Ultra-thin electrolytic copper foil defect intelligent detection method and system for new energy field

Abstract

The invention discloses a new energy field-oriented ultrathin electrolytic copper foil defect intelligent detection method and system, and belongs to the technical field of intelligent detection, and the method comprises the following steps: continuously collecting an original image sequence at a fixed line frequency, carrying out the visual detection of a preprocessed image, and carrying out the judgment to obtain a suspected defect region set; performing space-time alignment on a suspected defect area set obtained by visual detection, generating an electrical verification task queue according to the operation process of the coiled material, and executing scheduling; and generating an electrical measurement strategy for an electrical verification task to be triggered, triggering local electrical measurement according to the generated electrical measurement strategy, obtaining local measurement data of the suspected area, and performing suspected result judgment on the suspected defect area. According to the method, the problem of low defect type detection accuracy caused by a single detection means in a high-speed production process in the prior art is solved by constructing a collaborative detection process taking visual detection as an entrance and taking electrical detection as a targeted verification means.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, and in particular to an intelligent detection method and system for defects in ultrathin electrolytic copper foil for the new energy field. Background Technology

[0002] Ultrathin electrolytic copper foil is an extremely thin copper material produced by electrolytic deposition. It is in roll form and its thickness is typically in the micrometer range, commonly around 4–12 μm. It features high purity, high conductivity, high ductility, and a smooth surface. In the new energy field, copper foil serves as the carrier substrate for negative electrode active materials, effectively transferring electrons to the external circuit. The thinner the electrolytic copper foil, the more active material can be loaded in the same space, thereby increasing the battery's energy density. However, when ultrathin electrolytic copper foil has pinholes, inclusions, or local thickness anomalies, it can cause local current density anomalies and heat concentration. When combined with the battery system, this can potentially evolve into micro-short circuit risks or performance degradation. Therefore, intelligent defect detection of ultrathin electrolytic copper foil is crucial. The goal is to discover and quantify surface, edge, morphology, and thickness-related defects on high-speed production lines for continuous copper foil production and slitting and rewinding. This allows for the transformation of defects into risk classification and online handling capabilities for battery consequences. It enables real-time detection of various defects on large-format, high-speed copper foil production lines and precise classification, location, and labeling of various micrometer-level defects, reducing the risk of safety accidents caused by material defects.

[0003] For example, Chinese invention patent CN113325001B discloses an automatic detection device and method for identifying surface defects on copper foil. This device is installed between the copper foil processing end and the slitting end. It includes a CCD line scan camera installed above and below the copper foil production line to perform real-time synchronous scanning of the front and back sides of the copper foil; a visual inspection image processing system to analyze and judge the image features of the copper foil surface; a light source generating device; and a grading module that classifies and grades copper foil defects based on their size, the number of defects per specified unit length, and the density of defects. The visual inspection image processing system identifies whether pinholes, spots, paint marks, or indentations exist on the top and bottom surfaces of the copper foil.

[0004] For example, Chinese Invention Patent CN112415013B discloses a copper foil defect detection system, which includes: an acquisition module for acquiring an initial image corresponding to the copper foil to be tested; a processing module for preprocessing the initial image to obtain a standard initial image; a matching module for performing contour matching on the standard initial image to obtain a detection image of the copper foil to be tested; and a comparison module for generating a defect label associated with the copper foil to be tested based on the difference coefficient between the detection image and a standard copper foil image being greater than a threshold value. The system can extract the detection image of the copper foil to be tested and compare the detection image with the standard copper foil image. During the comparison process, the pixel unit of the detection image is used as the smallest comparison unit.

[0005] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems:

[0006] In existing technologies, for ultra-thin electrolytic copper foil used in new energy applications, optical imaging is commonly used for online inspection of the copper foil surface. These methods typically rely on grayscale, texture, color, and local statistical perturbations as primary criteria, and can effectively detect surface defects such as pinholes, scratches, and wrinkles. However, in the high-speed continuous production of ultra-thin electrolytic copper foil, defect risks not only arise from surface abnormalities but also include functional defects such as internal thinning, localized electrical conductivity anomalies, uneven metallurgical structure, and material continuity disruption. These defects often exhibit weak contrast and instability at the imaging level. The presence of insignificant visual features makes it difficult for visual inspection to provide evidence directly corresponding to electrical performance risks, leading to missed detections of minor but high-risk functional defects. At the same time, minor contamination, oil mist, water marks, light fluctuations, micro-vibrations caused by tension, and changes in viewing angle under high-speed operating conditions can easily generate significant responses in visual images that are similar to real defects, increasing the false judgment rate of visual inspection. Existing technologies lack a linkage mechanism that can drive verification based on the results of visual inspection, resulting in a technical problem of low accuracy in defect type detection due to the reliance on a single inspection method during high-speed production. Summary of the Invention

[0007] To address the technical problem of low accuracy in defect detection due to the reliance on a single detection method during high-speed production, this invention provides an intelligent defect detection method and system for ultra-thin electrolytic copper foil in the new energy field. The technical solution is as follows:

[0008] On the one hand, a method for intelligent defect detection of ultrathin electrolytic copper foil in the new energy field is provided. The method includes: when the ultrathin electrolytic copper foil passes through the field of view of an industrial camera, continuously acquiring original image sequences at a fixed line frequency; preprocessing the original images; performing visual inspection on the preprocessed images to determine a set of suspected defect areas; performing spatiotemporal alignment on the set of suspected defect areas obtained by visual inspection; generating an electrical verification task queue according to the roll material operation process; and simultaneously scheduling the task queue; generating an electrical measurement strategy for the upcoming electrical verification task, the electrical measurement strategy including the selected electrical measurement mode, the set electrical scanning path, and the number of points; triggering local electrical measurement through an electrical detection probe according to the generated electrical measurement strategy; acquiring local measurement data of the suspected areas; and determining the suspected defect areas that have completed the electrical measurement.

[0009] On the other hand, an intelligent defect detection system for ultra-thin electrolytic copper foil in the new energy field is provided. This system includes: a module for identifying suspected defect areas, a module for generating an electrical verification task queue, a module for generating an electrical measurement strategy, and a module for determining suspected results. The module for identifying suspected defect areas continuously acquires a sequence of original images at a fixed line frequency when the ultra-thin electrolytic copper foil passes through the field of view of an industrial camera, preprocesses the original images, and performs visual inspection on the preprocessed images to determine a set of suspected defect areas. The module for generating an electrical verification task queue performs spatiotemporal alignment on the set of suspected defect areas obtained from visual inspection, generates an electrical verification task queue based on the roll material operation process, and schedules the task queue. The module for generating an electrical measurement strategy generates an electrical measurement strategy for the upcoming electrical verification task, which includes a selected electrical measurement mode, a set electrical scanning path, and a number of points. The module for determining suspected results triggers local electrical measurements using an electrical detection probe according to the generated electrical measurement strategy, acquires local measurement data of the suspected areas, and determines the suspected result for the completed electrical measurements of the suspected defect areas.

