An interlayer bubble data identification method based on multi-piece folding and gluing

Abstract

The application relates to the technical field of data recognition, and discloses an interlayer bubble data recognition method based on multi-piece folding and gluing, which comprises the following steps: layer position separation is performed on the optical signals of a target sample at multiple folding layer positions to obtain independent signals of each layer; then, texture interference is suppressed according to the texture characteristics of each layer of material, pure bubble candidate signals are generated, and two-dimensional bubble distribution maps of each layer are constructed; adjacent layer maps are aligned in the normal direction, and the normal projection overlapping degree of bubble center points is calculated; finally, all layer position data are combined, the overlapping degree is used to distinguish through bubbles and in-layer bubbles, and a recognition result set is output; the application can improve the reliability of the interlayer bubble data recognition based on multi-piece folding and gluing.

Description

Technical Field

[0001] This invention relates to the field of data recognition technology, and in particular to a method for identifying interlayer bubble data based on multi-layer folded bonding. Background Technology

[0002] In the folding and bonding process of multilayer sheets, interlayer bubbles are a typical structural defect, significantly weakening interlayer bonding strength and affecting product appearance quality. However, because folded and bonded products contain multiple stacked bonding interfaces, these interfaces are spatially tightly fitted, causing bubble signals from different layers to overlap during acquisition. Simultaneously, the inherent folding textures and indentation patterns on the sheet surface share similar morphological characteristics with bubble defects, further increasing the difficulty of signal identification. These factors collectively constitute the technical challenge in this field: how to effectively distinguish bubble signals from the material's own texture signals in a multilayered coupled signal environment, and accurately locate the specific spatial layer where the bubble is located, is a pressing technical problem that needs to be solved.

[0003] To address the aforementioned need for identifying interlayer bubbles, existing technical solutions rely on inspectors using magnified observation or standard illumination comparison to determine the presence of bubbles. However, manual inspection has inherent limitations, such as significant subjective judgment differences, fatigue from prolonged work leading to increased false negative rates, and the sampling inspection model cannot cover all products, resulting in insufficient statistical representativeness of the inspection results. Alternatively, the distribution of internal defects can be inferred by measuring the attenuation of light intensity after passing through the sheet. However, such solutions typically only output the total area or number of defects on the entire sheet, failing to pinpoint the identification result to a specific bonding interface layer, thus lacking layer-specific guidance for subsequent process adjustments. Edge detection or region growing algorithms can also be used to extract bubble contours from the acquired surface images. However, these methods are prone to misinterpreting inherent folding textures and indentation patterns on the sheet surface as bubble contours, significantly increasing the false positive rate in areas with dense textures. Furthermore, when the product to be identified contains multiple folded structures, the defect projections at each interface overlap on the acquisition plane, making it impossible to distinguish the depth level of the signal source in a single two-dimensional image acquisition. In these methods, data acquisition and signal processing are separated, and only the transmission results of a single surface or the whole image of the final product are analyzed. There is no mechanism for separating multi-layer signals, which leads to crosstalk between interlayer signals from different depths. The identification results cannot reflect the true spatial layer of the bubble. Secondly, there is a lack of systematic means to suppress interference from material texture. In application scenarios with complex folded textures or multiple superimposed layers, the introduction of a large number of false defect signals seriously affects the identification accuracy and the reliability of the results. Summary of the Invention

[0004] This invention provides a method for identifying interlayer air bubbles based on multi-layer folded bonding, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for identifying interlayer air bubbles based on multi-layer folded bonding, comprising: Step 1: Perform multi-layer signal separation processing on the initial optical signal dataset of the target sample at multiple fold layers to obtain independent layer signal data of the fold layers; Step 2: Based on the material texture characteristics of each folded layer, perform texture interference suppression processing on the signal data of the independent layer to generate pure bubble candidate signals for each layer. Combine the spatial coordinate mapping information of the pure bubble candidate signals to construct a two-dimensional bubble distribution map for each layer. Step 3: Align the two-dimensional bubble distribution maps of adjacent folded layers in space with normal directions, and calculate the projection overlap of the bubble center point of the current layer with the bubble center point of the adjacent layer in the normal direction. Step 4: Combine the two-dimensional bubble distribution maps of all folded layers, and output the identified through-bubble data and intra-layer bubble data as a bubble data identification result set based on the overlap of the bubble center point projection in the interlayer normal direction.

[0006] In one feasible implementation, an online optical bubble detection production line is used for multi-layer folded adhesive separator films in power batteries: The production line folding isolation film consists of three layers of high-polymer transparent sheets that are sequentially hot-pressed, folded, and bonded together to form a two-layer bonding interface. The refractive index of the sheet substrate is fixed at 1.48, the physical thickness of a single sheet is 0.2mm, and the bonding gap between layers is uniformly 0.1mm. A parallel backlight high-speed line scan camera is used to acquire full-area optical images, and the image pixel to physical size calibration ratio is 1 pixel = 0.01mm.

[0007] A bubble-free standard folded film standard sample was prepared in advance, and two pre-calibrations were completed: first, optical signals at different depths of the standard sample were collected to establish a one-to-one mapping table between signal depth and folded layer position; second, a database of the inherent texture direction and folding indentation position of each layer of sheet was collected as a benchmark for texture interference suppression comparison.

[0008] The algorithm has the following preset fixed parameters: texture similarity judgment threshold 0.82, texture signal attenuation coefficient 0.85, and effective judgment length threshold for bubble projection overlap 0.03mm; the signal analysis area is uniformly divided into 50×50 pixel local windows, and the equivalent diameter is calculated using a signal abrupt change amplitude weighted fitting method.

[0009] During production line operation, the line scan camera continuously acquires optical images of the three-layer folded film, which are then aggregated to form an initial optical signal dataset of multiple folded layers. This dataset is then input into the edge industrial control computer to perform multi-layer signal separation processing.

[0010] Based on the fixed physical spacing between layers of the sheet material and the preset refractive index of the substrate, a deep interlayer signal attenuation model based on light transmission loss is constructed. The model uses layer thickness and refractive index as core variables to characterize the light signal transmission attenuation and the superposition law of interlayer signal crosstalk at different depths. The attenuation model is used to generate a convolution kernel that matches the film structure, and a deep demixing deconvolution operation with Tikhonov regularization constraint is performed on the mixed optical signal to suppress the noise amplification problem caused by the deconvolution operation and decompose it into multiple independent signal components corresponding to different depths.

