A battery pole piece drying defect detection method based on infrared thermal imaging

Abstract

This invention discloses a method for detecting drying defects in battery electrodes based on infrared thermal imaging, relating to the field of battery electrode technology. The method includes: acquiring a sequence of infrared thermal images of the battery electrode during continuous transport after drying, and simultaneously acquiring the spatial position and motion state information of the image frames to construct a spatial thermal image data array bound to the frame index; performing boundary extraction and contour analysis on the thermal image data array to identify boundary disturbance regions and generate corresponding structural deformation target regions; within the structural deformation target regions, extracting thermal diffusion directionality information based on temperature distribution characteristics in the multi-frame image sequence, and identifying abnormal thermal diffusion regions. This invention identifies local abnormal heat conduction regions by simultaneously extracting boundary disturbance features and temperature change vector information from the thermal image sequence, constructing a main thermal diffusion direction vector field, and calculating a thermal diffusion symmetry breaking factor based on a preset factor fusion rule.

Description

Technical Field

[0001] This invention relates to the field of battery electrode technology, specifically to a method for detecting drying defects in battery electrodes based on infrared thermal imaging. Background Technology

[0002] In the manufacturing process of lithium-ion batteries, the drying quality of the battery electrodes directly affects the battery's performance, safety, and lifespan. The drying process typically employs a continuous conveying system to heat and dehumidify the electrodes. Due to the fast production cycle, large electrode area, and complex operating environment, electrodes are prone to structural deformation problems such as edge warping, stress concentration, and adhesion during the post-drying conveying stage. These structural abnormalities may interfere with the heat conduction process, thereby affecting the overall drying uniformity of the electrodes. How to identify drying defect areas in real time and non-contact during electrode conveying has become an important research direction for manufacturing quality control.

[0003] Infrared thermal imaging technology has been widely used for temperature detection and surface defect identification of industrial products due to its advantages such as non-contact, full-field, and high resolution. In existing technologies, physical defect areas, such as surface damage, foreign objects, or local undried phenomena, are often identified by analyzing temperature distribution anomalies in infrared images. However, due to the slight structural disturbances that occur during the transport of electrodes, the changes in the heat conduction path may lead to changes in the local heat diffusion direction. These thermal anomalies caused by structural deformation do not have typical temperature deviation characteristics and are difficult to accurately identify and quantify in existing methods based on static thermal image analysis, which may reduce the accuracy of defect identification.

[0004] Existing technologies for detecting drying defects in battery electrodes using infrared thermal imaging primarily focus on the static identification of temperature anomaly regions in thermal images. Defect types are often inferred by judging local temperature rise, hot spot distribution, or cold zone deviation in a single frame image. However, due to dynamic factors such as structural disturbances and posture changes that may occur during electrode transport, relying solely on the temperature distribution of a single frame is often insufficient to accurately distinguish between genuine defects and occasional anomalies caused by background noise. Furthermore, changes in heat diffusion direction caused by structural deformation may spatially couple with thermal anomaly regions due to poor drying. Therefore, this invention proposes a method for detecting drying defects in battery electrodes based on infrared thermal imaging. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting drying defects in battery electrodes based on infrared thermal imaging, so as to solve the problems mentioned in the background art.

[0006] This invention can be achieved through the following technical solution: a method for detecting drying defects in battery electrodes based on infrared thermal imaging, comprising:

[0007] Step 1: Collect infrared thermal image sequences during the continuous transport process after the battery electrodes have dried, and simultaneously acquire the spatial position and motion state information of the image frames to construct a spatial thermal image data array bound to the frame index.

[0008] Step 2: Perform boundary extraction and contour analysis on the thermal image data array to identify boundary disturbance areas and generate corresponding structural deformation target areas;

[0009] Step 3: Within the target area of ​​structural deformation, extract thermal diffusion directionality information based on the temperature distribution features in the multi-frame image sequence, and identify areas of abnormal thermal diffusion.

[0010] Step 4: Perform spatial orientation consistency judgment and temporal phase synchronization analysis on the thermal diffusion anomaly area and the structural deformation target area. If the two are similar in direction and their initial frame time difference is within the preset tolerance range, they are determined to be a composite defect area.

[0011] Step 5: Perform frame order persistence verification on the spatial location and temperature status of the composite defect region in the image frame sequence, filter out the composite defect regions that meet the stability requirements, and generate structured annotation data containing their location, frame index and defect label.

[0012] A further technical improvement of the present invention lies in: the method for constructing the spatial thermal image data array in step one, comprising:

[0013] The continuously acquired infrared thermal images are assigned corresponding frame indices according to the acquisition order, and each frame of infrared thermal image is associated with its spatial location data and motion state information corresponding to its acquisition time, so as to obtain an image frame data unit containing frame index, image data, spatial location field and motion state field.

[0014] Based on the frame index, the image frame data units are arranged sequentially to obtain an ordered data set by frame index;

[0015] Based on the motion state information corresponding to adjacent image frame data units, the order of image frame data units in the data set is corrected to ensure that the image frame data units in the data set are consistent with the actual transport position of the battery electrode.

[0016] The set of image frame data units after sequential correction is determined as a spatial thermal image data array bound to the frame index.

[0017] A further technical improvement of the present invention is that step two, generating the structural deformation target region, includes:

[0018] Boundary contour extraction processing is performed on each image frame in the spatial thermal image data array to obtain the boundary curve data of the battery electrode in the image coordinate system;

[0019] Boundary point sequences are extracted from the boundary curve data at preset sampling intervals to obtain the boundary disturbance change sequence.

[0020] Periodic variation characteristics of the boundary disturbance variation sequence are analyzed to obtain the disturbance amplitude parameters and disturbance frequency parameters at each boundary point;

[0021] Identify continuous boundary segments where the disturbance amplitude is greater than the disturbance amplitude threshold and the disturbance frequency is within the frequency stability range, as the identified boundary disturbance regions;

[0022] For each boundary point in the boundary disturbance region, calculate the local outward normal direction vector based on the spatial arrangement of its neighboring boundary points;

[0023] Based on the outward normal direction vector and according to the preset expansion distance, a region expansion operation is performed in the spatial direction in the image coordinate system to obtain the expanded region;

[0024] The extended region is used as the target region for structural deformation in subsequent thermal anomaly detection steps.