[0010] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0011] 1. The intelligent defect detection method for ultra-thin electrolytic copper foil in the new energy field provided by this invention constructs a cascaded collaborative detection process with visual inspection as the entry point and electrical inspection as the targeted verification method. Under the condition of high-speed roll material operation, a unified spatiotemporal alignment, task-driven and joint judgment mechanism is introduced, which enables the suspected defect area obtained by visual inspection to be accurately mapped and corresponding verification evidence to be obtained by electrical inspection. Thus, without increasing the burden of full-width electrical scanning, it can stably distinguish between appearance defects and functional defects of ultra-thin electrolytic copper foil. Compared with the existing detection schemes that rely only on single optical inspection or visual and electrical parallel but independent detection schemes, it can significantly reduce the missed detection rate of functional defects and the misjudgment rate of appearance anomalies, and maintain the consistency and interpretability of detection results in high-speed production scenarios, thereby improving the engineering practical value of the entire detection system in the quality control of ultra-thin copper foil in the new energy field.

[0012] 2. This invention establishes a spatiotemporal mapping relationship between image coordinates, physical coordinates of the roll material, and the coordinate system of the electrical probe based on the roll material's running trajectory. This transforms the suspected defect area obtained during the visual inspection stage from the image domain into an electrical verification task with a clear trigger location and trigger time window. The electrical inspection behavior is scheduled and controlled in the form of a task queue, thus solving the problem of the inability to correspond one-to-one between visual anomalies and electrical signals in the prior art. This mechanism makes electrical inspection no longer dependent on a fixed sampling rhythm or blind scanning strategy, but triggers local measurement only within the effective time window when the suspected defect area reaches the probe. This effectively avoids verification failures caused by multimodal data misalignment and scale mismatch, enabling the visual inspection results to be accurately located and verified in the electrical signals, significantly improving the executability and reliability of multimodal collaborative detection.

[0013] 3. For each upcoming electrical verification task, instead of using a fixed electrical measurement method, the electrical verification evidence target is determined based on the initial judgment of the defect type of the suspected defect area, visual confidence level, and multi-feature description vector. On this basis, the electrical measurement mode, scanning path, number of sampling points, and measurement parameters are adaptively selected, so that the electrical measurement process can revolve around verifying whether the suspected defect has structural anomalies. Compared with the existing technology of statically configuring electrical detection parameters or setting them only according to equipment capabilities, this invention can prioritize the use of electrical probe resources to obtain the measurement evidence with the highest discriminative value in high-risk areas within a limited triggering time window, thereby improving the pertinence and discrimination efficiency of electrical verification for different defect morphologies without increasing hardware complexity.

[0014] 4. After completing local electrical measurements, visual feature sub-vectors and electrical feature sub-vectors are constructed, and fusion judgment is performed while maintaining the distinguishability of feature sources. A defect verification mechanism with the combination of visual confidence and electrical confidence as the core is established. This mechanism can provide clear judgment results that distinguish between structural defects, non-structural anomalies, and potential hidden defects in conflict scenarios that are difficult to handle in existing technologies, such as "visual anomalies are obvious but electrical functions are normal" and "visual anomalies are not significant but electrical functions are abnormal". Compared with existing detection systems that only output a single anomaly conclusion, this invention can identify functional risk defects that are difficult to identify visually in a stable manner in advance, thereby improving the guiding value of the detection results in actual production decision-making. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field provided in this application embodiment.

[0017] Figure 2 A flowchart for the combined electrical fine inspection decision provided in the embodiments of this application.

[0018] Figure 3 A flowchart for determining the defect type provided in the embodiments of this application.

[0019] Figure 4 This is a schematic diagram of the structure of an intelligent defect detection system for ultra-thin electrolytic copper foil in the new energy field, provided in an embodiment of this application. Detailed Implementation

[0020] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0021] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. In embodiments of the present invention, sometimes subscripts such as W1 may be written in a non-subscript form such as W1; unless the distinction is emphasized, their intended meanings are the same.

[0022] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0023] Embodiment 1 of the present invention provides an intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field, such as... Figure 1 The diagram shown is a flowchart of an intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field, provided by an embodiment of this application. The method includes the following steps:

[0024] S1. When the copper foil passes through the camera's field of view, the original image sequence is continuously acquired at a fixed line frequency. Due to the high reflectivity and fine texture of the ultra-thin electrolytic copper foil surface, and under the influence of tension fluctuations, slight shaking, light source aging, or lens contamination, the original image is prone to uneven brightness, fixed pattern interference, and geometric distortion. This results in poor stability and a high false alarm rate when directly identifying defects based on the original image. Therefore, this embodiment first preprocesses the original image to restore the surface texture information that can be stably analyzed. The preprocessing includes: firstly, performing illumination and background compensation on the original image based on a set reference image to suppress fixed pattern interference and improve the comparability between images acquired at different time periods.The reference image can be a bright-field reference image and / or a dark-field reference image pre-acquired when the production line is running normally and the copper foil is in a defect-free or acceptable state. The reference image is used to characterize the stable background and illumination distribution characteristics of the imaging system. By normalizing and compensating the original image and the reference image, which refers to scaling and correcting the gray values ​​of each pixel or local area in the original image using the reference image as the brightness and background benchmark, the difference between the illumination distribution and the stable background is unified to the same scale. This effectively eliminates repetitive interference textures caused by uneven light sources, lens contamination, or sensor fixed noise, allowing subsequent inspection to focus more on the real texture details that change with the surface of the roll material. Subsequently, the compensated image undergoes limited contrast adaptive enhancement processing. The image is divided into multiple local regions according to a preset window size. Local statistical features such as the mean brightness, variance brightness, and gradient energy are calculated in each local region. Based on the magnitude and direction of the deviation between the local contrast and the target contrast threshold, a corresponding enhancement adjustment factor is generated through a monotonic and limited mapping relationship. Regions with local contrast below the set contrast threshold receive a higher enhancement amplitude according to the enhancement adjustment factor mapping, while regions with contrast above the contrast threshold receive a higher enhancement amplitude according to the enhancement adjustment factor mapping. Sub-mapping reduces the enhancement amplitude, making subtle textures, edges of minor scratches, or brightness abrupt changes around pinholes on the copper foil surface more easily highlighted. Simultaneously, by calculating the high-frequency energy percentage or local gradient variance within a local area as a noise indicator, when the noise indicator exceeds a preset threshold, the enhancement gain of that local area is reduced or frozen to restore it to a set baseline level, preventing excessive noise amplification. This improves detail visibility while maintaining stable texture statistics. Furthermore, spatial distortion correction is performed on the image, and pixel coordinates are uniformly mapped to the physical coordinate system of the roll material. Spatial distortion correction refers to… The geometric mapping relationship obtained by camera calibration is used to correct nonlinear deformation caused by lens distortion, mounting tilt angle, or field projection, so that the image position corresponding to the same physical position remains consistent in different frames. The physical coordinate system of the roll material is a coordinate system consisting of the longitudinal coordinate defined along the roll material running direction and the transverse coordinate defined along the transverse width direction of the roll material. Through this mapping, the pixel position in the image can be converted into the actual position on the roll material, thereby providing a consistent spatial reference for cross-frame correlation, defect position tracking, and subsequent alignment with the position of the electrical probe, avoiding the difficulty in stable identification due to the same defect drifting in multiple frames caused by jitter or distortion.

