcomings of traditional fixed-angle combined imaging, which suffers from fixed and untargeted perspectives, and improves the clarity of micro-defect imaging, defect localization accuracy, and film layer differentiation capabilities. By performing layered micro-step shooting according to the target attitude sequence and completing multi-angle deflection sampling based on the baseline attitude, the system can dynamically increase or decrease sampling attitudes in conjunction with real-time imaging benefits. Therefore, it can supplement sampling for high-value imaging angles to ensure detection accuracy and reduce the number of samples for attitudes with no imaging benefits, achieving a dynamic balance between detection accuracy and detection efficiency.

[0049] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0051] It should be noted that the device in the embodiments of this application includes a virtual device and a physical device. The virtual device can be a multi-pose micro-stepping imaging device for detecting abnormalities in transparent films, and the physical device can include electronic devices, computer storage media, and computer program products.

[0052] The multi-pose microstepping imaging method for transparent film anomaly detection provided in this application can be executed by a multi-pose microstepping imaging device for transparent film anomaly detection, or by a control module within that device for multi-pose microstepping imaging of transparent film anomaly detection. This application uses the execution of the multi-pose microstepping imaging method for transparent film anomaly detection as an example to illustrate the multi-pose microstepping imaging device for transparent film anomaly detection provided in this application.

[0053] It should be noted that, according to the above method examples, the multi-pose micro-stepping imaging device for transparent film anomaly detection can be divided into functional modules. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. Optionally, the module division in this embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0054] like Figure 2As shown in the figure, this application embodiment provides a multi-pose micro-stepping imaging device 200 for transparent film anomaly detection. The multi-pose micro-stepping imaging device 200 for transparent film anomaly detection includes: an acquisition module 201 and a processing module 202. The acquisition module 201 can be used to acquire a full-field visual image of the transparent film, identify anomaly candidate target regions with defects in the full-field visual image, and extract defect feature parameters of the anomaly candidate target regions. The processing module 202 can be used to construct a local pose search space based on the center point of the anomaly candidate target region and the defect feature parameters, the local pose search space containing multiple sets of shooting poses; prioritize the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; drive the actuator to perform layered micro-stepping imaging on the anomaly candidate target region according to the target pose sequence; wherein, the layered micro-stepping imaging refers to multi-angle deflection sampling based on a reference pose, and dynamically increasing or decreasing the sampling pose according to the imaging gain; the preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target positioning gain, and mechanical motion cost.

[0055] Optionally, the defect feature parameters include scale parameters, target orientation, and anomaly category; the extraction of defect feature parameters of the anomaly candidate target region includes: locating the center coordinates of each anomaly candidate target region, determining the target bounding box size, and obtaining the scale parameters; extracting the extension direction of the strip-shaped defect to obtain the target orientation; and distinguishing the anomaly category based on the defect optical imaging characteristics, wherein the anomaly category includes dot-like particles, scratches, and internal impurities.

[0056] Optionally, the step of constructing a local pose search space based on the center point of the candidate abnormal target region and the defect feature parameters includes: using the center point of the candidate abnormal target region as the imaging reference origin, and setting a safe observation distance range between the camera and the membrane surface in conjunction with the scale parameters; defining the micro-angle deflection boundary of the camera relative to the normal of the membrane surface in conjunction with the joint motion limit of the actuator and the camera stable imaging constraint; setting a differentiated angle sampling density according to the abnormality category and the target orientation to generate candidate observation angles suitable for imaging and recognition of various defects; generating multiple sets of candidate shooting postures around the defect region within the preset local constraint space according to a preset step size, and summarizing all valid shooting postures after legality verification to obtain the local pose search space.

[0057] Optionally, the step of prioritizing the shooting postures in the local posture search space based on a preset posture value function to obtain a target posture sequence includes: substituting each group of shooting postures into the preset posture value function to calculate the corresponding comprehensive evaluation value; arranging all shooting postures in descending order according to the comprehensive evaluation value from high to low; and removing shooting postures with a comprehensive evaluation value lower than a preset threshold to generate the target posture sequence.

[0058] Optionally, the step of driving the actuator to perform layered micro-step imaging of the abnormal candidate target region according to the target posture sequence includes: controlling the actuator to adjust to the reference imaging posture corresponding to the membrane surface normal, acquiring an initial reference image with the reference imaging posture, and extracting the corresponding first defect imaging gain; taking the reference imaging posture as the center, sequentially switching each effective imaging posture in the target posture sequence according to a preset step size for step-by-step imaging acquisition; calculating the second defect imaging gain of the image under each effective imaging posture in real time; if the second defect imaging gain is higher than the first defect imaging gain, then adding a candidate imaging posture around the corresponding effective imaging posture.