[0025] A further technical improvement of the present invention is that the process of extracting thermal diffusion directionality information and identifying anomalous thermal diffusion regions in step three includes the following steps:

[0026] In the multi-frame spatial thermal image data array corresponding to the structural deformation target area, a temperature change vector field in a two-dimensional spatial coordinate system is constructed based on the pixel temperature change between adjacent image frames.

[0027] The vector direction angles of each pixel position in the temperature change vector field are statistically analyzed to extract the main heat diffusion direction vector within the structural deformation target area;

[0028] Based on the temperature change vector field, the main heat diffusion direction vector, and the spatial geometric main direction vector of the structural deformation target area, pixel-by-pixel calculations are performed within the structural deformation target area to obtain the first sub-factor and the second sub-factor corresponding to each pixel position.

[0029] The first sub-factor is used to reflect the degree of thermal diffusion anisotropy at the pixel location, and the second sub-factor is used to reflect the directional coupling consistency between the main thermal diffusion direction vector and the main spatial geometric direction vector at the pixel location.

[0030] According to the preset factor fusion rules, the first sub-factor and the second sub-factor corresponding to each pixel position are combined to form the thermal diffusion symmetry breaking factor of that pixel position, thereby forming a spatial distribution map of the thermal diffusion symmetry breaking factor in the structural deformation target area.

[0031] Identify connected pixel regions in the spatial distribution map whose thermal diffusion symmetry breaking factor value exceeds the thermal diffusion anomaly identification threshold, and determine the connected pixel regions as thermal diffusion anomaly regions.

[0032] A further technical improvement of the present invention is that the factor fusion rule includes the following steps:

[0033] Within the target area of ​​structural deformation, the first mean and the first standard deviation, as well as the second mean and the second standard deviation, are calculated based on the pixel value distribution of the first sub-factor and the second sub-factor, respectively.

[0034] The first sub-factor is standardized based on the first mean and the first standard deviation to generate the first standardized sub-factor.

[0035] The second sub-factor is standardized based on the second mean and the second standard deviation to generate the second standardized sub-factor.

[0036] Based on the contour coordinate data of the structural deformation target region, the local boundary curvature of each pixel position is calculated;

[0037] Along the direction of the main heat diffusion direction vector, calculate the Euclidean distance from each pixel position to the edge of the structural deformation target region, and use it as the heat diffusion feature distance;

[0038] The corresponding fusion weight coefficients are calculated based on the local boundary curvature and thermal diffusion feature distance at each pixel location.

[0039] Based on the fusion weight coefficient, the first normalized sub-factor and the second normalized sub-factor at each pixel position are weighted and fused to obtain the initial fusion factor distribution map;

[0040] Anisotropic diffusion filtering based on directional structure tensor is performed on the initial fusion factor distribution map;

[0041] The values ​​of each pixel position in the filtered distribution map are used as the thermal diffusion symmetry breaking factor values ​​for that pixel position.

[0042] A further technical improvement of the present invention lies in the following: the process of determining spatial orientation consistency and analyzing temporal phase synchronization in step four includes:

[0043] In the image coordinate system, the main thermal diffusion direction vectors of the thermal diffusion anomaly region and the structural deformation target region are extracted respectively;

[0044] Based on the angular relationship between the main thermal diffusion direction vectors, the spatial directional deviation angle between the thermal diffusion anomaly region and the structural deformation target region is determined.

[0045] The spatial orientation deviation angle is compared with the preset orientation consistency tolerance angle. If the spatial orientation deviation angle is not greater than the orientation consistency tolerance angle, then the two are determined to meet the orientation consistency condition.

[0046] Extract the frame index of the first occurrence of the thermal diffusion anomaly region and the structural deformation target region in the image frame sequence, and calculate the difference between the two frame indexes;

[0047] The frame index difference is compared with the preset frame index tolerance range. If the frame index difference is within the preset frame index tolerance range, it is determined that the two meet the time phase synchronization condition.

[0048] When both the directional consistency condition and the time phase synchronization condition are met, the region corresponding to both the thermal diffusion anomaly region and the structural deformation target region is determined as the composite defect region.

[0049] A further technical improvement of the present invention is that, in step four, when the frame index difference between the thermal diffusion anomaly region and the structural deformation target region is within a preset frame index tolerance range, a time phase synchronization enhancement determination is performed, including:

[0050] Extract the pixel temperature change sequences of the thermal diffusion anomaly region and the structural deformation target region in multiple consecutive image frames after their first appearance;

[0051] Based on the pixel temperature change sequence, the consistency of the fluctuations in the temperature change trends of the two is analyzed to determine whether the two exhibit synchronous change characteristics within a continuous frame interval.

[0052] When the temperature change trend meets the preset trend synchronization judgment condition within a continuous frame interval, it is confirmed that the thermal diffusion anomaly region and the structural deformation target region meet the enhanced time phase synchronization condition.

[0053] A further technical improvement of the present invention is that the process of performing frame sequence persistence verification on the composite defect region in step five includes the following steps:

[0054] In the image frame sequence, the composite defect region is tracked across frames, and its regional location parameter sequence and regional temperature feature sequence are extracted in multiple consecutive image frames.

[0055] Based on the regional location parameter sequence, the spatial displacement of the composite defect region between adjacent image frames is calculated, and the cumulative displacement change value in multiple consecutive image frames is statistically analyzed.

[0056] The cumulative displacement change value is compared with the preset spatial displacement stability threshold.

[0057] Based on the regional temperature feature sequence, the temperature change amplitude of the composite defect region in consecutive image frames is calculated, and its temperature fluctuation range in multiple consecutive image frames is statistically analyzed.

[0058] Compare the temperature fluctuation range with a preset temperature stability threshold;

[0059] When the cumulative displacement change value does not exceed the spatial displacement stability threshold and the temperature fluctuation range does not exceed the temperature stability threshold, the corresponding composite defect region is determined as a stable composite defect region that meets the frame sequence continuity requirement.