[0025] After preprocessing, a cross-scale local texture perturbation response set is constructed based on the preprocessed image to characterize anomalies on the copper foil surface. "Cross-scale" refers to extracting responses sensitive to anomalies at different spatial scales, capable of sensing both fine-scale texture abrupt changes caused by small defects such as pinholes and microcracks, and medium- or large-scale texture changes caused by larger structural perturbations such as wrinkles and indentations. The local texture perturbation response set represents multiple complementary anomalous response quantities formed within the local neighborhood of the image, used to characterize three types of anomalies: local texture abrupt changes, directional continuity imbalance, and local statistical structure deviation. The anomalous response quantities reflect whether there is a brightness abrupt change, whether there is an abnormal directional structure, and whether its texture statistical characteristics deviate from the normal background. Local texture abrupt changes reflect abnormal changes in local brightness or texture over short distances, suitable for characterizing anomalies with abrupt boundaries such as pinhole boundaries and scratch edges. Multi-scale high-pass component extraction and gradient energy analysis characterize local texture abrupt changes, local unevenness changes, and abrupt features at pinhole boundaries and scratch contours. Directional continuity imbalance reflects the disruption or abnormal enhancement of texture continuity in a certain main direction. Furthermore, directional filter banks are applied to the image to identify texture fragments with unbalanced directional consistency and enhance linear structures that change continuously along a single direction. This is used to capture perturbation patterns with directional features, such as scratches and wrinkles, and is suitable for characterizing linear structures with obvious directionality, such as scratches and wrinkles. Local statistical structure deviation is used to reflect the abrupt change in the statistical distribution characteristics of local texture relative to the background. It is suitable for characterizing anomalies such as color difference, contamination, or roughness anomalies that do not necessarily have clear edges but change the statistical distribution of texture. Using the local neighborhood as the analysis unit, the pixel value distribution, fluctuation dispersion, and statistical stability characteristics of the central local region and its adjacent reference regions are calculated separately. The deviation between the two in the corresponding statistical dimensions is calculated item by item, and each deviation is normalized to make them on a uniform comparable scale. Then, the pixel value distribution, fluctuation dispersion, and statistical stability characteristics are summed to obtain the texture consistency index. The texture consistency index is used to quantify the degree of difference between the statistical distribution of texture in a local region and the statistical distribution of its corresponding adjacent regions, and to measure the statistical structure deviation of the local region, thereby identifying abrupt changes in the statistical distribution of color difference, contamination, and roughness anomaly regions.

[0026] To ensure comparability of values ​​from different scales and response terms, a unified scale alignment and difference enhancement process needs to be performed on the cross-scale local texture perturbation response set. Scale alignment eliminates differences in the value range and units of different response terms, preventing unreasonable amplification or submersion of each response term during fusion due to differences in amplitude. Difference enhancement strengthens the distinction between anomalous responses and local background responses, making anomalous regions more prominent in the fusion result. Simultaneously, it suppresses interference from background regions on anomalous responses, avoiding erroneous saliency increases caused by large-area uniform textures or residual illumination. Subsequently, the processed response results are fused to generate a saliency distribution map, which characterizes the local regions of the image relative to their surrounding regions. The significance of anomalies in the neighborhood is determined by constructing a neighborhood background reference using local regions as the analysis unit. The anomaly response values ​​at the current location are compared with the neighborhood reference to obtain the enhanced response results. The response results of multiple types at the same scale and across scales are summed and averaged to obtain a significance value that characterizes the significance of the location relative to its neighborhood background. The higher the significance value, the more likely the location is to correspond to a suspected defect area. This unifies multiple anomaly clues into an intermediate result that can be used for region extraction. Through the above construction and fusion process, a significant characterization that is sensitive to defects and robust to the background can be stably generated under complex lighting and high reflectivity backgrounds, improving the reliability of subsequent suspected region extraction.

[0027] After obtaining the saliency distribution map, a set of suspected defect regions is further determined. Specifically, based on the saliency criteria used to distinguish between abnormal responses and background responses, and combined with the spatial correlation between adjacent pixels or local regions, connected regions with clustered saliency responses are extracted from the saliency distribution map. A connected region is a continuous region in the saliency distribution map composed of spatially adjacent pixels or local regions whose saliency responses jointly satisfy the abnormality criteria. The saliency criteria can be understood as a set of judgment rules used to classify saliency responses into abnormal candidates and background. It can be implemented by a fixed threshold, an adaptive threshold, or a threshold strategy based on local statistics, but the specific form is not limited. Spatial correlation is used to merge spatially adjacent pixels or small regions with interconnected saliency responses into the same candidate region to obtain connected regions with spatial continuity, thereby avoiding the division of the same defect into multiple fragmented regions. The system may misidentify discrete noise as defect areas. Subsequently, it calculates regional features for each connected region to characterize its morphology, texture perturbation intensity, and response consistency within a preset frame window, forming a region-level multi-feature description vector. The region morphology describes the candidate region's area, aspect ratio, main direction, and other geometric attributes to reflect the spatial structure characteristics of the anomaly. The texture perturbation intensity describes the response amplitude and distribution characteristics of the region on a cross-scale local texture perturbation response set to reflect the salience of the anomaly. The response consistency within the preset frame window describes the stability of the salience response at corresponding positions in several consecutive frames. The preset frame window can be determined based on line frequency and line velocity to cover the effective occurrence duration of the defect within the camera's field of view. This consistency evaluation can suppress false anomalies caused by instantaneous reflective flicker, random noise, or brief jitter, thereby reducing false alarms and improving the stability of the candidate region. Based on regional multi-feature description vectors, visual confidence scores are generated to characterize the credibility of a connected region as a real defect candidate. Simultaneously, connected regions are categorized according to the components representing regional morphology and texture perturbation in the multi-feature description vectors. For each connected region, based on its regional morphological features and texture perturbation features, a score is calculated based on preset criteria to determine the degree of matching between the region and preset anomaly categories such as point anomalies, linear deformations, surface perturbations, and roughness anomalies. The connected region is then classified into the anomaly category with the highest feature matching degree, providing prior information for the selection of measurement modes in the subsequent electrical verification stage. The category identification includes at least one of point anomalies, linear deformations, surface perturbations, or roughness anomalies. Point anomalies are used to characterize local small-scale anomalies such as pinholes and pitting; linear deformations are used to characterize anomalies such as scratches and wrinkles that extend continuously along a certain direction; surface perturbations are used to characterize regional anomalies caused by large-scale contamination, indentations, or uneven coating; and roughness anomalies are used to characterize statistical distribution anomalies such as overall coarsening or thinning of surface microtexture.Ultimately, a set of suspected defect areas is formed, including the location and / or boundaries, visual confidence level, and category identifier. The location and / or boundaries are given based on the aforementioned mapping from pixel coordinates to the physical coordinate system of the roll material. This allows suspected areas to be located not only in the image but also directly associated with the actual location of the roll material. This provides executable input for establishing the spatiotemporal alignment with the electrical probe, generating electrical verification tasks, and realizing event-driven local electrical precision inspection. Through the S1 processing flow of this embodiment, high-value suspected areas can be stably screened across the entire range with vision as the entry point without increasing the burden of full-width electrical scanning. The structured results with spatial coordinates, confidence level, and type information are output, thereby significantly reducing the risk of false positives and false negatives caused by reflection, noise, and operating condition drift. This provides a clear and schedulable target set for subsequent electrical verification, improving the practicality and reliability of the entire cascaded collaborative inspection process under high-speed production lines and resource-constrained conditions.