[0011] The depth-layer mapping relationship of the pre-stored standard sample is retrieved. Following the folding arrangement from the surface inwards, precise layer-by-layer matching and calibration are performed on the signal components at each depth, outputting independent layer-by-layer signal data corresponding to the surface, middle, and bottom layers. The processed data completely eliminates optical signal crosstalk from other layers and amplifies the amplitude of characteristic signals generated by abrupt changes in refractive index at the bubble interface, achieving precise layer-by-layer decoupling of multi-layer aliased optical signals.

[0012] Texture interference suppression and bubble map reconstruction are performed layer by layer for the three independent layer signal data.

[0013] First, the pre-stored material texture distribution information for each layer is retrieved, including data on the inherent texture direction of the sheet and the location and extent of folds and indentations generated during hot pressing. The signal data of each independent layer is evenly divided into several 50×50 pixel signal analysis regions according to planar coordinates, with each region corresponding to a local physical area of ​​the film material.

[0014] For each signal analysis region, the signal gradient and continuous distribution features within the region are extracted and compared with the pre-stored standard texture features of that region. Using a texture similarity threshold of 0.82, the signal fluctuations are distinguished as either caused by inherent sheet texture / indentation or abrupt signals from actual bubbles. Regions identified as texture-induced signal interference are marked as texture interference signals. Texture direction data for these regions is read to determine the planar extension direction and range of the texture interference signals. The continuous, gradually changing signal trajectory that perfectly matches the texture direction within the region is precisely located, and all signals on this trajectory are marked as signals to be attenuated.

[0015] A quantitative attenuation coefficient of 0.85 is used to suppress the intensity of the signal to be attenuated within the trajectory, while retaining the irregular local abrupt signal outside the trajectory in the region. At the same time, the residual components of abnormal signals that are not completely eliminated after attenuation are also retained. The effective bubble components after processing all signal analysis areas are fully fused to obtain a pure bubble candidate signal without texture pseudo-defects.

[0016] Traverse all signal abrupt change points within the candidate pure bubble signal, and complete the original acquisition coordinate space positioning based on the camera pixel-physical coordinate calibration relationship; combine the signal abrupt change amplitude and signal duration width of each abrupt change point to fit and calculate the equivalent diameter and fine edge contour of each candidate bubble, and construct a standardized two-dimensional bubble distribution map composed of closed contour lines in each independent planar coordinate system. Each map independently stores the complete information of the spatial distribution, size, and contour of bubbles in a single layer of membrane material.

[0017] The two-dimensional bubble distribution maps of two adjacent folded layers, namely the surface-middle layer and the middle-bottom layer, are paired sequentially, and spatial normal alignment and overlap quantification calculations are performed.

[0018] Using the current layer's coordinate system as a reference, translation and rotation spatial registration are performed on adjacent layers to eliminate coordinate misalignment caused by imaging and membrane assembly offsets, ensuring complete alignment of the two layers' planar coordinate systems along the interlayer normal reference axis. After alignment, the two-dimensional coordinates of all bubble center points in the current layer and all bubble center points in adjacent layers are extracted, and all center point coordinates are uniformly mapped to the interlayer normal reference axis, generating the projected coordinate sets for each bubble layer.

[0019] Traverse and match each target bubble in the current layer with all the contrast bubbles in the adjacent layers to form bubble pairs, and calculate the difference in projected coordinates for each bubble pair; combine the equivalent diameters of the target bubble and the contrast bubbles to solve the corresponding projection coverage area of ​​the two bubbles on the normal axis; determine whether there is an intersection between the two sets of projection intervals, and define the length of the intersection interval as the projection overlap of the bubble pair. If the overlap length is greater than 0.03mm, it is determined that there is a valid cross-layer correspondence between the two sets of bubbles.

[0020] The three layers of two-dimensional bubble maps—surface, middle, and bottom—are stacked in order from the surface inwards to form a complete layer sequence. Based on the bubble projection overlap calculated in the previous step, bubble classification is completed.

[0021] Iterate through all bubbles in each layer. If a bubble has a valid projection overlap with bubbles in adjacent layers, mark it as a through-candidate bubble and bind matching bubbles in the upper and lower layers to form a cross-layer bubble association group. Bubbles without any projection overlap with adjacent layers are directly marked as independent bubbles within the layer. All bubbles in the cross-layer association group are uniformly marked as through-bubble data.

[0022] All penetrating bubbles and independent bubbles within a layer are aggregated across three layers to generate a complete bubble recognition result set. For each bubble, a structured data entry is encapsulated: a spatial layer label is generated based on the bubble's layer number; the number of adjacent layers with consecutive projection overlaps is counted to determine the penetration category; bubbles without overlap are considered intra-layer bubbles, while those with ≥1 consecutive overlapping layers are penetrating bubbles. The number of layers penetrated and the corresponding planar coordinates of each layer are recorded simultaneously; the bubble's center point coordinates, equivalent diameter, and closed edge contour are extracted as bubble morphology parameters. The spatial layer label is bound to the morphology parameters for storage, and finally, a bubble recognition result set with complete structured labels is output.

[0023] The solution was successfully tested and verified on a high-speed mass production line for 8 consecutive hours. The line can perform imaging and inspection of 120 three-layer folded adhesive release films per minute. Traditional single-layer planar optical inspection solutions can only identify bubbles on a single layer and cannot distinguish between independent bubbles within a layer and bubbles that penetrate across the adhesive interface. At the same time, the inherent texture of the sheet and the indentations caused by hot pressing and folding will continuously generate a large number of false defects that interfere with the identification results. The average number of false detections per hour in the actual test reached 136. It does not have the ability to classify defects, and subsequent manual re-inspection of each sheet is still required to distinguish between minor and major defects.

[0024] After adopting the complete identification method of this invention, the multi-layer signal separation algorithm completely eliminates the crosstalk caused by the superposition of optical signals from different layers. Combined with the partitioned directional texture attenuation mechanism, it significantly suppresses the misidentification problems caused by texture and indentation, reducing the number of false detections per hour to 11 and the overall false detection rate to 91.9%. Relying on inter-layer map normal registration and quantitative calculation of bubble projection overlap, it can automatically distinguish between single-layer bubbles, two-layer through bubbles, and three-layer through bubbles, and simultaneously output complete structured data such as the bubble's layer, equivalent size, and closed contour. The entire algorithm only relies on two-dimensional optical images to complete the calculation locally on the edge device, without the need for high-cost equipment such as three-dimensional tomography. The time taken for the entire process of bubble identification on a single sheet of film material is stably controlled within 18ms, which is perfectly matched with the operating cycle of existing high-speed mass production lines. The system outputs classified bubble data that can be directly connected to the production line's automatic sorting equipment. Products with multiple layers of through-bubbles are judged as severely defective and directly rejected, while products with only a single layer of tiny bubbles are marked as slightly defective and sorted for further processing. This significantly reduces the workload of manual re-inspection and achieves fully automated graded quality control of bubble defects in folded laminated film materials.