[0060] Generate structured annotation data containing the spatial location of stable composite defect regions, corresponding frame indexes, and defect labels.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] This invention identifies local abnormal heat conduction regions and improves the detectability of weak defects by simultaneously extracting boundary perturbation features and temperature change vector information from thermal image sequences, constructing a main heat diffusion direction vector field, and calculating the heat diffusion symmetry breaking factor based on a preset factor fusion rule.

[0063] Furthermore, this invention proposes to perform spatial orientation consistency judgment and temporal phase synchronization analysis on the structural deformation target area and the thermal diffusion anomaly area. Based on the coupling relationship between the spatial orientation angle and the first occurrence time frame index, composite defect areas with synchronous response characteristics are identified. Compared with the method that relies solely on a single frame thermal image for judgment, this method can effectively suppress misjudgments caused by instantaneous disturbances or environmental factors, and improve the robustness and stability of defect identification.

[0064] On the other hand, the present invention also designs a frame sequence persistence verification mechanism for composite defect regions. By analyzing the spatial displacement stability and temperature fluctuation range of defect regions in multi-frame images, stable defect regions with continuous response characteristics are screened out and structured annotation data is output. This mechanism provides a reliable basis for subsequent defect source tracing analysis and quality control, and has good engineering application prospects and industrialization value. Attached Figure Description

[0065] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0066] Figure 1 This is a schematic diagram of the method logic of the present invention. Detailed Implementation

[0067] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0068] Please see Figure 1 As shown, this invention provides a method for detecting drying defects in battery electrodes based on infrared thermal imaging, comprising:

[0069] Step 1: Acquire a sequence of infrared thermal images of the battery electrodes during continuous transport after drying, and simultaneously acquire the spatial location and motion state information of each image frame to construct a spatial thermal image data array bound to a frame index. This step utilizes an infrared imaging system to dynamically and continuously acquire the thermal state of the battery electrodes during transport after drying, forming a time-seriesd infrared thermal image data. Simultaneously, the synchronous acquisition of spatial location and motion state information corresponding to each image frame imbues the image data with traceable physical spatial semantics, providing a foundation for subsequent spatiotemporal correlation analysis of electrode structural disturbances and thermal diffusion behavior. By constructing a spatial thermal image data array bound to a frame index, the orderliness and location of image frames within the sequence are ensured, enabling unified scheduling and analysis of the thermal image data in both time and space dimensions for subsequent processing operations.

[0070] Methods for constructing spatial thermal image data arrays include:

[0071] The continuously acquired infrared thermal images are assigned corresponding frame indices according to the acquisition order, and each frame of infrared thermal image is associated with its spatial location data and motion state information corresponding to its acquisition time, so as to obtain an image frame data unit containing frame index, image data, spatial location field and motion state field.

[0072] Based on the frame index, the image frame data units are arranged sequentially to obtain an ordered data set by frame index;

[0073] Based on the motion state information corresponding to adjacent image frame data units, the order of image frame data units in the data set is corrected to ensure that the image frame data units in the data set are consistent with the actual transport position of the battery electrode.

[0074] The set of image frame data units after sequential correction is determined as a spatial thermal image data array bound to the frame index.

[0075] Specifically, in one embodiment, when monitoring and analyzing the dynamic thermal state of the battery electrode during continuous transport after drying, in order to ensure the consistency of subsequent processing steps in the time and space dimensions, a spatial thermal image data array bound to the frame index is first constructed, and the specific implementation method is as follows.

[0076] Infrared thermal images acquired by the infrared thermal imaging device during the continuous transport of battery electrodes are sequentially assigned corresponding frame indices according to the actual acquisition time. Each frame of infrared thermal image corresponds to a unique frame index value. Simultaneously with frame index assignment, the spatial location data and motion state information corresponding to the acquisition time of that frame of infrared thermal image are recorded. The spatial location data describes the position coordinates of the battery electrode in the image coordinate system, and the motion state information describes the displacement direction and speed of the battery electrode during transport. By associating and recording the frame index, image data, spatial location field, and motion state field, a structured image frame data unit is formed.

[0077] Based on the frame index, all image frame data units are sequentially arranged to obtain an ordered data set by frame index, so as to maintain the continuity of infrared thermal images in the acquisition time dimension and form a clear correspondence between adjacent image frames.

[0078] For a frame-indexed ordered dataset, sequence correction is performed on the image frame data units based on the motion state information recorded in adjacent image frame data units. Specifically, the theoretical displacement of the battery electrode within the time interval is calculated based on the motion speed information recorded between two adjacent image frame data units and the corresponding frame index interval. The theoretical displacement is then compared with the spatial position data recorded in adjacent image frames to obtain the position offset. Subsequently, the spatial position data in subsequent image frame data units is corrected based on the position offset, so that the spatial arrangement of each image frame data unit is consistent with the continuous positional relationship of the battery electrode in the actual transport path.

[0079] The set of image frame data units, after sequential correction, is identified as a spatial thermal image data array bound to the frame index. Through the above processing, the spatial thermal image data array is consistent with the actual transport state of the battery electrodes in both time sequence and spatial location, thus providing a reliable data foundation for subsequent analysis and processing based on structural deformation and thermal diffusion characteristics.

[0080] Step 2: Perform boundary extraction and contour analysis on the thermal image data array to identify boundary disturbance regions and generate corresponding structural deformation target regions. This step extracts and analyzes the outer boundary contours of the battery electrodes in the image to identify regions with periodic or abnormal boundary disturbances, thereby determining local areas with potential structural anomalies (such as warping, adhesion, or curling). Structural deformation may interfere with the thermal diffusion process on the electrode surface, becoming a potential cause of drying defects. Expanding the boundary disturbance regions into structural deformation target regions can effectively limit the spatial scope of subsequent analysis, improve processing efficiency, and enhance the specificity and robustness of anomaly detection.

[0081] Step two, which generates the target region for structural deformation, includes:

[0082] Boundary contour extraction processing is performed on each image frame in the spatial thermal image data array to obtain the boundary curve data of the battery electrode in the image coordinate system;

[0083] Boundary point sequences are extracted from the boundary curve data at preset sampling intervals to obtain the boundary disturbance change sequence.