[0028] S2. Spatiotemporal alignment of the suspected defect area set obtained from visual inspection. Since the ultra-thin electrolytic copper foil is in continuous motion during production, while the electrical probe has a fixed or limited position and processing capacity on the production line, if the specific time and spatial location of the suspected defect area reaching the electrical probe cannot be accurately predicted, effective electrical measurement of the target area cannot be performed within a limited time, easily leading to measurement misalignment, missed detection, or wasted probe resources. By spatiotemporally aligning the suspected defect area, the visual inspection results are transformed into a verification task that can be executed by the electrical probe and has clear triggering conditions. Specifically, for the suspected defect area set obtained from visual inspection, a spatiotemporal mapping relationship between the image coordinate system and the electrical probe coordinate system is established based on the roll's running trajectory. The roll's running trajectory represents the roll's movement path in the production line and its subsequent... The temporal and spatial relationships of these regions, which can be provided by linear speed, roller position, encoder signals, or production line control systems, allow for the conversion of the suspected defect area's position in the image into its actual physical location on the production path. Furthermore, it enables the estimation of when and where this area will reach the effective detection range of the electrical probe in the future, thus yielding the corresponding expected trigger position and trigger time window. The trigger position indicates the detection position of the electrical probe in the horizontal or vertical direction, while the trigger time window defines the time interval within which the suspected area is within the probe's measurable range. This time window reflects the combined results of factors such as the roll material running speed, the probe's effective range, and the area size. Through this mapping, the suspected defect area, which originally existed only in the image domain, is transformed into a spatiotemporal mapping result with clear physical meaning, providing a foundation for subsequent electrical verification.

[0029] An electrical verification task queue is generated based on the roll material operation process. After obtaining the spatiotemporal mapping results, a corresponding electrical verification task is generated for each suspected defect area. Each electrical verification task includes at least the trigger position and trigger time window for the electrical probe, and further includes mode information related to electrical measurement. This mode information indicates the measurement method or configuration type that the electrical probe should use when performing measurements, such as the electrical verification means required to distinguish different defect types. By encapsulating suspected defect areas into electrical verification tasks, the conversion from area results to executable tasks is achieved, enabling visual inspection and electrical measurement to be integrated in data... The structure and execution logic are decoupled. Based on the arrival order of the roll material during operation, multiple electrical verification tasks are organized according to the order of their trigger time windows to form an electrical verification task queue. At the same time, the task queue can also be sorted or rearranged in combination with a preset scheduling strategy. The preset scheduling strategy can be determined according to visual confidence, defect category identification, area scale or other business needs, and is used to distinguish the importance of different tasks when probe resources are limited. By constructing an electrical verification task queue, the verification requirements corresponding to multiple suspected defect areas can be managed uniformly on the timeline, avoiding conflicts or omissions between tasks.

[0030] Simultaneously, task queue scheduling is performed. After the task queue is formed, the actual available processing capacity of the electrical probe within the trigger time window and the allowed mode switching constraints per unit time are further considered. The electrical verification task queue is then scheduled. Available processing capacity represents the number of measurements or measurement duration that the electrical probe can complete within a given time window. Mode switching constraints refer to the time or number of times the electrical probe needs to switch between different measurement modes. When scheduling the electrical verification task queue, the required measurement duration or number of measurements within the trigger time window for each task is estimated and compared with the actual available processing capacity of the electrical probe within the time window to obtain the task's executability index. The executability index characterizes the feasibility of executing the electrical verification task within the trigger time window. Simultaneously, the measurement mode switching requirements between adjacent tasks are evaluated. The corresponding switching correction amount is determined based on the pre-calibrated mode switching time parameters, and the task execution cost is adjusted based on the switching correction amount according to the time required for mode switching. Based on this, the executability index is used to screen... A set of executable tasks within the corresponding trigger time window is selected. Within this set, tasks are sorted according to the modified execution cost, prioritizing tasks with lower execution costs and greater executability margins. This includes scheduled execution, delayed execution, or, when necessary, a degraded measurement mode with low resource consumption. This prioritizes the completion of electrical verification of high-value suspected defect areas under limited probe resources, thus avoiding the resource waste and efficiency bottlenecks of traditional full-width or blind-scan electrical inspection. Ultimately, through this scheduling result, the electrical probe can trigger local measurements in a predetermined order within the corresponding trigger time window when each suspected defect area reaches its detection position. This achieves an event-driven electrical verification method, effectively transforming visually suspected defect areas on continuous rolls into electrical verification tasks with clear spatiotemporal constraints. Under limited probe resources, tasks are uniformly scheduled and managed, allowing the electrical probe to perform measurements on the target area only at the necessary time and location. This significantly improves the relevance, timeliness, and feasibility and reliability of the overall inspection system in high-speed production line environments.

[0031] S3 generates an electrical measurement strategy for the upcoming electrical verification task. The electrical measurement strategy includes the electrical measurement mode, the set electrical scan path, and the number of points. For ultra-thin electrolytic copper foil running continuously at high linear speeds on the production line, the effective measurement time of the electrical probe for a single suspected area is strictly limited by the trigger time window. At the same time, resource conditions such as the number of probe channels, sampling bandwidth, and mode switching time also constrain the executable measurement intensity. Therefore, if only a fixed measurement method or fixed sampling density is used, two types of problems are likely to occur if the electrical measurement strategy is used to effectively solve the problem: first, the measurement cannot be completed within the time window, resulting in missed measurements; second, the measurement is too coarse, making it difficult to form electrical evidence that can be used for verification and confirmation. The electrical measurement strategy is defined as a set of measurement control schemes that can be executed within the trigger time window and can specifically support defect verification. The electrical measurement strategy includes at least the selection result of the electrical measurement mode and the electrical scan path and number of sampling points that match the mode. It can further include a set of parameters such as dwell time or integration time, excitation frequency / band or sampling rate, etc., to maximize the discrimination capability of electrical verification under resource-constrained conditions.

[0032] For an upcoming electrical verification task, based on the preliminary defect type assessment, visual confidence level, and region-level multi-feature description vector obtained in the visual stage for the suspected defect area corresponding to the task, the electrical verification evidence target is determined. The electrical verification evidence target is the key electrical evidence type and its discrimination focus required to complete verification confirmation or graded handling. It is used to clarify the core questions that electrical measurement needs to answer. Based on the electrical verification evidence target, a search is performed in a preset measurement mode selection strategy library to obtain candidate electrical measurement modes. The measurement mode selection strategy library can be understood as a set of rules or strategies that map the preliminary defect type assessment and evidence target to the selectable electrical measurement methods. Its content can be pre-configured according to the probe type, production line conditions, and quality control requirements. The candidate electrical measurement mode can be a single measurement method or a combination of multiple measurement methods, used to obtain complementary electrical evidence within the same time window. It should be noted that in this embodiment, the candidate electrical measurement mode is not the same as the final executable mode, because the final mode must also meet the spatiotemporal constraints and probe resource constraints. In terms of obtaining candidate electrical measurement modes, data-driven selection or hierarchical decision-making methods can also be used.