[0025] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. The method provided by this invention performs multi-layer signal separation processing on the initial optical signal dataset, enabling the signal components of each bonding interface in a folded glued product to be peeled off layer by layer according to their origin depth. This ensures that each folded layer obtains independent signal data free from crosstalk from other layers, thus providing a clear signal foundation for subsequent bubble identification. Furthermore, based on the texture distribution characteristics of each layer's material, this method specifically suppresses signal components in the independent layer signal data that match the texture direction and fold indentation, effectively preserving candidate regions where signal abrupt changes are caused solely by bubble interfaces. Therefore, this method can achieve independent extraction and accurate localization of bubble signals in each layer under the dual conditions of multi-layer signal mixing and complex texture backgrounds.

[0026] 2. This invention constructs a two-dimensional bubble distribution map by mapping the candidate pure bubble signals of each layer to a planar coordinate system. Based on this, it performs normal alignment and projection overlap calculation of adjacent layer maps, achieving quantitative determination of bubbles penetrating multi-layer bonding interfaces. This method can record the number of layers in which each bubble continuously overlaps in the layer sequence, thereby distinguishing between independent bubbles within a layer, double-layer penetrating bubbles, and multi-layer penetrating bubbles, ensuring the identification results fully reflect the actual span of the bubbles in space. Simultaneously, this method constructs a quantitative calculation method for projection overlap based on the difference in projection coordinates of the center points of each bubble and the equivalent diameter parameter, providing a reproducible quantitative basis for determining the degree of penetration. The final output bubble data identification result set includes the spatial layer label and morphological parameters of each bubble, enabling quality assessment personnel to directly locate the specific bonding interface layer and its penetration depth based on the identification results. This provides clear data support for the directional adjustment of subsequent bonding processes, effectively improving the precision level of interlayer bubble data identification and the usability of the results. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below refer to only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0028] Figure 1 This is a flowchart illustrating a method for identifying interlayer bubble data based on multi-layer folded bonding, according to an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0030] Reference Figure 1 The diagram shown is a flowchart illustrating a method for identifying interlayer air bubbles based on multi-layer folded bonding according to an embodiment of the present invention. In this embodiment, the method for identifying interlayer air bubbles based on multi-layer folded bonding includes: Step 1: Perform multi-layer signal separation processing on the initial optical signal dataset of the target sample at multiple fold layers to obtain independent layer signal data of the fold layers; In this embodiment of the invention, the step of performing multi-layer signal separation processing on the initial optical signal dataset of the target sample at multiple fold layers to obtain independent layer signal data for each fold layer includes: The multi-level bit signal separation processing is used to eliminate superposition crosstalk between signals of different levels; Based on the interlayer physical distance after sheet folding and the preset refractive index of the material, an interlayer optical signal attenuation model is constructed. The interlayer signal attenuation model is used to generate a convolution kernel, and a regularized deconvolution calculation is performed on the mixed signal in the initial optical signal dataset to separate the signal components corresponding to different depths. According to the folding arrangement order of the samples from top to bottom, the layer matching and calibration of each depth signal component is completed, and the independent layer signal data corresponding to each fold layer is output. The independent layer signal data eliminates crosstalk of other layer signals and enhances the characteristic signal amplitude corresponding to the abrupt change in refractive index at the bubble interface.

[0031] The interlayer signal attenuation model is a depth attenuation model based on optical propagation transmission loss, and the deconvolution calculation is a depth-unmixed deconvolution operation with regularization constraints, including: The interlayer signal attenuation model is constructed with the interlayer physical thickness and substrate refractive index of each fold layer as the core variables, and is used to characterize the transmission attenuation and crosstalk superposition law of optical signals at different depth layers. A mapping relationship between signal depth and fold layer is established in advance using standard bubble-free folded samples. Based on the mapping relationship, the layer assignment of each depth signal component is accurately determined.

[0032] The execution objective of multi-layer signal separation processing is limited to eliminating optical signal superposition crosstalk from different folded layers. The criterion for superposition crosstalk is that there are two or more optical signals transmitted from layers of different depths at the same imaging position on the same plane. This criterion is formed based on the physical structure of the multi-layer folded stack of the sheet, and all signal separation operations are carried out around this criterion.

[0033] The data collected included the fixed physical distance between each layer of the folded sheet and the fixed refractive index of the material calibrated at the factory stage. These two types of data were used as the input for the construction process of the interlayer optical signal attenuation model. The model was formed based on the light propagation and transmission loss law, which is a depth attenuation model. The core input variables of the depth attenuation model were fixed as the physical thickness between each folded layer and the refractive index of the substrate. After the model was formed, it could fully bear the transmission attenuation change law of optical signals at different depths during the process of penetrating the sheet and the change law of crosstalk formed by the superposition of multiple signals. Both types of laws were fully stored inside the formed interlayer optical signal attenuation model.

[0034] The transmission attenuation and crosstalk superposition rules carried within the formed interlayer optical signal attenuation model are extracted. The complete rules are used as the sole basis for generating the convolution kernel. After the convolution kernel is generated, the optical signal transmission change characteristics recorded by the interlayer optical signal attenuation model are completely preserved. After the convolution kernel is generated, it is directly connected to all the mixed signals stored in the initial optical signal dataset. The generated convolution kernel is completely applied to all the optical data corresponding to the mixed signal to perform deep demixing and deconvolution operation. This operation is bound to regularization constraint logic throughout. The execution benchmark of the regularization constraint is to suppress the noise signal newly generated during the operation. During the operation, all the optical information contained in the mixed signal is split according to the depth level of the optical signal propagation inside the sheet. After the splitting operation is completed, independent signal components corresponding to different depth positions of the sheet are output.