[0084] Periodic variation characteristics of the boundary disturbance variation sequence are analyzed to obtain the disturbance amplitude parameters and disturbance frequency parameters at each boundary point;

[0085] Identify continuous boundary segments where the disturbance amplitude is greater than the disturbance amplitude threshold and the disturbance frequency is within the frequency stability range, as the identified boundary disturbance regions;

[0086] For each boundary point in the boundary disturbance region, calculate the local outward normal direction vector based on the spatial arrangement of its neighboring boundary points;

[0087] Based on the outward normal direction vector and according to the preset expansion distance, a region expansion operation is performed in the spatial direction in the image coordinate system to obtain the expanded region;

[0088] The extended region is used as the target region for structural deformation in subsequent thermal anomaly detection steps.

[0089] Specifically, in this embodiment, the process of generating the structural deformation target region in step two is implemented as follows, according to the technical steps defined in the claims.

[0090] For the spatial thermal image data array constructed in step one, boundary contour extraction processing is performed on each image frame to obtain the boundary curve data of the battery electrode in the image coordinate system. This boundary curve data is used to describe the outer contour shape of the battery electrode in the corresponding image frame, and it can be formed by sequentially constructing pixel coordinate points.

[0091] After obtaining the boundary curve data, the boundary point sequence is extracted according to a preset sampling interval. Specifically, boundary points are selected sequentially along the boundary curve at fixed spatial intervals to form an ordered sequence of boundary points, which is then used as the basis data for the boundary perturbation change sequence.

[0092] For the boundary perturbation change sequence, a position change sequence of each boundary point in the sequence is constructed across multiple consecutive image frames. Based on this, periodic change feature analysis is performed on the position change sequence corresponding to each boundary point. By analyzing the change characteristics of the position change sequence of each boundary point in the time dimension, the perturbation amplitude parameter and perturbation frequency parameter corresponding to that boundary point are obtained. The perturbation amplitude parameter reflects the maximum positional offset of the boundary point in consecutive frames, and the perturbation frequency parameter reflects the periodicity of the position change of the boundary point.

[0093] After obtaining the disturbance amplitude parameters and disturbance frequency parameters corresponding to each boundary point, the boundary points are judged one by one. The boundary points with disturbance amplitude parameters greater than the disturbance amplitude threshold and disturbance frequency parameters within the preset frequency stability range are identified. Adjacent boundary points that meet the above conditions are combined according to their continuity in the boundary point sequence to form a continuous boundary segment. The continuous boundary segment is then identified as the identified boundary disturbance region.

[0094] For each boundary point in the boundary disturbance region, the corresponding local outward normal direction vector is calculated based on the spatial arrangement relationship between the boundary point and its neighboring boundary points in the image coordinate system. Specifically, for any boundary point in the boundary disturbance region, the preceding and following boundary points in the boundary point sequence are selected, and the local tangent direction at the boundary point is determined based on the spatial coordinate relationship of the three. Furthermore, a direction vector perpendicular to the local tangent direction and pointing outward from the boundary is determined as the local outward normal direction vector of the boundary point.

[0095] After obtaining the local outward normal direction vectors corresponding to each boundary point within the boundary disturbance region, based on the local outward normal direction vectors and according to the preset expansion distance, a spatial expansion operation is performed on the boundary disturbance region in the image coordinate system, thereby generating a corresponding expanded region in the outward normal direction of each boundary point.

[0096] The extended region obtained by the above-mentioned region expansion operation is determined as the structural deformation target region, and the structural deformation target region is used in the subsequent thermal diffusion anomaly detection step.

[0097] Step 3: Within the target area of ​​structural deformation, extract the directionality of heat diffusion based on the temperature distribution features in a multi-frame image sequence and identify areas of abnormal heat diffusion. This step utilizes the temperature variation trend of pixels in multiple consecutive image frames within the target area of ​​structural deformation to extract the directional features of heat diffusion on the electrode surface, and constructs a quantitative index to reflect the local heat conduction mode. If there are abnormal changes in the direction of heat diffusion or enhanced anisotropy, it may indicate that the heat conduction in this area has been affected by physical structural disturbances. By identifying areas that significantly deviate from normal diffusion behavior, potential areas of thermal anomaly can be preliminarily screened, providing a basis for subsequent multimodal fusion judgment.

[0098] Step 3, which involves extracting thermal diffusion directionality information and identifying anomalous thermal diffusion regions, includes the following steps:

[0099] In the multi-frame spatial thermal image data array corresponding to the structural deformation target area, a temperature change vector field in a two-dimensional spatial coordinate system is constructed based on the pixel temperature change between adjacent image frames.

[0100] The vector direction angles of each pixel position in the temperature change vector field are statistically analyzed to extract the main heat diffusion direction vector within the structural deformation target area;

[0101] Based on the temperature change vector field, the main heat diffusion direction vector, and the spatial geometric main direction vector of the structural deformation target area, pixel-by-pixel calculations are performed within the structural deformation target area to obtain the first sub-factor and the second sub-factor corresponding to each pixel position.

[0102] The first sub-factor is used to reflect the degree of thermal diffusion anisotropy at the pixel location, and the second sub-factor is used to reflect the directional coupling consistency between the main thermal diffusion direction vector and the main spatial geometric direction vector at the pixel location.

[0103] According to the preset factor fusion rules, the first sub-factor and the second sub-factor corresponding to each pixel position are combined to form the thermal diffusion symmetry breaking factor of that pixel position, thereby forming a spatial distribution map of the thermal diffusion symmetry breaking factor in the structural deformation target area.

[0104] Identify connected pixel regions in the spatial distribution map whose thermal diffusion symmetry breaking factor value exceeds the thermal diffusion anomaly identification threshold, and determine the connected pixel regions as thermal diffusion anomaly regions.

[0105] Factor fusion rules include the following steps:

[0106] Within the target area of ​​structural deformation, the first mean and the first standard deviation, as well as the second mean and the second standard deviation, are calculated based on the pixel value distribution of the first sub-factor and the second sub-factor, respectively.

[0107] The first sub-factor is standardized based on the first mean and the first standard deviation to generate the first standardized sub-factor.

[0108] The second sub-factor is standardized based on the second mean and the second standard deviation to generate the second standardized sub-factor.