[0033] Based on the trigger location and trigger time or time interval given by the spatiotemporal mapping results, and according to the available resource constraints of the electrical probe, the candidate measurement modes are screened for feasibility. Available resource constraints characterize the measurement load the probe can withstand within the current trigger time window, such as the number of available channels, available sampling bandwidth, allowed number of mode switches, and the measurement budget allocated to the task within that time window. By combining candidate modes with the trigger time window, trigger location, and resource constraints, candidate measurement modes that cannot be completed within the time window or that would lead to task conflicts and resource overload can be eliminated, ensuring that the final selected mode is feasible. Among the feasible electrical measurement modes, the corresponding electrical measurement mode is selected based on the evaluation results of the discriminative ability of the electrical verification evidence target as the electrical measurement mode for the suspected defect area. The discriminative ability evaluation results characterize the degree of distinguishability of the candidate measurement mode for the evidence target, the expected risk of misjudgment, or verifiable historical performance. The electrical measurement mode can be obtained in any of the following ways or in combination:

[0034] Method 1: Discriminative ability evaluation based on statistical separability distance. For each executable measurement mode, determine the hypothesis set H corresponding to the evidence target (at least including the normal hypothesis H0 and the target defect hypothesis H1, or including multiple defect type hypotheses), and obtain the mean output feature μ of the mode under different hypotheses based on historical labeled samples or offline calibration samples. m,i Simultaneously, under field noise conditions, the system estimates the noise covariance Σm of the mode's output characteristics based on reference area data, and uses the minimum inter-class Mahalanobis distance as the indicator of the mode's discriminative ability.

[0035] ;

[0036] D m As one of the results of the discriminative ability evaluation, and meeting the preset threshold D m ≥D min Given the premise, choose D. m The largest measurement mode is taken as the electrical measurement mode; or D m The resource consumption cost is incorporated into the comprehensive selection rule to output the final mode, min i≠j This indicates that the minimum value is taken for all different pairs of assumptions (Hi, Hj); m is the electrical measurement mode identifier, representing a specific electrical measurement configuration scheme. Represents the covariance matrix The inverse matrix, , These represent the mean vectors of the output features corresponding to the i-th and j-th hypotheses under measurement mode m, respectively, which are obtained by statistical analysis of historical labeled samples or offline calibration samples. Under the same measurement mode m, each hypothesis corresponds to a unique set of mean vectors.

[0037] Method 2: Based on information gain, the discriminative ability evaluation maps visual confidence and the initial defect type judgment result to a prior probability distribution P(H) of the hypothesis set H. For each executable measurement mode m, P(Y) is established based on historical or calibration data. m |H) (where Y) m Calculate the expected reduction in hypothesis uncertainty caused by the measurement model (for the evidence features or evidence vectors output by the measurement model), as an indicator of discriminative capability:

[0038] ;

[0039] Among them, H( H H( is the entropy of the assumption before measurement) H |Ym) represents the posterior entropy after combining the measurement output; IG(m) is used as one of the results of the discrimination capability evaluation, and the measurement mode with higher information gain and satisfying resource constraints is preferentially selected as the electrical measurement mode, E Ym [ ] represents the mathematical expectation of the random variable of the measurement output, which means averaging the reduction in uncertainty over all possible measurement results.

[0040] When the measurement patterns obtained by the two discrimination capability evaluation methods are inconsistent, a decision shall be made based on the current operating scenario: when the on-site noise is high or the defect type discrimination is low, the statistical separability distance evaluation result shall be given priority; when the prior information is sufficient or the probability distribution of defect types is significantly different, the information gain evaluation result shall be given priority. After the electrical measurement mode is determined, a scanning path and sampling point configuration matching the mode are further generated. Specifically, the scanning path form is determined based on the spatial morphology and equivalent scale of the suspected defect area. The spatial morphology can be described by the boundary, main direction, aspect ratio, or area of ​​the suspected area. The equivalent scale is used to transform irregular areas into scale representations that can be used for measurement planning, enabling the path planning to maintain consistent configurable logic across different defect morphologies. For example, point anomalies are more suitable for local point measurement or small-area grid measurement, linear deformations are more suitable for line scanning along the main direction, and area disturbances are more suitable for coverage or multi-line scanning paths. The number of sampling points and the point spacing are configured based on the relationship between the scale of the suspected defect area and the sensitive footprint of the electrical probe. By comparing the equivalent scale of the suspected defect area with the sensitive footprint of the electrical probe, the sampling point spacing is determined according to the principle of footprint covering defects, and the required number of sampling points is calculated accordingly. The sensitive footprint refers to the area covered by the electrical probe in a single sampling. The material generates an effective response and forms the spatial range of the measurement output or the equivalent sampling coverage area. By associating the regional scale with the sensitive footprint, the number and interval of sampling points can be reasonably determined, so that the set scanning path can cover the suspected area within the trigger time window, while avoiding excessive sampling that leads to timeout or resource congestion, thereby meeting the preset measurement coverage criteria. The preset measurement coverage criteria are used to limit the minimum requirements for measurement coverage, such as requiring the suspected area to obtain at least a certain number of effective sampling points in its key directions or to reach a certain coverage ratio. After completing the path and point planning, the dwell time or integration time is further configured according to the electrical verification evidence target and the on-site noise conditions, so that the measurement uncertainty is within the discriminable range. On-site noise conditions include signal fluctuations caused by electromagnetic interference, contact fluctuations, temperature drift, and mechanical vibration of the production line. The dwell time or integration time is used to reduce the influence of random noise by extending sampling or integration, but its setting is limited by the trigger time window.

[0041] Therefore, this embodiment configures the evidence target and noise conditions in a linked manner, enabling the measurement to obtain sufficient signal-to-noise ratio and stability under feasible conditions. At the same time, it configures parameters such as excitation frequency, frequency band, or sampling rate in combination with the physical mechanism of the selected electrical measurement mode. The excitation frequency or frequency band is used to match the sensitive range of different modes to thickness, conductivity, or continuity disruption, and the sampling rate is used to ensure the ability to capture response changes. Through the above parameter configuration, an electrical measurement strategy is finally formed to drive the electrical probe to perform verification within the corresponding trigger time window, transforming electrical measurement from general detection to targeted verification for the evidence target. Thus, even under resource-constrained conditions, it can still output electrical evidence that can be used to judge consistency with visual results.