[0035] A standard folded sheet sample free of any bubble defects is prepared in advance. All optical signals output from different depth positions of the standard sample are collected. The depth information of the collected optical signals is bound and stored one by one with the actual folded layer position of the standard sample to form a fixed signal depth and folded layer position mapping relationship. This mapping relationship serves as a unified judgment benchmark for all depth signal component matching and calibration operations. After the mapping relationship is formed, it is permanently retained for subsequent layer position calibration work of all sheets of the same specification.

[0036] The pre-formed signal depth and fold layer mapping relationship is retrieved as the calibration basis. According to the physical folding order of the sheet from the outer surface to the inner bottom layer, each set of depth signal components output by the depth demixing and deconvolution operation is matched and bound one by one with the corresponding fold layer information stored in the mapping relationship. After the matching and binding operation is completed, the independent layer signal data corresponding to each fold layer is output separately. After the independent layer signal data is generated, all optical signal crosstalk information generated by non-fold layers will be completely eliminated. At the same time, the characteristic amplitude of the optical signal generated at the interface of two media with different refractive indices at the bubble position inside the sheet is amplified. The criterion for the amplification of the characteristic amplitude is that there is a fixed difference between the refractive index of the bubble interface medium and the refractive index of the sheet substrate. This difference is used as the basis for the execution of the characteristic amplitude amplification operation.

[0037] Step 2: Based on the material texture characteristics of each folded layer, perform texture interference suppression processing on the signal data of the independent layer to generate pure bubble candidate signals for each layer. Combine the spatial coordinate mapping information of the pure bubble candidate signals to construct a two-dimensional bubble distribution map for each layer. In this embodiment of the invention, the step of performing texture interference suppression processing on the independent layer signal data based on the material texture characteristics of each fold layer to generate pure bubble candidate signals for each layer, and constructing a two-dimensional bubble distribution map for each layer by combining the spatial coordinate mapping information of the pure bubble candidate signals, includes: The texture direction data and folding indentation location data of the corresponding sheet surface in each fold layer are used as material texture distribution information; Based on the material texture distribution information, signal components that match the material texture are identified in the independent layer signal data, and the identified signal components are used as texture interference signals for suppression processing to obtain pure bubble candidate signals; Each signal mutation point in the pure bubble candidate signal is mapped to the spatial coordinate system of the sheet plane. Based on the coordinate position and signal distribution range of each signal mutation point, a two-dimensional bubble distribution map composed of closed contour lines is constructed in the spatial coordinate system of each folded layer.

[0038] The process of obtaining the pure bubble candidate signal further includes: The independent layer signal data is divided into multiple signal analysis regions according to the spatial coordinates of the sheet plane, and each signal analysis region corresponds to a local area on the sheet surface. The signal features within the signal analysis area are compared with the texture features corresponding to the material texture distribution information within the same area to determine whether the signal change within the signal analysis area is caused by the material texture. The corresponding signal components in the region where the signal change is determined to be caused by material texture are marked as texture interference signals, and the texture interference signals are attenuated in the region. The remaining signal components in the region after attenuation are used as candidate bubble signal components in the signal analysis region. The candidate bubble signal components of each signal analysis region are merged to obtain a pure bubble candidate signal that fully covers the entire region of the folded layer.

[0039] The step of using the signal components remaining after attenuation processing within the region as candidate bubble signal components within the signal analysis region includes: The texture direction data in the material texture distribution information within the signal analysis area is obtained, and the extension direction and extension range of the texture interference signal along the sheet plane within the area are determined based on the texture direction data. Based on the extension direction and extension range, a signal trajectory consistent with the texture direction is located in the signal data of the region, and the signal component located on the signal trajectory is taken as the signal to be attenuated. An attenuation operation is applied to the signal to be attenuated to reduce the signal strength of the signal to be attenuated. The signal components located outside the signal trajectory in the region and the signal residue components that are not completely eliminated after the attenuation operation are collectively regarded as candidate bubble signal components.

[0040] Data on the entire texture orientation and folding indentation location of the sheet surface corresponding to a single fold layer are collected. The two types of data are integrated and uniformly named as material texture distribution information. The material texture distribution information storage carrier is bound to the independent layer signal data of the corresponding fold layer, so as to realize the one-to-one binding and matching of single layer texture information and layer optical signal.

[0041] Based on the already bound material texture distribution information as a unified comparison benchmark, all signal components within the complete independent layer signal data of the current folded layer are traversed. Signal components whose signal change trajectory and spatial distribution position completely match the material texture distribution information are selected one by one. After the selection is completed, the matched signal components are uniformly named as texture interference signals. A fixed intensity attenuation suppression operation is performed on all texture interference signals. After the attenuation suppression operation is completed, the clean bubble candidate signal corresponding to the current folded layer is output.

[0042] All signal abrupt change points within the candidate pure bubble signal are extracted. The criterion for determining signal abrupt change points is that there is a fixed amplitude difference between the optical signals acquired before and after the same spatial coordinate position. The original acquired pixel coordinates corresponding to each signal abrupt change point are synchronously converted into physical dimension coordinates of the sheet plane. After the coordinate conversion is completed, all are entered into the sheet plane's dedicated spatial coordinate system for mapping and storage. The coordinate position and signal spatial distribution range corresponding to each signal abrupt change point in the spatial coordinate system are retrieved. The coordinate points are connected sequentially as boundary nodes to form a continuous closed contour line without breaks. After all the closed contour lines are integrated, a two-dimensional bubble distribution map is generated within the dedicated spatial coordinate system of the corresponding folded layer.

[0043] The system reads the sheet plane spatial coordinate division rules that correspond to the complete and independent layer signal data of the current folded layer. The division rules have built-in local area division standards of fixed size. According to the division standards, the complete and independent layer signal data is evenly divided into several independent signal analysis regions. Each signal analysis region corresponds to a continuous local region of fixed size on the sheet surface. The division boundary of the signal analysis region is completely aligned with the physical boundary of the sheet plane.

[0044] All optical signal features stored within a single signal analysis region are retrieved. Simultaneously, the exclusive texture features corresponding to the material texture distribution information matching the spatial coordinates of the signal analysis region are retrieved. The two sets of features are compared point by point according to the same spatial position. The judgment criterion for the comparison operation is a fixed percentage threshold of the overlap points between the signal features and texture features in the total number of points in the region. If the percentage of overlap points reaches the fixed percentage threshold, it is determined that the signal change within the signal analysis region is caused by the material texture.