[0109] Based on the contour coordinate data of the structural deformation target region, the local boundary curvature of each pixel position is calculated;

[0110] Along the direction of the main heat diffusion direction vector, calculate the Euclidean distance from each pixel position to the edge of the structural deformation target region, and use it as the heat diffusion feature distance;

[0111] The corresponding fusion weight coefficients are calculated based on the local boundary curvature and thermal diffusion feature distance at each pixel location.

[0112] Based on the fusion weight coefficient, the first normalized sub-factor and the second normalized sub-factor at each pixel position are weighted and fused to obtain the initial fusion factor distribution map;

[0113] Anisotropic diffusion filtering based on directional structure tensor is performed on the initial fusion factor distribution map;

[0114] The values ​​of each pixel position in the filtered distribution map are used as the thermal diffusion symmetry breaking factor values ​​for that pixel position.

[0115] Specifically, in the multi-frame spatial thermal image data array corresponding to the structural deformation target region, a temperature change vector field in a two-dimensional spatial coordinate system is constructed based on the pixel temperature change between adjacent image frames. This vector field is used to characterize the direction and relative intensity of temperature change at each pixel location during the heat conduction process.

[0116] The main heat diffusion direction vector within the target area of ​​structural deformation is extracted by statistically analyzing the vector direction angles of each pixel position in the temperature change vector field. This main heat diffusion direction vector can be obtained by statistically analyzing the frequency of the direction angles of all pixel vectors and selecting the direction vector corresponding to the direction angle with the highest frequency.

[0117] Based on the temperature change vector field, the main heat diffusion direction vector, and the spatial geometric main direction vector of the structural deformation target region, pixel-by-pixel calculations are performed within the structural deformation target region to obtain the first sub-factor and the second sub-factor corresponding to each pixel position.

[0118] The calculation method for the first sub-factor is as follows: taking the current pixel position as the center, extract the temperature change direction of the pixel, and select multiple surrounding pixel positions in its neighboring area to obtain the temperature change direction angle of each position; then, taking the direction angle of the current pixel as the reference, calculate the angle difference between each surrounding pixel and the reference direction in turn; take all angle differences as an angle difference set, and calculate its standard deviation based on the set. The standard deviation is used as the first sub-factor of the current pixel to characterize the dispersion of the heat diffusion direction around the pixel.

[0119] The second sub-factor is calculated as follows: extract the temperature change direction vector of the current pixel position as the local heat diffusion direction; at the same time, obtain the geometric principal direction vector of the structural deformation target area; then calculate the angle between the two direction vectors, and use the cosine value of the angle as the second sub-factor of the pixel position; the cosine value is used to characterize the consistency between the heat diffusion direction and the structural direction, and the closer the value is to 1, the higher the degree of directional coupling.

[0120] Then, according to the preset factor fusion rules, the first sub-factor and the second sub-factor corresponding to each pixel position are combined to form the thermal diffusion symmetry breaking factor of that pixel position, thereby forming a spatial distribution map of the thermal diffusion symmetry breaking factor in the structural deformation target area.

[0121] Anisotropic diffusion filtering based on a directional structure tensor is performed on the initial fusion factor distribution map. This filtering process includes: first, constructing a structure tensor based on the gray-level gradient at each pixel location to determine the main diffusion direction; then, performing anisotropic diffusion according to the dominant direction and local gradient information reflected in the structure tensor to enhance the image's main directional structure and suppress noise perturbations. The filtering can be implemented using a modified Perona-Malik diffusion model.

[0122] Finally, in the filtered thermal diffusion symmetry breaking factor distribution map, connected pixel regions whose values ​​exceed the set thermal diffusion anomaly identification threshold are identified and designated as thermal diffusion anomaly regions.

[0123] The method for setting the threshold for identifying thermal diffusion anomalies is as follows: In multiple normal sample regions, the thermal diffusion symmetry breaking factor values ​​corresponding to all pixel positions are statistically analyzed, and their mean and standard deviation are calculated. The mean plus three times the standard deviation is used as the threshold, thereby ensuring that abnormal regions deviating from normal diffusion behavior are identified under the premise of high statistical confidence. This threshold can also be set to 2.5 based on experiments.

[0124] For example, in a structural deformation target region with a resolution of 256×256 and an image sampling time interval of 0.1s, a thermal diffusion symmetry breaking factor distribution map can be generated using the above method. When the breaking factor values ​​of multiple pixels within a certain connected region are consistently higher than 2.5, the region can be identified as a thermal diffusion anomaly region.

[0125] Step 4: Perform spatial orientation consistency and temporal phase synchronization analysis on the thermal diffusion anomaly region and the structural deformation target region. If their directions are similar and the difference in their initial frame time is within a preset tolerance range, they are identified as a composite defect region. This step, based on the spatiotemporal correspondence between the first two types of anomaly regions, determines whether they belong to different physical representations of the same potential defect event. By comparing whether the main thermal diffusion directions of the thermal diffusion anomaly region and the structural deformation target region are consistent, and assessing whether their first appearance times in the image sequence are close, misjudgments caused by occasional disturbances or background thermal field fluctuations can be ruled out. When the two have coupling characteristics in the spatial direction and exhibit synchronous response in time, the thermal anomaly in this region can be considered to originate from localized poor drying caused by structural deformation, thus identifying it as a composite defect region.

[0126] Step four involves determining spatial orientation consistency and analyzing temporal phase synchronization, including:

[0127] In the image coordinate system, the main thermal diffusion direction vectors of the thermal diffusion anomaly region and the structural deformation target region are extracted respectively;

[0128] Based on the angular relationship between the main thermal diffusion direction vectors, the spatial directional deviation angle between the thermal diffusion anomaly region and the structural deformation target region is determined.

[0129] The spatial orientation deviation angle is compared with the preset orientation consistency tolerance angle. If the spatial orientation deviation angle is not greater than the orientation consistency tolerance angle, then the two are determined to meet the orientation consistency condition.

[0130] Extract the frame index of the first occurrence of the thermal diffusion anomaly region and the structural deformation target region in the image frame sequence, and calculate the difference between the two frame indexes;

[0131] The frame index difference is compared with the preset frame index tolerance range. If the frame index difference is within the preset frame index tolerance range, it is determined that the two meet the time phase synchronization condition.