[0059] Optionally, the step of driving the actuator to perform layered microstepping imaging of the abnormal candidate target region according to the target posture sequence further includes: if the second defect imaging gain of the image under N consecutive effective shooting postures is lower than the first defect imaging gain, then after all effective shooting postures have been traversed, switching to the next abnormal candidate target region for shooting; where N is a positive integer greater than 1.

[0060] In this embodiment, by first acquiring a full-area visual image of the transparent film, identifying abnormal candidate target areas, and extracting corresponding defect feature parameters, the location of film surface defects can be accurately located and the inherent properties of the defects themselves can be obtained. Therefore, it can provide accurate data support for subsequent pose space construction, avoiding blindly conducting full-area imaging. Since the center point of the abnormal candidate target area is used as a reference, and a local pose search space is constructed specifically based on the defect feature parameters, the candidate shooting poses are only arranged around the defect area. There is no need to perform indiscriminate full-area pose scanning on the entire transparent film. Therefore, a large number of invalid shooting actions in the full-area uniform scanning mode are completely avoided, effectively reducing invalid movement of the actuator and reducing detection time costs and mechanical movement losses. Since it adopts a method that includes defect imaging benefits and layer determination benefits, it can achieve the following results. The system employs a preset attitude value function that prioritizes candidate shooting attitudes based on at least one of the following indicators: point target localization benefit and mechanical motion cost. This function prioritizes and selects the optimal shooting sequence, enabling targeted selection of observation angles with better imaging effects and lower motion costs. This overcomes the shortcomings of traditional fixed-angle combined imaging, which suffers from fixed and untargeted perspectives, and improves the clarity of micro-defect imaging, defect localization accuracy, and film layer differentiation capabilities. By performing layered micro-step shooting according to the target attitude sequence and completing multi-angle deflection sampling based on the baseline attitude, the system can dynamically increase or decrease sampling attitudes in conjunction with real-time imaging benefits. Therefore, it can supplement sampling for high-value imaging angles to ensure detection accuracy and reduce the number of samples for attitudes with no imaging benefits, achieving a dynamic balance between detection accuracy and detection efficiency.

[0061] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communications bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other through the communications bus 340. The processor 310 can call logic instructions in the memory 330 to execute a multi-pose microstepping imaging method for anomaly detection of a transparent film. This method includes: acquiring a global visual image of the transparent film, identifying candidate anomaly regions with defects in the global visual image, and extracting defect feature parameters of the candidate anomaly regions; constructing a local pose search space based on the center point of the candidate anomaly regions and the defect feature parameters, the local pose search space containing multiple shooting poses; prioritizing the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; and driving an actuator to perform layered microstepping imaging on the candidate anomaly regions according to the target pose sequence. The layered microstepping imaging refers to multi-angle deflection sampling based on a reference pose, and dynamically increasing or decreasing the sampling pose according to the imaging gain. The preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target localization gain, and mechanical motion cost.

[0062] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0063] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-pose microstepping imaging method for transparent film anomaly detection provided by the above methods. The method includes: acquiring a full-field visual image of the transparent film, identifying anomaly candidate target regions with defects in the full-field visual image, and extracting defect feature parameters of the anomaly candidate target regions; constructing a local pose search space based on the center point of the anomaly candidate target regions and the defect feature parameters, the local pose search space containing multiple sets of shooting poses; prioritizing the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; driving an actuator to perform layered microstepping imaging on the anomaly candidate target regions according to the target pose sequence; wherein, the layered microstepping imaging refers to performing multi-angle deflection sampling based on a reference pose and dynamically increasing or decreasing the sampling pose according to the imaging gain; the preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target positioning gain, and mechanical motion cost.

[0064] On another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a multi-pose microstepping imaging method for transparent film anomaly detection provided by the methods described above. The method includes: acquiring a global visual image of the transparent film, identifying anomaly candidate target regions with defects in the global visual image, and extracting defect feature parameters of the anomaly candidate target regions; constructing a local pose search space based on the center point of the anomaly candidate target regions and the defect feature parameters, the local pose search space containing multiple sets of shooting poses; prioritizing the shooting poses in the local pose search space based on a preset pose value function to obtain a target pose sequence; driving an actuator to perform layered microstepping imaging on the anomaly candidate target regions according to the target pose sequence; wherein, the layered microstepping imaging refers to performing multi-angle deflection sampling based on a reference pose and dynamically increasing or decreasing the sampling pose according to the imaging gain; the preset pose value function includes at least one of the following: defect imaging gain, layer determination gain, point target positioning gain, and mechanical motion cost.