[0045] All signal components matching texture features within the signal analysis area where the proportion of overlapping points reaches a fixed threshold are marked as texture interference signals. Fixed amplitude attenuation processing is performed only on the marked texture interference signals within the spatial boundary of the current signal analysis area. After the attenuation processing is completed, the remaining signal components within the signal analysis area that did not participate in the attenuation operation are retained. The retained signal components are uniformly named as the candidate bubble signal components corresponding to the signal analysis area.

[0046] The candidate bubble signal components generated by each signal analysis region under the same fold layer are spliced ​​and fused according to the original planar spatial arrangement order of each signal analysis region. The splicing and fusion operation completely covers the entire planar area of ​​the sheet corresponding to the fold layer. After the splicing and fusion operation is completed, a pure bubble candidate signal that completely covers the entire area of ​​the fold layer is generated.

[0047] Lock the single signal analysis area to be processed, retrieve the pre-bound and stored material texture distribution information of the area, extract the texture direction data separately from the material texture distribution information, and record the extension path reference of the sheet texture in the plane of the area completely.

[0048] Based on the extracted texture direction data as a spatial positioning reference, the extension direction and range of the texture interference signal within the sheet plane are completely delineated along the texture extension path. The boundaries of the extension direction and range are determined solely by the physical boundaries recorded in the texture direction data.

[0049] Using the defined extension direction and range as the search boundary, all stored signal data within the signal analysis area are traversed. The overlap between the spatial extension path of the signal data and the texture direction data recording path is checked point by point. The set of continuous points where the spatial extension path completely matches the texture direction is uniformly defined as the signal trajectory.

[0050] All signal components falling within the signal trajectory coverage area are uniformly marked as signals to be attenuated. After marking, an independent data storage area for the signals to be attenuated is separately defined, and the signals are physically spatially distinguished from the signal components in other locations within the signal analysis area.

[0051] For the divided signals to be attenuated, a fixed attenuation amplitude is uniformly executed. The attenuation operation will reduce the inherent signal strength value of the signal to be attenuated according to a preset uniform standard. The standard for reducing the signal strength is a uniform and fixed standard for the entire area, and there are no differentiated adjustment rules.

[0052] All signal components whose spatial location is not within the coverage area of ​​the signal trajectory are separately identified within the signal analysis area. Simultaneously, the residual signal components that remain after the signal to be attenuated has been processed by a unified standard attenuation operation are extracted. After the spatial data of the two types of signal components are merged, they are uniformly named as the candidate bubble signal components corresponding to the signal analysis area.

[0053] Step 3: Align the two-dimensional bubble distribution maps of adjacent folded layers in space with normal directions, and calculate the projection overlap of the bubble center point of the current layer with the bubble center point of the adjacent layer in the normal direction. In this embodiment of the invention, the step of aligning the two-dimensional bubble distribution maps of adjacent folded layers in space according to their normal directions and calculating the projection overlap of the bubble center point of the current layer with the bubble center points of adjacent layers in the normal direction includes: Obtain the two-dimensional bubble distribution map of the current folded layer and the two-dimensional bubble distribution map of the adjacent layers of the adjacent folded layers; The two-dimensional bubble distribution maps of adjacent layers are translated and rotated in space so that the spatial coordinate system of the two-dimensional bubble distribution maps of the adjacent layers is aligned with the spatial coordinate system of the two-dimensional bubble distribution map of the current layer in the normal direction. After completing spatial alignment, the first center point position of each bubble in the current layer two-dimensional bubble distribution map and the second center point position of each bubble in the adjacent layer two-dimensional bubble distribution map are extracted respectively. The first center point position and the second center point position are mapped onto the normal reference axis respectively to obtain the first projection position and the second projection position of each bubble on the normal reference axis. The projection overlap length of the bubble pair is determined according to the projection position relationship of the same group of bubble pairs on the normal reference axis, and the projection overlap length is used as the projection overlap degree of the bubble pair. Based on the relationship between the projection overlap and the bubble size, both bubbles in a bubble pair that meet the penetration condition are marked as penetrating bubble data, and bubbles that do not meet the penetration condition are marked as intralayer bubble data.

[0054] The step of determining the projection overlap length of a bubble pair based on the projection position relationship of the same group of bubble pairs on the normal reference axis, and using the projection overlap length as the projection overlap degree of the bubble pair, includes: The two-dimensional coordinates of each bubble center point in its respective layer plane coordinate system are mapped to the normal reference axis to obtain the projected coordinate values ​​of the bubble center point on the normal reference axis. The projected coordinate values ​​of all bubbles belonging to the same fold layer are combined into a projected coordinate set. Select the projection coordinate value of a target bubble from the projection coordinate set of the current layer, and select the projection coordinate values ​​of comparison bubbles one by one from the projection coordinate sets of adjacent layers. Compare the projection coordinate values ​​of the target bubble with the projection coordinate values ​​of the comparison bubbles to obtain the projection coordinate difference of each bubble pair. Based on the size of the target bubble and the size of the comparison bubble, determine the respective projection coverage area of ​​the target bubble and the comparison bubble on the normal reference axis; Based on the difference in projection coordinates, it is determined whether there is an intersection between the projection coverage area of ​​the target bubble and the projection coverage area of ​​the comparison bubble, and the length of the intersection interval is taken as the projection overlap of the bubble pair.

[0055] Retrieve the current layer's two-dimensional bubble distribution map and the adjacent layer's two-dimensional bubble distribution map after the two-dimensional bubble distribution map construction operation has been completed. Both types of maps are equipped with the corresponding folded layer's exclusive planar spatial coordinate system, and the maps completely store all data related to the bubble's closed outline and center point coordinates.

[0056] Using the planar spatial coordinate system carried by the current layer's two-dimensional bubble distribution map as the spatial alignment reference carrier, the two-dimensional bubble distribution maps of adjacent layers are simultaneously subjected to global spatial translation and global spatial rotation operations. The termination criteria for translation and rotation operations are that all coordinate scales of the adjacent layer's planar coordinate system and the scales of the current layer's planar coordinate system completely coincide along the normal direction perpendicular to the sheet plane. The criteria for scale coincidence are that the deviation of the normal coordinates of the same physical point in the two coordinate systems reaches the preset zero-reset threshold. After the two coordinate systems complete scale coincidence, the normal alignment state is achieved.