[0132] When both the directional consistency condition and the time phase synchronization condition are met, the region corresponding to both the thermal diffusion anomaly region and the structural deformation target region is determined as the composite defect region.

[0133] Furthermore, in step four, when the frame index difference between the thermal diffusion anomaly region and the structural deformation target region is within the preset frame index tolerance range, a time phase synchronization enhancement determination is performed, including:

[0134] Extract the pixel temperature change sequences of the thermal diffusion anomaly region and the structural deformation target region in multiple consecutive image frames after their first appearance;

[0135] Based on the pixel temperature change sequence, the consistency of the fluctuations in the temperature change trends of the two is analyzed to determine whether the two exhibit synchronous change characteristics within a continuous frame interval.

[0136] When the temperature change trend meets the preset trend synchronization judgment condition within a continuous frame interval, it is confirmed that the thermal diffusion anomaly region and the structural deformation target region meet the enhanced time phase synchronization condition.

[0137] Specifically, in the image coordinate system, the main thermal diffusion direction vectors of the thermal diffusion anomaly region and the structural deformation target region are extracted respectively. The main thermal diffusion direction vector is a directional feature obtained in the previous step and associated with the corresponding region. It is used to characterize the main propagation direction of thermal diffusion in this region, and this direction vector serves as the basic input for subsequent spatial direction consistency judgment.

[0138] Based on the angular relationship between the main heat diffusion direction vectors, the spatial directional deviation angle between the heat diffusion anomaly region and the structural deformation target region is determined. This spatial directional deviation angle reflects the degree of deviation between the main heat diffusion directions of the two regions. When this spatial directional deviation angle is not greater than a preset directional consistency tolerance angle, the two regions are deemed to meet the directional consistency condition.

[0139] The frame indices of the first appearance of the thermal diffusion anomaly region and the structural deformation target region in the image frame sequence are extracted, and the frame index difference between the two is calculated. The frame index difference is used to describe the time interval between the appearance of the two regions, so as to quantify their temporal phase relationship.

[0140] The frame index difference is compared with the preset frame index tolerance range. If the frame index difference is within the preset frame index tolerance range, it is determined that the two meet the time phase synchronization condition.

[0141] When both the directional consistency condition and the time phase synchronization condition are met, the region corresponding to both the thermal diffusion anomaly region and the structural deformation target region is determined as the composite defect region.

[0142] In the above basic judgment process, when the frame index difference between the thermal diffusion abnormal area and the structural deformation target area is within the preset frame index tolerance range, a time phase synchronization enhancement judgment can be further performed to improve the reliability of composite defect identification.

[0143] Pixel temperature change sequences were extracted from the thermal diffusion anomaly region and the structural deformation target region in multiple consecutive image frames after their first appearance. The pixel temperature change sequence is composed of the temperature change values ​​of the corresponding pixel in adjacent image frames in chronological order, and is used to describe the thermal response evolution process of the region within consecutive frame intervals.

[0144] The temperature change sequences of the two pixels are normalized to ensure that the temperature change sequences of different regions are within a uniform numerical scale, thereby eliminating the influence of absolute temperature differences on trend judgment.

[0145] Determine whether two pixel temperature change sequences meet the trend synchronization criteria. Specifically, within a continuous frame interval, compare the temperature change direction of the two sequences at the same frame position frame by frame. The temperature change direction is determined based on the temperature change state of the current frame relative to the previous frame, including rising, falling, or remaining essentially unchanged. Count the proportion of frames in which the two sequences have the same temperature change direction within the continuous frame interval out of the total number of frames.

[0146] When the proportion is higher than the preset trend synchronization judgment threshold (e.g., 75%), it is determined that the temperature change sequences of the two pixels exhibit synchronous change characteristics within a continuous frame interval, thereby confirming that the thermal diffusion abnormal region and the structural deformation target region meet the enhanced time phase synchronization condition.

[0147] In one specific embodiment, the initial positions of the thermal diffusion anomaly region and the structural deformation target region in the image coordinate system are (100, 150) and (105, 155), respectively. The spatial direction deviation angle between the corresponding main thermal diffusion direction vectors is 5°, which is less than the direction consistency tolerance angle of 10°, thus satisfying the direction consistency condition.

[0148] The first appearance of the two regions in the image frame sequence is at frame 100 and frame 103, respectively. The difference in frame index is 3, which is within the frame index tolerance range of 5 frames, thus satisfying the time phase synchronization condition.

[0149] During the enhancement determination process, the pixel temperature change sequence of the two regions for three consecutive frames after the first appearance of the frame is extracted. After normalization, the frame-by-frame directional consistency is statistically analyzed. The proportion of frames with consistent temperature change direction exceeds 75%, which finally confirms that the enhanced time phase synchronization condition is met, thereby improving the reliability and accuracy of the final composite defect identification.

[0150] Step 5: Perform frame-sequence persistence verification on the spatial location and temperature status of composite defect regions in the image frame sequence, filter out composite defect regions that meet stability requirements, and generate structured annotation data containing their location, frame index, and defect label. This step performs dynamic tracking and stability analysis on the time series of the initially identified composite defect regions. By evaluating their positional changes and temperature fluctuations in multiple consecutive image frames, short-term sporadic interference and unstable hotspots are eliminated, thereby improving the reliability of defect identification. When a region maintains a relatively stable spatial location in consecutive frames and its temperature characteristics do not fluctuate drastically, it indicates that the region may have a real physical defect. The final output of structured annotation data helps to achieve automatic annotation, subsequent statistical analysis, and defect data-driven production optimization.

[0151] The process of performing frame sequence persistence verification on composite defect regions includes the following steps:

[0152] In the image frame sequence, the composite defect region is tracked across frames, and its regional location parameter sequence and regional temperature feature sequence are extracted in multiple consecutive image frames.

[0153] Based on the regional location parameter sequence, the spatial displacement of the composite defect region between adjacent image frames is calculated, and the cumulative displacement change value in multiple consecutive image frames is statistically analyzed.

[0154] The cumulative displacement change value is compared with the preset spatial displacement stability threshold.

[0155] Based on the regional temperature feature sequence, the temperature change amplitude of the composite defect region in consecutive image frames is calculated, and its temperature fluctuation range in multiple consecutive image frames is statistically analyzed.