[0065] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0066] Obviously, those skilled in the art should understand that the various units or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0067] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for detecting anomalies in transparent films using multi-pose micro-stepping imaging, characterized in that, include: Acquire a full-field visual image of the transparent film, identify abnormal candidate target regions with defects in the full-field visual image, and extract the defect feature parameters of the abnormal candidate target regions; Using the center point of the abnormal candidate target region as a reference, a local pose search space is constructed based on the defect feature parameters. The local pose search space contains multiple sets of shooting poses. Based on a preset attitude value function, the shooting attitudes in the local attitude search space are prioritized to obtain the target attitude sequence; The actuator is driven to perform layered micro-step imaging of the abnormal candidate target region according to the target posture sequence; The layered micro-stepping shooting refers to multi-angle deflection sampling based on a reference posture, and dynamically increasing or decreasing the sampling posture according to the imaging benefits. The preset attitude value function includes at least one of the following: defect imaging benefit, layer determination benefit, point target positioning benefit, and mechanical motion cost.

2. The multi-pose micro-stepping imaging method for transparent membrane anomaly detection according to claim 1, characterized in that, The defect feature parameters include scale parameters, target trajectory, and anomaly category; The extraction of defect feature parameters from the abnormal candidate target region includes: Locate the center coordinates of each anomaly candidate target region, determine the target bounding box size, and obtain the scale parameter; Extract the extension direction of the strip-shaped defect to obtain the target orientation; Anomalies are categorized based on their optical imaging characteristics, including dot-like particles, scratches, and internal impurities.

3. The multi-pose micro-stepping imaging method for transparent membrane anomaly detection according to claim 2, characterized in that, The step of constructing a local pose search space based on the defect feature parameters, using the center point of the abnormal candidate target region as a reference, includes: Using the center point of the abnormal candidate target region as the imaging reference origin, and combining the scale parameters, the safe observation distance range between the camera and the membrane surface is set. By combining the joint motion limits of the actuator and the camera's stable imaging constraints, the micro-angle deflection boundary of the camera relative to the normal of the film surface is defined; Differentiated angle sampling density is set according to the anomaly category and the target orientation to generate candidate observation angles that are suitable for various types of defect imaging and recognition; Within a preset local constraint space, multiple sets of candidate shooting postures around the defect area are generated according to a preset step size. After legality verification, all valid shooting postures are summarized to obtain the local posture search space.

4. The multi-pose micro-stepping imaging method for transparent membrane anomaly detection according to claim 1, characterized in that, The step of prioritizing the shooting postures within the local posture search space based on a preset posture value function to obtain a target posture sequence includes: Substitute each shooting posture into the preset posture value function to calculate the corresponding comprehensive evaluation value. All shooting poses are sorted in descending order of their overall evaluation scores, from highest to lowest. Shooting postures with a comprehensive evaluation value lower than a preset threshold are removed to generate the target posture sequence.

5. The multi-pose micro-stepping imaging method for transparent membrane anomaly detection according to claim 1, characterized in that, The step of driving the actuator to perform layered micro-stepping imaging of the abnormal candidate target region according to the target posture sequence includes: The control actuator is adjusted to the reference shooting posture corresponding to the membrane surface normal, the initial reference image is acquired in the reference shooting posture, and the corresponding first defect imaging gain is extracted; Using the baseline shooting posture as the center, step-by-step imaging acquisition is performed by sequentially switching each effective shooting posture in the target posture sequence according to a preset step size. Real-time calculation of the second defect imaging gain of images under each effective shooting posture; If the imaging benefit of the second defect is higher than that of the first defect, then a new candidate imaging posture is added around the corresponding effective imaging posture.

6. The multi-pose micro-stepping imaging method for transparent membrane anomaly detection according to claim 5, characterized in that, The method further includes: If the imaging benefit of the second defect in images under N consecutive valid shooting postures is lower than the imaging benefit of the first defect, then after all valid shooting postures have been traversed, switch to the next abnormal candidate target area for shooting. Where N is a positive integer greater than 1.

7. A multi-pose micro-stepping imaging device for detecting anomalies in a transparent membrane, characterized in that, include: Acquisition module and processing module; The acquisition module is used to acquire a full-field visual image of the transparent film, identify abnormal candidate target regions with defects in the full-field visual image, and extract the defect feature parameters of the abnormal candidate target regions. The processing module is used to construct a local pose search space based on the center point of the abnormal candidate target region and the defect feature parameters, and the local pose search space includes multiple shooting poses. Based on a preset attitude value function, the shooting attitudes in the local attitude search space are prioritized to obtain a target attitude sequence; the actuator is driven to perform layered micro-step shooting of the abnormal candidate target area according to the target attitude sequence; The layered micro-stepping shooting refers to multi-angle deflection sampling based on a reference posture, and dynamically increasing or decreasing the sampling posture according to the imaging benefits. The preset attitude value function includes at least one of the following: defect imaging benefit, layer determination benefit, point target positioning benefit, and mechanical motion cost.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the multi-pose micro-stepping imaging method for transparent film anomaly detection as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-pose microstepping imaging method for transparent film anomaly detection as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-pose microstepping imaging method for transparent film anomaly detection as described in any one of claims 1 to 6.

Images