[0057] After the two sets of two-dimensional bubble distribution maps are aligned in the normal direction, all closed bubble contours inside the current layer of the two-dimensional bubble distribution map are traversed, and the coordinate information corresponding to the geometric center point of each closed contour is extracted and uniformly named as the first center point position. Simultaneously, all closed bubble contours inside the adjacent layer of the two-dimensional bubble distribution map are traversed, and the coordinate information corresponding to the geometric center point of each closed contour is extracted and uniformly named as the second center point position.

[0058] A fixed normal reference axis is constructed perpendicular to the plane of all folded layers. The normal reference axis is uniformly adapted to the physical stacking direction of all folded layers. The planar coordinates corresponding to the first center point position of each group are projected onto the normal reference axis along the vertical plane direction to generate the first projection position. The planar coordinates corresponding to the second center point position of each group are projected onto the normal reference axis along the vertical plane direction to generate the second projection position. A bubble pair is formed by the same target bubble and the same comparison bubble. The length of the intersection area of ​​the first projection position and the second projection position on the normal reference axis inside the comparison bubble pair is defined as the projection overlap length. The generated projection overlap length is directly used as the projection overlap degree corresponding to the bubble pair.

[0059] The fixed size criterion corresponding to the equivalent diameter of the bubble itself is used as the sole basis for determining the penetration condition. The corresponding projection overlap value of the bubble pair is matched and verified with the bubble size criterion. If the matching result meets the preset penetration threshold, both bubbles in the bubble pair are marked as penetrating bubble data. If the matching result does not meet the preset penetration threshold, the bubble is marked as an intra-layer bubble data.

[0060] Extract the two-dimensional plane coordinate information of the center point of a single bubble, and map it completely onto a unified normal reference axis along a fixed projection direction perpendicular to the plane of the sheet. After the mapping operation is completed, the unique projection coordinate value of the bubble is obtained. The projection coordinate values ​​generated by all bubbles in the same fold layer are integrated and collected according to the original arrangement order of the bubbles in the layer. After collection, the unique projection coordinate set of the corresponding layer is generated.

[0061] Select the projection coordinate value corresponding to a single bubble within the current layer's projection coordinate set as the target bubble's corresponding value. Sequentially retrieve the projection coordinate value corresponding to each bubble within the adjacent layer's projection coordinate set as the comparison bubble's corresponding value. Perform a point comparison operation on the normal reference axis scale line for the two sets of coordinate values. The scale difference generated by the point comparison operation is uniformly named the projection coordinate difference.

[0062] The equivalent diameter information of the target bubble and the equivalent diameter information of the comparison bubble recorded in the two-dimensional bubble distribution map are retrieved. The projection coordinates of the bubble center point are used as the midpoint of the interval, and the normal coverage area corresponding to the equivalent diameter of the bubble is used as the extension boundary of the interval. The exclusive projection coverage interval of the target bubble and the exclusive projection coverage interval of the comparison bubble are defined respectively. The normal reference axis scale is used as the boundary judgment benchmark for both types of intervals.

[0063] Using the scale offset corresponding to the difference in projected coordinates as the reference for interval position comparison, check whether there are continuous shared scale segments on the normal reference axis scale between the target bubble projection coverage interval and the comparison bubble projection coverage interval. The existence of continuous shared scale segments means that the two sets of intervals have an intersection. The scale span length corresponding to the continuous shared scale segments is defined as the projection overlap of the bubble pair.

[0064] Step 4: Combine the two-dimensional bubble distribution maps of all folded layers, and output the identified through-bubble data and intra-layer bubble data as a bubble data identification result set based on the overlap of the bubble center point projection in the interlayer normal direction.

[0065] In this embodiment of the invention, the two-dimensional bubble distribution map of all folded layers is used to output the identified pervasive bubble data and intralayer bubble data as a bubble data identification result set based on the overlap of the bubble center point projection in the interlayer normal direction, including: The two-dimensional bubble distribution maps of each fold layer are arranged in the stacking order to form a layer sequence; Based on the projection overlap and the layer sequence, determine one by one whether each bubble in the current layer has projection overlap with bubbles in adjacent layers; Bubbles with overlapping projections are marked as through-bubbles, and bubbles without overlapping projections are marked as independent bubbles. The through-bubbles are associated with bubbles in adjacent layers that have overlapping projections to form cross-layer bubble association groups, and all bubbles in the cross-layer bubble association groups are marked as through-bubbles data. The data of bubbles marked as through-bubble data and bubbles marked as independent bubbles in all folded layers are aggregated to generate a bubble data identification result set containing through-bubble data and intra-layer bubble data; The bubble data recognition result set is output as the recognition result, which includes bubble spatial layer labels and morphological parameters.

[0066] The identification results include a set of bubble data identification results containing bubble spatial layer labels and morphological parameters, including: Obtain the fold layer attribution information and penetration category information of the current bubble, and combine the fold layer attribution information and penetration category information into the spatial layer label of the current bubble; The fold layer attribution information is determined by the layer number of the fold layer where the bubble is located; The penetration category information is determined based on the number of adjacent layers in the layer sequence where the bubble has consecutive projection overlap: when the number of adjacent layers with consecutive projection overlap is zero, the penetration category information of the bubble is determined to be an intralayer bubble. When the number of adjacent layers with consecutive projection overlap is greater than or equal to one, the penetration category information of the bubble is determined as a penetrating bubble, and the number of layers penetrated by the bubble and the position coordinates of the bubble in each penetrated layer are recorded. The center point coordinates, equivalent diameter, and edge contour data of the current bubble in the two-dimensional bubble distribution map of its respective fold layer are combined to form the morphological parameters of the current bubble; The spatial layer label of the current bubble is associated with and stored with the morphological parameters to form the bubble data entry corresponding to the bubble.

[0067] Retrieve the two-dimensional bubble distribution map after all folded layers have been aligned with the normal direction, and arrange and store it in an orderly manner according to the actual physical stacking order of the sheet from the surface layer to the bottom layer. After the arrangement is completed, it is uniformly named as the layer sequence. The arrangement order of the map inside the layer sequence is completely consistent with the actual folding and stacking order of the material, which serves as the unified sorting benchmark for subsequent cross-layer matching of bubbles.

[0068] Using the order of map arrangement within the layer sequence as the traversal criterion, the projection overlap of each group of adjacent two-dimensional bubble distribution maps is retrieved. For each bubble in the current layer's two-dimensional bubble distribution map, a matching and verification operation is performed individually. The verification criterion is that the projection overlap of the bubble reaches a preset effective overlap threshold, thereby confirming whether there is a projection overlap relationship between the current bubble and the bubbles in the adjacent layers.