[0042] S4, the electrical detection probe triggers local electrical measurements according to the generated electrical measurement strategy to acquire local measurement data of the suspected area. The electrical detection probe can be a non-contact electrical measurement device, which applies controlled electromagnetic excitation to the copper foil and collects response signals, outputting equivalent parameters that have a definite correspondence with local resistance changes. The equivalent parameters include impedance-related quantities, equivalent conductivity parameters, response amplitude, or phase characteristics. The designated probe channel refers to the set of channels selected to ensure consistency with the trigger position and scanning path when the probe has multiple channels or multiple measurement units. The scanning path is the sampling sequence and sampling coordinates of the probe on the lateral position or coverage trajectory of the suspected area. The set of reference points; trigger measurement indicates that the probe sampling is initiated within the trigger time window based on the spatiotemporal mapping results and the sampling is executed according to the set parameters, so that the measurement behavior is strictly aligned with the physical arrival process of the suspected area, avoiding measurement misalignment or deviation of the measured object; to make the electrical measurement results comparable, a local reference benchmark is constructed at the nearest neighbor position on the same roll material, and the difference characteristics characterizing the degree of anomaly of the suspected defect area are calculated based on the local measurement data and the local reference benchmark. The nearest neighbor position refers to the reference area that is adjacent to the suspected area on the roll material and close in time. This reference area is usually under the same production batch, similar temperature and tension conditions as the suspected area. This can serve as a background baseline; a local reference baseline can be composed of the average resistance or conductivity, average thickness or morphology of the reference area, and the corresponding fluctuation range; difference features are used to quantify the degree of deviation of the suspected area from the reference baseline, such as deviation amplitude, deviation direction, spatial distribution inhomogeneity, or continuity disruption features along the scanning path; after obtaining the difference features, based on the preliminary judgment and category identification of the defect type corresponding to the suspected area in the suspected defect area set, the difference features are structurally aggregated and physically consistent with each other, generating an electrical confidence level to characterize the credibility of structural anomalies. The structural aggregation arranges the difference features according to the evidence structure related to the defect type. Merging and summarizing are crucial. For example, for point anomalies, more attention is paid to local peak deviation and peak width; for linear deformations, more attention is paid to continuous deviation and fracture characteristics along the main direction; for area disturbances, more attention is paid to regional mean shift and consistency; for roughness anomalies, more attention is paid to the statistical fluctuation pattern of the difference characteristics. Physical consistency verification is used to determine whether the observed difference characteristics conform to the expected physical mechanism of the selected electrical measurement mode. For example, when the measurement mode is sensitive to thickness changes, the difference characteristics should show directional changes consistent with the thinning or thickening of the thickness. If the difference characteristics show a pattern that is significantly inconsistent with the mechanism, the credibility of it as evidence of structural defects is reduced.

[0043] like Figure 2The diagram shows a flowchart of the combined electrical fine inspection decision-making process provided in this application embodiment. Under the combined electrical measurement mode, it describes a complete decision-making process for performing local electrical fine inspection and evidence judgment on suspected defect areas. Its core idea is to achieve adaptive and reliable control of electrical verification by screening first, then upgrading, and then verifying consistency under resource-constrained and high-speed operation conditions. Specifically, when a suspected defect area enters the effective measurement window of the electrical detection probe, it first enters the local electrical measurement stage and prioritizes the rapid screening measurement with lower resource consumption and faster response speed to obtain the preliminary electrical response characteristics of the area. Based on the pre-set upgrade criteria (such as electrical response amplitude, deviation from the background, or uncertainty level), it is determined whether it is necessary to further improve the measurement resolution. If the upgrade conditions are met, the system will switch to the precision measurement mode, using higher resolution or higher sensitivity parameter configurations to perform in-depth measurements on suspected areas. If the conditions are not met, the screening results will be maintained and the system will proceed to the subsequent evidence quantification stage. After screening or precision measurement is completed, the measurement results will be uniformly processed for difference feature calculation and verification. The system will further determine whether the obtained difference features are consistent with the physical mechanism of the current measurement mode, thereby improving the overall engineering reliability and interpretability of the detection results.

[0044] like Figure 3The flowchart shown is a flowchart for determining the type of defect area provided in an embodiment of this application. It includes: determining the suspected defect area after completing electrical measurements; after forming an electrical confidence score, constructing a visual feature sub-vector and an electrical feature sub-vector for each suspected defect area that has completed electrical fine-tuning; wherein the visual feature sub-vector at least includes the geometric morphology description, texture perturbation features, and visual confidence score of the area, used to reflect the abnormal performance of the area at the visual level; the electrical feature sub-vector at least includes the resistance or conductivity deviation, thickness or morphology deviation, and electrical confidence score of the area, used to reflect the abnormal evidence of the area at the electrical level; and combining the visual feature sub-vector and the electrical feature sub-vector according to a preset structure. The system performs fusion and standardization to prevent certain features from being unreasonably amplified in the fusion decision due to their large amplitude. The pre-defined structure indicates that fixed segment positions or grouping orders are reserved for visual source feature components and electrical source feature components in the fusion feature vector. The standardization process includes: for each feature component, scaling the measured value based on the value range or fluctuation scale obtained from historical calibration samples or online reference areas, so that the numerical distribution of each feature component is at a uniform order of magnitude, thereby improving the stability of the fusion decision and forming a fusion feature vector that maintains the distinguishability of feature sources. The pre-defined structure indicates that fixed field positions are reserved for visual source features and electrical source features in the fusion vector. Grouping by field allows subsequent judgments to utilize the combined information of both types of features and trace the contribution of each type of evidence to the final conclusion. Interpretable hierarchical judgments are achieved based on the combined relationship between visual and electrical confidence scores: when both visual and electrical confidence scores meet their respective confirmation thresholds, it indicates that the area has significant anomalies at the appearance level and consistent structural evidence at the electrical level; therefore, the suspected defective area is classified as a structural defective area. When visual confidence scores meet the confirmation threshold but electrical confidence scores are below the electrical threshold, it indicates that the area has obvious appearance anomalies but lacks structural electrical evidence; the area is classified as a non-structural anomaly area or marked as an appearance anomaly requiring downgrading, in order to suppress the influence of visual confidence scores. False alarms caused by surface contamination, light disturbance, or process fluctuations; when the visual confidence level is lower than the visual threshold but the electrical confidence level meets the confirmation threshold, it indicates that the area is difficult to identify stably at the visual level but has significant anomalies at the electrical level. This area is identified as a potential hidden defect area or marked as a hidden anomaly requiring review and handling, in order to capture risk defects such as thinning, microcracks, or material anomalies that are difficult to detect visually but may affect battery consistency or safety. When both the visual confidence level and the electrical confidence level are lower than the corresponding threshold, it indicates that the area lacks valid evidence to support the establishment of defects in both visual and electrical aspects. This area is identified as a background area and can be discarded or only low-priority records can be retained according to the scheduling strategy.

[0045] By using an event-triggered approach, the electrical probe does not need to scan the entire roll of material. Instead, it performs local measurements only in high-value areas indicated by visual signals. This significantly reduces the occupancy of electrical channels and improves feasibility on high-speed production lines. It can transform electrical measurements from absolute value judgments into relative anomaly evidence, thereby significantly suppressing false alarms caused by overall material fluctuations or operating condition drift. This improves the stability of verification evidence and enables electrical verification and confirmation of visually suspected defects through a judgment mechanism.

[0046] In Example 2, while keeping everything else unchanged from Example 1, the acquisition method for candidate electrical measurement modes can be achieved through data-driven selection or hierarchical decision-making to enhance the system's adaptability and verification efficiency under different operating conditions and different defect distribution characteristics.