[0156] Compare the temperature fluctuation range with a preset temperature stability threshold;

[0157] When the cumulative displacement change value does not exceed the spatial displacement stability threshold and the temperature fluctuation range does not exceed the temperature stability threshold, the corresponding composite defect region is determined as a stable composite defect region that meets the frame sequence continuity requirement.

[0158] Generate structured annotation data containing the spatial location of stable composite defect regions, corresponding frame indexes, and defect labels.

[0159] Specifically, cross-frame tracking is performed on identified composite defect regions within the image frame sequence. To achieve accurate association of the same composite defect region across consecutive frames, a cross-frame tracking strategy based on a joint determination of spatial centroid location matching and region shape similarity is adopted. Specifically, in the current image frame, the centroid coordinates of each identified composite defect region are extracted and recorded. In the next image frame, the Euclidean distance between the centroid coordinates of all candidate regions and the centroid of the current region is calculated, and targets within a preset threshold range (e.g., 10 pixels) are selected. Simultaneously, the contour shapes of the current region and candidate regions are compared, and the intersection-union ratio (IU / R) of the contour envelope is used for auxiliary filtering. When the IU / R is greater than a preset similarity threshold (e.g., 0.5), it is determined to be a cross-frame corresponding object of the same composite defect region. Through the above strategy, each composite defect region is matched sequentially between frames, thereby extracting a stable sequence of associated regions across multiple consecutive image frames.

[0160] Based on this, for each composite defect region tracked across frames, its regional location parameters and regional temperature feature sequence are extracted. Regional location parameters include, but are not limited to, centroid coordinates, area, and contour shape. The regional temperature feature sequence is a time series composed of the average temperature values ​​of all pixels within that region in each frame, used to reflect the trend of regional temperature change over time. The average temperature is calculated by summing the temperature values ​​of all pixels within the region and then dividing by the total number of pixels in that region; the result is the regional temperature feature value for that frame. The continuous sequence of these feature values ​​constitutes the regional temperature feature sequence.

[0161] Based on the region location parameter sequence, the spatial displacement of the composite defect region between adjacent image frames is calculated frame by frame. This displacement is obtained by calculating the Euclidean distance between the centroid coordinates of the region in the current frame and the previous frame. The displacements between all consecutive frames are accumulated to obtain the cumulative displacement change value of the region over the entire consecutive frame interval, which is used to measure the spatial stability of the region within the observation period. If the cumulative displacement change value is less than or equal to a preset spatial displacement stability threshold (e.g., 5 pixels), the region has good spatial stability.

[0162] Based on the regional temperature feature sequence, the temperature variation amplitude of the region in consecutive image frames is analyzed. The difference between the maximum and minimum temperature values ​​in the temperature feature sequence is calculated, and this difference is the temperature fluctuation range, used to measure the thermal stability of the region within the observation period. If the temperature fluctuation range does not exceed a preset temperature stability threshold (e.g., 1.5℃), the region is determined to have sufficient stability in terms of thermal response.

[0163] When the cumulative displacement change of a composite defect region does not exceed the spatial displacement stability threshold and its temperature fluctuation range does not exceed the temperature stability threshold throughout the entire continuous frame interval, the region is determined to meet the frame sequence continuity requirement and is identified as a stable composite defect region. This stable composite defect region has higher defect determination reliability and can be output as the final detection result.

[0164] Based on this, the spatial location information (such as centroid coordinates and boundary contours), frame index intervals (start frame and end frame), regional temperature feature sequence, and corresponding defect labels of stable composite defect regions that meet the frame sequence continuity requirements are structured and organized to form structured labeled data, which is used for subsequent downstream processes such as defect storage, alarm triggering, or defect tracing.

[0165] For example, if a composite defect region is stably tracked in frames 110 to 120 of the image, and its centroid has a total displacement of 4.8 pixels between the start and end frames, and the maximum temperature feature sequence is 61.3℃ and the minimum is 60.2℃, then its temperature fluctuation range is 1.1℃. In this case, if the spatial displacement stability threshold is set to 5 pixels and the temperature stability threshold is set to 1.5℃, then the composite defect region meets the frame sequence persistence verification condition and can be confirmed as a stable composite defect region. Its starting coordinates are output as (132, 218), the frame index range is 110 to 120, and the defect label is "structural thermal coupling type drying defect".

[0166] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0167] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for detecting drying defects in battery electrodes based on infrared thermal imaging, characterized in that, include: Step 1: Collect infrared thermal image sequences during the continuous transport process after the battery electrodes have dried, and simultaneously acquire the spatial position and motion state information of the image frames to construct a spatial thermal image data array bound to the frame index. Step 2: Perform boundary extraction and contour analysis on the thermal image data array to identify boundary disturbance areas and generate corresponding structural deformation target areas; Step 3: Within the target area of ​​structural deformation, extract thermal diffusion directionality information based on the temperature distribution features in the multi-frame image sequence, and identify areas of abnormal thermal diffusion. Step 4: Extract the main thermal diffusion direction vector of the thermal diffusion anomaly region and the main thermal diffusion direction vector of the structural deformation target region in the image coordinate system, calculate the direction angle between the two main thermal diffusion direction vectors to obtain the spatial direction deviation angle, and compare the spatial direction deviation angle with the preset direction consistency tolerance angle. When the spatial orientation deviation angle is not greater than the orientation consistency tolerance angle, it is determined that the two meet the spatial orientation consistency condition. At the same time, the frame index of the first appearance of the heat diffusion anomaly region and the structural deformation target region in the image frame sequence is extracted and the frame index difference is calculated. The frame index difference is compared with the preset frame index tolerance range. When the frame index difference is within the frame index tolerance range, it is determined that the two meet the time phase synchronization condition. When both the spatial orientation consistency condition and the time phase synchronization condition are met, the region corresponding to both the heat diffusion anomaly region and the structural deformation target region is determined as a composite defect region. Step 5: Perform frame order persistence verification on the spatial location and temperature status of the composite defect region in the image frame sequence, filter out the composite defect regions that meet the stability requirements, and generate structured annotation data containing their location, frame index and defect label.

2. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 1, characterized in that, The method for constructing the spatial thermal image data array in step one includes: The continuously acquired infrared thermal images are assigned corresponding frame indices according to the acquisition order, and each frame of infrared thermal image is associated with its spatial location data and motion state information corresponding to its acquisition time, so as to obtain an image frame data unit containing frame index, image data, spatial location field and motion state field. Based on the frame index, the image frame data units are arranged sequentially to obtain an ordered data set by frame index; Based on the motion state information corresponding to adjacent image frame data units, the order of image frame data units in the data set is corrected to ensure that the image frame data units in the data set are consistent with the actual transport position of the battery electrode. The set of image frame data units after sequential correction is determined as a spatial thermal image data array bound to the frame index.

3. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 1, characterized in that, Step two, which generates the target region for structural deformation, includes: Boundary contour extraction processing is performed on each image frame in the spatial thermal image data array to obtain the boundary curve data of the battery electrode in the image coordinate system; Boundary point sequences are extracted from the boundary curve data at preset sampling intervals to obtain the boundary disturbance change sequence. Periodic variation characteristics of the boundary disturbance variation sequence are analyzed to obtain the disturbance amplitude parameters and disturbance frequency parameters at each boundary point; Identify continuous boundary segments where the disturbance amplitude is greater than the disturbance amplitude threshold and the disturbance frequency is within the frequency stability range, as the identified boundary disturbance regions; For each boundary point in the boundary disturbance region, calculate the local outward normal direction vector based on the spatial arrangement of its neighboring boundary points; Based on the outward normal direction vector and according to the preset expansion distance, a region expansion operation is performed in the spatial direction in the image coordinate system to obtain the expanded region; The extended region is used as the target region for structural deformation in subsequent thermal anomaly detection steps.

4. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 1, characterized in that, Step 3, which involves extracting thermal diffusion directionality information and identifying anomalous thermal diffusion regions, includes the following steps: In the multi-frame spatial thermal image data array corresponding to the structural deformation target area, a temperature change vector field in a two-dimensional spatial coordinate system is constructed based on the pixel temperature change between adjacent image frames. The vector direction angles of each pixel position in the temperature change vector field are statistically analyzed to extract the main heat diffusion direction vector within the structural deformation target area; Based on the temperature change vector field, the main heat diffusion direction vector, and the spatial geometric main direction vector of the structural deformation target area, pixel-by-pixel calculations are performed within the structural deformation target area to obtain the first sub-factor and the second sub-factor corresponding to each pixel position. The first sub-factor is used to reflect the degree of thermal diffusion anisotropy at the pixel location, and the second sub-factor is used to reflect the directional coupling consistency between the main thermal diffusion direction vector and the main spatial geometric direction vector at the pixel location. According to the preset factor fusion rules, the first sub-factor and the second sub-factor corresponding to each pixel position are combined to form the thermal diffusion symmetry breaking factor of that pixel position, thereby forming a spatial distribution map of the thermal diffusion symmetry breaking factor in the structural deformation target area. Identify connected pixel regions in the spatial distribution map whose thermal diffusion symmetry breaking factor value exceeds the thermal diffusion anomaly identification threshold, and determine the connected pixel regions as thermal diffusion anomaly regions.

5. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 4, characterized in that, Factor fusion rules include the following steps: Within the target area of ​​structural deformation, the first mean and the first standard deviation, as well as the second mean and the second standard deviation, are calculated based on the pixel value distribution of the first sub-factor and the second sub-factor, respectively. The first sub-factor is standardized based on the first mean and the first standard deviation to generate the first standardized sub-factor. The second sub-factor is standardized based on the second mean and the second standard deviation to generate the second standardized sub-factor. Based on the contour coordinate data of the structural deformation target region, the local boundary curvature of each pixel position is calculated; Along the direction of the main heat diffusion direction vector, calculate the Euclidean distance from each pixel position to the edge of the structural deformation target region, and use it as the heat diffusion feature distance; The corresponding fusion weight coefficients are calculated based on the local boundary curvature and thermal diffusion feature distance at each pixel location. Based on the fusion weight coefficient, the first normalized sub-factor and the second normalized sub-factor at each pixel position are weighted and fused to obtain the initial fusion factor distribution map; Anisotropic diffusion filtering based on directional structure tensor is performed on the initial fusion factor distribution map; The values ​​of each pixel position in the filtered distribution map are used as the thermal diffusion symmetry breaking factor values ​​for that pixel position.

6. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 1, characterized in that, In step four, when the frame index difference between the thermal diffusion anomaly region and the structural deformation target region is within the preset frame index tolerance range, a time phase synchronization enhancement determination is performed, including: Extract the pixel temperature change sequences of the thermal diffusion anomaly region and the structural deformation target region in multiple consecutive image frames after their first appearance; Based on the pixel temperature change sequence, the consistency of the fluctuations in the temperature change trends of the two is analyzed to determine whether the two exhibit synchronous change characteristics within a continuous frame interval. When the temperature change trend meets the preset trend synchronization judgment condition within a continuous frame interval, it is confirmed that the thermal diffusion anomaly region and the structural deformation target region meet the enhanced time phase synchronization condition.

7. The method for detecting drying defects in battery electrodes based on infrared thermal imaging according to claim 1, characterized in that, Step five involves performing frame sequence persistence verification on the composite defect region, which includes the following steps: In the image frame sequence, the composite defect region is tracked across frames, and its regional location parameter sequence and regional temperature feature sequence are extracted in multiple consecutive image frames. Based on the regional location parameter sequence, the spatial displacement of the composite defect region between adjacent image frames is calculated, and the cumulative displacement change value in multiple consecutive image frames is statistically analyzed. The cumulative displacement change value is compared with the preset spatial displacement stability threshold. Based on the regional temperature feature sequence, the temperature change amplitude of the composite defect region in consecutive image frames is calculated, and its temperature fluctuation range in multiple consecutive image frames is statistically analyzed. Compare the temperature fluctuation range with a preset temperature stability threshold; When the cumulative displacement change value does not exceed the spatial displacement stability threshold and the temperature fluctuation range does not exceed the temperature stability threshold, the corresponding composite defect region is determined as a stable composite defect region that meets the frame sequence continuity requirement. Generate structured annotation data containing the spatial location of stable composite defect regions, corresponding frame indexes, and defect labels.

Images