[0069] Bubbles that meet the preset overlap threshold after verification are uniformly marked as through candidate bubbles, and bubbles that do not meet the preset overlap threshold after verification are uniformly marked as independent bubbles. A fixed binding relationship is established between through candidate bubbles that meet the projection overlap judgment conditions in the same group. After binding, an independent cross-layer bubble association group is generated. All bubbles in the same cross-layer bubble association group are uniformly marked as through bubble data.

[0070] The bubble data identification results are generated by traversing all two-dimensional bubble distribution maps within the stratigraphic sequence and collecting all labeled through bubble data and all labeled independent bubbles. After the two types of bubble data are fully collected and integrated, a bubble data identification result set is generated. The bubble data identification result set fully includes all the basic identification information of the two types of bubbles with different attributes.

[0071] The integrated bubble data recognition result set is output as the final output carrier of the entire recognition process. The output bubble data recognition result set uniformly carries two types of exclusive information: bubble spatial layer label and bubble morphology parameters, which fully carry all the recognition content of the layer affiliation, penetration attribute and shape size of a single bubble.

[0072] The folded layer attribution information is extracted by locating the layer sequence position corresponding to the two-dimensional bubble distribution map to which a single bubble to be archived belongs. Simultaneously, the penetration judgment result obtained by matching the bubble is retrieved to extract the penetration category information. The two types of information are bound and integrated to generate a unique spatial layer label for the bubble. Each bubble corresponds to a unique set of spatial layer labels and there is no reuse.

[0073] The criterion for extracting and determining the folded stratigraphic attribution information is that the two-dimensional bubble distribution map of the bubble is arranged in a fixed sequence within the stratigraphic sequence. Each layer map corresponds to a unique fixed stratigraphic attribution number, which is directly used as the unique recorded content of the folded stratigraphic attribution information.

[0074] The total number of adjacent layer maps that can be matched with effective projection overlap in the stratigraphic sequence is used as the criterion for determining effective projection overlap. The criterion for determining effective projection overlap follows the preset effective overlap determination threshold of the bubble. This statistical value is directly used to generate through category information.

[0075] When the number of adjacent layer patterns with effective projection overlap of bubble matching is zero, the penetration category information corresponding to the bubble is uniformly recorded as an intra-layer bubble. This recording result corresponds to the physical state that the bubble exists only inside a single folded layer.

[0076] When the number of adjacent layers of the map that effectively overlap with the bubble reaches a preset value of one or more, the penetration category information corresponding to the bubble is uniformly recorded as a penetration bubble. At the same time, the total number of adjacent layers of the map that the bubble can form an effective projection overlap is recorded, and the planar position coordinate information of the bubble in each folded layer map where there is an effective overlap is stored.

[0077] Read the three types of original record information stored in the two-dimensional bubble distribution map of a single bubble: bubble center point coordinates, bubble equivalent diameter, and closed edge contour. After integrating and aggregating the three types of information, name them uniformly as the morphological parameters of the bubble. Each bubble is associated with a unique and complete morphological parameter record.

[0078] A permanent binding storage relationship is established between the spatial layer label generated by a single bubble and the corresponding morphological parameters. After the two types of information are bound and archived, an independent and complete bubble data entry is generated. Each identified bubble corresponds to a unique bubble data entry, and all bubble data entries together constitute the storage body of the complete bubble data recognition result set.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0080] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying interlayer air bubbles based on multi-layer folded bonding, characterized in that, The method includes: Step 1: Perform multi-layer signal separation processing on the initial optical signal dataset of the target sample at multiple fold layers to obtain independent layer signal data for each fold layer, including: The multi-level bit signal separation processing is used to eliminate superposition crosstalk between signals of different levels; Based on the interlayer physical distance after sheet folding and the preset refractive index of the material, an interlayer optical signal attenuation model is constructed. The interlayer optical signal attenuation model is used to generate a convolution kernel, and a regularized deconvolution calculation is performed on the mixed signal in the initial optical signal dataset to separate the signal components corresponding to different depths. According to the folding arrangement order of the samples from top to bottom, the layer matching and calibration of each depth signal component is completed, and the independent layer signal data corresponding to each fold layer is output; the independent layer signal data eliminates crosstalk of other layer signals and enhances the characteristic signal amplitude corresponding to the abrupt change in refractive index at the bubble interface. The interlayer optical signal attenuation model is a depth attenuation model based on optical propagation and transmission loss, and the deconvolution calculation is a depth-unmixed deconvolution operation with regularization constraints, including: The interlayer optical signal attenuation model is constructed with the interlayer physical thickness and substrate refractive index of each fold layer as the core variables, and is used to characterize the transmission attenuation and crosstalk superposition law of optical signals at different depth layers. A mapping relationship between signal depth and fold layer is established in advance using standard bubble-free folded samples. Based on the mapping relationship, the layer assignment of each depth signal component is accurately determined. Step 2: Based on the material texture characteristics of each folded layer, perform texture interference suppression processing on the signal data of the independent layer to generate pure bubble candidate signals for each layer. Combine the spatial coordinate mapping information of the pure bubble candidate signals to construct a two-dimensional bubble distribution map for each layer. Step 3: Align the two-dimensional bubble distribution maps of adjacent folded layers in space with normal directions, and calculate the projection overlap of the bubble center point of the current layer with the bubble center point of the adjacent layer in the normal direction. Step 4: Combine the two-dimensional bubble distribution maps of all folded layers, and output the identified through-bubble data and intra-layer bubble data as a bubble data identification result set based on the overlap of the bubble center point projection in the interlayer normal direction.

2. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 1, characterized in that, The process involves performing texture interference suppression processing on the independent layer signal data based on the material texture characteristics of each fold layer to generate clean bubble candidate signals for each layer. Combining the spatial coordinate mapping information of the clean bubble candidate signals, a two-dimensional bubble distribution map for each layer is constructed, including: The texture direction data and folding indentation location data of the corresponding sheet surface in each fold layer are used as material texture distribution information; Based on the material texture distribution information, signal components that match the material texture are identified in the independent layer signal data, and the identified signal components are used as texture interference signals for suppression processing to obtain pure bubble candidate signals; Each signal mutation point in the pure bubble candidate signal is mapped to the spatial coordinate system of the sheet plane. Based on the coordinate position and signal distribution range of each signal mutation point, a two-dimensional bubble distribution map composed of closed contour lines is constructed in the spatial coordinate system of each folded layer.

3. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 2, characterized in that, The process of obtaining the pure bubble candidate signal further includes: The independent layer signal data is divided into multiple signal analysis regions according to the spatial coordinates of the sheet plane, and each signal analysis region corresponds to a local area on the sheet surface. The signal features within the signal analysis area are compared with the texture features corresponding to the material texture distribution information within the same area to determine whether the signal change within the signal analysis area is caused by the material texture. The corresponding signal components in the region where the signal change is determined to be caused by material texture are marked as texture interference signals, and the texture interference signals are attenuated in the region. The remaining signal components in the region after attenuation are used as candidate bubble signal components in the signal analysis region. The candidate bubble signal components of each signal analysis region are merged to obtain a pure bubble candidate signal that fully covers the entire region of the folded layer.

4. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 3, characterized in that, The step of using the signal components remaining after attenuation processing within the region as candidate bubble signal components within the signal analysis region includes: The texture direction data in the material texture distribution information within the signal analysis area is obtained, and the extension direction and extension range of the texture interference signal along the sheet plane within the area are determined based on the texture direction data. Based on the extension direction and extension range, a signal trajectory consistent with the texture direction is located in the signal data of the region, and the signal component located on the signal trajectory is taken as the signal to be attenuated. An attenuation operation is applied to the signal to be attenuated to reduce the signal strength of the signal to be attenuated. The signal components located outside the signal trajectory in the region and the signal residue components that are not completely eliminated after the attenuation operation are collectively regarded as candidate bubble signal components.

5. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 1, characterized in that, The step of aligning the two-dimensional bubble distribution maps of adjacent folded layers spatially according to their normal directions and calculating the projection overlap of the bubble center point of the current layer with the bubble center points of adjacent layers in the normal direction includes: Obtain the two-dimensional bubble distribution map of the current folded layer and the two-dimensional bubble distribution map of the adjacent layers of the adjacent folded layers; The two-dimensional bubble distribution maps of adjacent layers are translated and rotated in space so that the spatial coordinate system of the two-dimensional bubble distribution maps of the adjacent layers is aligned with the spatial coordinate system of the two-dimensional bubble distribution map of the current layer in the normal direction. After completing spatial alignment, the first center point position of each bubble in the current layer two-dimensional bubble distribution map and the second center point position of each bubble in the adjacent layer two-dimensional bubble distribution map are extracted respectively. The first center point position and the second center point position are mapped onto the normal reference axis respectively to obtain the first projection position and the second projection position of each bubble on the normal reference axis. The projection overlap length of the bubble pair is determined according to the projection position relationship of the same group of bubble pairs on the normal reference axis, and the projection overlap length is used as the projection overlap degree of the bubble pair. Based on the relationship between the projection overlap and the bubble size, both bubbles in a bubble pair that meet the penetration condition are marked as penetrating bubble data, and bubbles that do not meet the penetration condition are marked as intralayer bubble data.

6. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 5, characterized in that, The step of determining the projection overlap length of a bubble pair based on the projection position relationship of the same group of bubble pairs on the normal reference axis, and using the projection overlap length as the projection overlap degree of the bubble pair, includes: The two-dimensional coordinates of each bubble center point in its respective layer plane coordinate system are mapped to the normal reference axis to obtain the projected coordinate values ​​of the bubble center point on the normal reference axis. The projected coordinate values ​​of all bubbles belonging to the same fold layer are combined into a projected coordinate set. Select the projection coordinate value of a target bubble from the projection coordinate set of the current layer, and select the projection coordinate values ​​of comparison bubbles one by one from the projection coordinate sets of adjacent layers. Compare the projection coordinate values ​​of the target bubble with the projection coordinate values ​​of the comparison bubbles to obtain the projection coordinate difference of each bubble pair. Based on the size of the target bubble and the size of the comparison bubble, determine the respective projection coverage area of ​​the target bubble and the comparison bubble on the normal reference axis; Based on the difference in projection coordinates, it is determined whether there is an intersection between the projection coverage area of ​​the target bubble and the projection coverage area of ​​the comparison bubble, and the length of the intersection interval is taken as the projection overlap of the bubble pair.

7. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 1, characterized in that, The two-dimensional bubble distribution map of all folded layers is combined, and based on the overlap of the projection of the bubble center point in the interlayer normal direction, the identified penetrating bubble data that penetrates the bonding interface of multiple folds and the intralayer bubble data are output as a bubble data identification result set, including: The two-dimensional bubble distribution maps of each fold layer are arranged in the stacking order to form a layer sequence; Based on the projection overlap and the layer sequence, determine one by one whether each bubble in the current layer has projection overlap with bubbles in adjacent layers; Bubbles with overlapping projections are marked as through-bubbles, and bubbles without overlapping projections are marked as independent bubbles. The through-bubbles are associated with bubbles in adjacent layers that have overlapping projections to form cross-layer bubble association groups, and all bubbles in the cross-layer bubble association groups are marked as through-bubbles data. The data of bubbles marked as through-bubble data and bubbles marked as independent bubbles in all folded layers are aggregated to generate a bubble data identification result set containing through-bubble data and intra-layer bubble data; The bubble data recognition result set is output as the recognition result, which includes bubble spatial layer labels and morphological parameters.

8. The method for identifying interlayer air bubbles based on multi-layer folded bonding as described in claim 7, characterized in that, The identification results include a set of bubble data identification results containing bubble spatial layer labels and morphological parameters, including: Obtain the fold layer attribution information and penetration category information of the current bubble, and combine the fold layer attribution information and penetration category information into the spatial layer label of the current bubble; The fold layer attribution information is determined by the layer number of the fold layer where the bubble is located; The penetration category information is determined based on the number of adjacent layers in the layer sequence where the bubble has consecutive projection overlap: when the number of adjacent layers with consecutive projection overlap is zero, the penetration category information of the bubble is determined to be an intralayer bubble. When the number of adjacent layers with consecutive projection overlap is greater than or equal to one, the penetration category information of the bubble is determined as a penetrating bubble, and the number of layers penetrated by the bubble and the position coordinates of the bubble in each penetrated layer are recorded. The center point coordinates, equivalent diameter, and edge contour data of the current bubble in the two-dimensional bubble distribution map of its respective fold layer are combined to form the morphological parameters of the current bubble; The spatial layer label of the current bubble is associated with and stored with the morphological parameters to form the bubble data entry corresponding to the bubble.

Images