[0047] In implementation method A, candidate electrical measurement modes are obtained by analyzing the correlation data between historical visual features and electrical verification results. During long-term operation, the system continuously accumulates multi-feature description vectors, visual confidence levels, defect category identifiers, corresponding electrical verification results, and final judgment conclusions output from the visual inspection stage. Statistical analysis of this historical data is performed as follows: historical samples are grouped according to visual feature description vectors, defect category identifiers, or visual confidence level intervals, and the judgment effect of different electrical measurement modes in the verification stage is statistically analyzed within each group. By comparing the verification success rate, stability, or misjudgment of different measurement modes under the same visual conditions, a correlation between visual feature conditions and electrical measurement modes is established. The correlation between the verification effect and the method is established. When a new suspected defect area appears, it can be matched with similar samples in historical data based on the visual features of the area. The electrical measurement mode with better verification effect in the sample is selected as the candidate mode. Compared with the method of simply relying on the preset measurement mode selection strategy library, this data-driven acquisition method can reflect the real working condition characteristics and material fluctuation law formed during the long-term operation of the production line. This allows the measurement mode selection to gradually fit the specific equipment status, material batch characteristics and defect distribution changes, and improve the matching degree between the candidate measurement mode and the actual verification needs. It is especially suitable for application scenarios with complex defect morphology, frequent changes in process conditions or difficulty in fully covering empirical rules.

[0048] In implementation method B, candidate electrical measurement modes are obtained through a hierarchical decision-making approach. First, a rapid screening measurement is performed on suspected defect areas, and then a precise measurement mode is triggered based on the screening results. Specifically, after a suspected defect area enters the triggering time window of the electrical probe, a low-resource-consuming and short-execution-time electrical measurement mode is first used to rapidly screen the area to obtain preliminary electrical response information. If the screening results show that the electrical response of the area is not significantly different from the background or does not meet the escalation criteria, the screening results are maintained as a verification conclusion to avoid further consuming probe resources. If the screening results show a significant abnormal trend or meet the preset escalation criteria... If the conditions are met, a higher resolution or higher discrimination capability fine measurement mode will be automatically triggered for the suspected area, and this fine measurement mode will be included in the candidate measurement mode set. Through this hierarchical decision-making method, the electrical measurement process is transformed from a one-time decision to a step-by-step verification process, so that probe resources are prioritized for suspected areas with high risk or high uncertainty. Compared with the method of directly retrieving measurement modes based on the strategy library, this hierarchical decision-making method can effectively reduce the over-measurement of low-risk areas while ensuring the sufficiency of verification in high-risk areas. Thus, it can significantly improve the overall detection throughput and operational stability under high-speed production lines and resource-constrained conditions.

[0049] Compared to the method of directly retrieving candidate measurement modes from a preset measurement mode selection strategy library based on electrical verification evidence targets, the data-driven selection method introduced in Embodiment 2 emphasizes the use of historical operating experience. Its advantage lies in its ability to continuously adapt to long-term changes in the production line and improve the targeting of mode selection. The hierarchical decision-making method emphasizes dynamic judgment in the real-time measurement process. Its advantage lies in the on-demand allocation of measurement resources through a screening and then fine-tuning mechanism. The three methods are not functionally interchangeable, but can be used independently or in combination in different implementation scenarios. This allows the present invention to ensure verification reliability while taking into account system efficiency, resource utilization, and adaptability to complex working conditions.

[0050] like Figure 4 The diagram shown is a structural schematic of an intelligent defect detection system for ultra-thin electrolytic copper foil in the new energy field provided in this application embodiment. It includes: a module for determining suspected defect areas, a module for generating an electrical verification task queue, a module for generating an electrical measurement strategy, and a module for judging suspected results.

[0051] The module for identifying suspected defect areas is used to continuously acquire raw image sequences at a fixed line frequency when the ultrathin electrolytic copper foil passes through the field of view of an industrial camera, preprocess the raw images, and perform visual inspection on the preprocessed images to determine a set of suspected defect areas.

[0052] The module for generating an electrical verification task queue is used to perform spatiotemporal alignment of the set of suspected defective areas obtained from visual inspection, generate an electrical verification task queue based on the roll material operation process, and simultaneously perform scheduling on the task queue.

[0053] The electrical measurement strategy generation module is used to generate electrical measurement strategies for the upcoming electrical verification task. The electrical measurement strategy includes the selected electrical measurement mode, the set electrical scan path, and the number of points.

[0054] The suspected result determination module is used to trigger local electrical measurements through an electrical detection probe according to a generated electrical measurement strategy, obtain local measurement data of the suspected area, and determine the suspected result of the suspected defect area after the electrical measurement is completed.

[0055] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0056] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0057] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0058] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the solution, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of 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.

[0060] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligent detection of defects in ultrathin electrolytic copper foil for the new energy field, characterized in that, Includes the following steps: S1. When the ultra-thin electrolytic copper foil passes through the field of view of the industrial camera, the original image sequence is continuously acquired at a fixed line frequency, the original image is preprocessed, and the preprocessed image is visually inspected to determine the set of suspected defect areas. S2, performs spatiotemporal alignment on the set of suspected defective areas obtained by visual inspection, generates an electrical verification task queue based on the roll material operation process, and performs scheduling on the task queue. S3, Generate an electrical measurement strategy for the upcoming electrical verification task. The electrical measurement strategy includes an electrical measurement mode and a set electrical scan path and number of points. S4. Local electrical measurements are triggered by the electrical detection probe according to the generated electrical measurement strategy to obtain local measurement data of the suspected area, and the suspected defect area that has completed the electrical measurement is judged as a suspected result.

2. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 1, characterized in that: The preprocessing of the original image and the visual inspection of the preprocessed image specifically involve: The original image is compensated for illumination and background based on a set reference image to suppress fixed pattern interference. Adaptive enhancement with limited contrast is performed on the compensated image to highlight the subtle texture features of the copper foil surface; It also performs spatial distortion correction on the image and maps the pixel coordinates to the physical coordinate system of the roll material, thereby providing a consistent spatial reference for cross-frame correlation and defect location tracking; Based on the preprocessed image, a cross-scale local texture perturbation response set is constructed to characterize anomalies on the surface of copper foil. The local texture perturbation response set is used to characterize at least local texture abrupt changes, directional continuity imbalance, and local statistical structure deviation. A uniform scale alignment and difference enhancement process is performed on the cross-scale local texture perturbation response set, and the interference of background regions on anomalous responses is suppressed. The processed response results are then fused to generate a saliency distribution map, which is used to characterize the degree of anomalous saliency of image pixels or local regions relative to their neighborhoods.

3. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 2, characterized in that: The determination yields a set of suspected defective regions, and the specific steps are as follows: Based on the saliency criterion that distinguishes between abnormal responses and background responses, and combined with the spatial correlation between adjacent pixels or local regions, the saliency distribution map is used to extract connected regions where saliency responses are clustered. These connected regions are used to organize pixel-level abnormal responses into candidate regions with spatial continuity, thereby serving as the basic unit for generating suspected defect regions. Calculate the region features for each connected region to form a region-level multi-feature description vector. The region features are used to characterize the region morphology, texture perturbation intensity and response consistency within a preset frame window. Visual confidence is generated based on multi-feature description vectors, and connected regions are classified according to the components representing shape and texture in the multi-feature description vectors. The classification includes at least one of point anomalies, linear deformations, surface perturbations, or roughness anomalies. A set of suspected defect areas is formed, which includes spatial representation information, visual confidence level, and category identification of the suspected defect areas; The spatial representation information includes at least the regional location and regional scale parameters, and additionally outputs regional boundary information under the condition that the regional boundary meets the set conditions.

4. The intelligent defect detection method for ultra-thin electrolytic copper foil in the new energy field as described in claim 1, characterized in that: The process of spatiotemporally aligning the set of suspected defective regions obtained from visual inspection and generating an electrical verification task queue based on the roll material operation process is as follows: For a set of suspected defective areas, a spatiotemporal mapping relationship between image coordinates and electrical probe coordinates is established based on the roll material running trajectory. The expected triggering position and triggering time window of each suspected defective area at the electrical probe are obtained and used as the spatiotemporal mapping result. Based on the spatiotemporal mapping results, an electrical verification task is generated for each suspected defect area. The electrical verification task includes at least the trigger location, the trigger time window, and mode information related to electrical measurement. Based on the arrival order during the roll material operation and / or the predetermined scheduling strategy, multiple electrical verification tasks are organized into an electrical verification task queue. Based on the available processing capacity of the electrical probe within the trigger time window and the allowed mode switching constraints per unit time, the electrical verification task queue is scheduled, and the execution order and processing strategy of each task are output, so that the electrical probe can trigger local measurement of the suspected defect area within the corresponding trigger time window.

5. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 1, characterized in that: The specific steps for generating the electrical measurement strategy for the upcoming electrical verification task are as follows: Based on the initial judgment of defect type in suspected defect areas, visual confidence and multi-feature description vectors, electrical verification evidence targets are determined. Candidate electrical measurement modes are obtained by searching in the preset measurement mode selection strategy library according to the electrical verification evidence targets. Based on the trigger location and trigger time information determined by the spatiotemporal mapping results, and according to the available resource constraints of the electrical probe, the feasibility of candidate measurement modes is screened. From the executable measurement modes, the corresponding measurement mode is selected based on the evaluation results of the ability to discriminate electrical verification evidence targets, and is used as the electrical measurement mode for the suspected defect area. The scanning path is determined based on the spatial morphology and equivalent scale of the suspected defect area. At the same time, the number of sampling points and the point spacing are configured based on the relationship between the scale of the suspected defect area and the sensitive footprint of the electrical probe to meet the preset measurement coverage criteria. Based on the electrical verification evidence target and the on-site noise conditions, configure the dwell time or integration time to keep the measurement uncertainty within a discriminable range, and combine the excitation frequency, frequency band or sampling rate with the physical mechanism of the electrical measurement mode to form an electrical measurement strategy for driving the electrical probe to perform verification.

6. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 5, characterized in that: Candidate electrical measurement modes also include data-driven selection or hierarchical decision-making. The data-driven selection is obtained by analyzing the correlation data between historical visual features and electrical verification results, while the hierarchical decision-making is obtained by performing rapid screening on suspected areas and triggering a precise measurement mode based on the screening results.

7. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 1, characterized in that: The process involves triggering local electrical measurements using an electrical detection probe according to a generated electrical measurement strategy to obtain local measurement data for the suspected area. Specifically: Within the trigger time or time interval corresponding to the suspected defect area, local electrical measurements are triggered on the specified probe channel and scanning path according to the electrical measurement strategy to obtain local measurement data for the suspected defect area. Local reference benchmarks are constructed in the vicinity of the same roll material, and the difference characteristics characterizing the degree of anomaly in the suspected defect area are calculated based on the local measurement data and the local reference benchmarks. Based on the set of suspected defective regions, the differential features are structurally aggregated and physically consistent with each other to generate an electrical confidence score that characterizes the credibility of structural anomalies. The electrical confidence level and the visual confidence level obtained in the visual inspection stage are combined to form joint evidence, which is used to verify and confirm suspected defective areas.

8. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 7, characterized in that: The triggering of local electrical measurements on the designated probe channel and scanning path according to the electrical measurement strategy also includes, when using a combined measurement mode, specifically: First, perform a rapid screening measurement with low resource consumption to obtain an initial electrical response. When the initial electrical response meets the preset upgrade criteria, then perform a more high-resolution fine detection measurement. When the difference features do not meet the physical consistency verification with the measurement mode mechanism, the corresponding suspected defect area is marked as insufficient evidence, and one or a combination of resampling, measurement mode upgrade or delayed review is triggered.

9. The intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in claim 7, characterized in that: The specific process for determining the suspected defect area after completing electrical measurements is as follows: For suspected defect areas that have completed electrical fine inspection, visual feature sub-vectors containing regional geometry, texture perturbation features and visual confidence are constructed, as well as electrical feature sub-vectors containing resistance or conductivity deviation, thickness or morphology deviation and electrical confidence. The visual feature vector and the electrical feature vector are fused according to a preset structure and then standardized to form a fused feature vector that retains the distinguishability of the feature sources. Based on the results of the fused feature vectors, the verification defect types are distinguished according to the combined relationship between visual confidence and electrical confidence. When both visual confidence and electrical confidence meet the corresponding confirmation threshold, the suspected defect area is determined to be a structural defect area. When the visual confidence level meets the confirmation threshold and the electrical confidence level is lower than the electrical threshold, the suspected defect area is judged as a non-structural abnormal area or marked as an appearance abnormality that needs to be downgraded. When the visual confidence level is lower than the visual threshold and the electrical confidence level meets the confirmation threshold, the suspected defect area is identified as a potential hidden defect area or marked as a hidden anomaly that needs to be reviewed and dealt with. When both visual confidence and electrical confidence are below the corresponding threshold, the area is identified as a background area, and the scheduling strategy can be used to either discard or retain only low-priority records.

10. A system applying the intelligent defect detection method for ultrathin electrolytic copper foil in the new energy field as described in any one of claims 1-9, characterized in that, include: The module includes modules for identifying suspected defect areas, generating an electrical verification task queue, generating electrical measurement strategies, and determining suspected results. The module for determining suspected defect areas is used to continuously acquire original image sequences at a fixed line frequency when the ultrathin electrolytic copper foil passes through the field of view of an industrial camera, preprocess the original images, and perform visual inspection on the preprocessed images to determine a set of suspected defect areas. The module for generating an electrical verification task queue is used to perform spatiotemporal alignment of the set of suspected defect areas obtained by visual inspection, generate an electrical verification task queue according to the roll material operation process, and simultaneously perform scheduling on the task queue. The electrical measurement strategy generation module is used to generate an electrical measurement strategy for an upcoming electrical verification task. The electrical measurement strategy includes the selected electrical measurement mode, the set electrical scan path, and the number of points. The suspected result determination module is used to trigger local electrical measurements through an electrical detection probe according to a generated electrical measurement strategy, obtain local measurement data of the suspected area, and determine the suspected result of the suspected defect area after the electrical measurement is completed.

Images