An image sensor-based sealed crack recognition method

Abstract

The application discloses a kind of sealing crack identification methods based on image sensor, it is related to visual nondestructive testing field, including the following steps: obtaining the sealing area information of the sealing structure to be detected, image acquisition condition and micro-pressure excitation condition and carry out space calibration;Sealing area image not subjected to micro-pressure excitation is processed and linear region initial screening;Generation and exert pseudo-random code micro-pressure excitation sequence, synchronous trigger image sensor exposure collection;The image of the same sealing detection area under different coding pressure states is aligned in response;Extract the crack response data of suspected linear region;Suspected linear region is screened and non-crack interference is excluded in synchronous response;Result integration is carried out to candidate crack region, and sealing crack identification result is generated.The application utilizes coding micro-pressure excitation and image sensing technology, identifies sealing crack dynamic response, has the advantages of accurate detection, strong anti-interference and high stability.

Description

Technical Field

[0001] This invention relates to the field of visual nondestructive testing, and more particularly to a method for identifying sealing cracks based on an image sensor. Background Technology

[0002] Sealing structures are widely used in components such as battery packs, pipe joints, valve end caps, and precision housings. Surface cracks in the sealing area can affect the airtightness and reliability of the structure. Current methods for identifying sealing cracks mainly rely on manual visual inspection, ordinary camera photography, grayscale threshold segmentation, or edge detection to identify linear marks, boundary breaks, and abrupt changes in grayscale on the sealing surface, and to determine the presence of cracks based on morphological features in static images.

[0003] However, the surface of the sealed area often has machining textures, oil film reflections, pressing marks, and slight scratches. Linear features in static images can easily be confused with real cracks. Some shallow cracks have small openings when not under stress, making them difficult to detect reliably based solely on grayscale differences and edge continuity. Existing methods lack dynamic response assessments corresponding to the pressure application process, and cannot effectively utilize the changes in crack opening, relative edge displacement, and residual opening features during loading, holding, and release, leading to missed detection of shallow cracks and misjudgments due to non-crack interference.

[0004] Therefore, how to provide a method for identifying sealing cracks based on image sensors is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a sealing crack identification method based on an image sensor. This invention utilizes coded micro-pressure excitation and image sensing technology to identify the dynamic response of sealing cracks, and has the advantages of accurate detection, strong anti-interference and high stability.

[0006] A sealing crack identification method based on an image sensor according to an embodiment of the present invention includes the following steps: The system acquires information on the sealing area of ​​the sealing structure to be tested, image acquisition conditions, and micro-pressure excitation conditions. It also spatially calibrates the field of view of the image sensor and the effective area of ​​the micro-pressure excitation unit to generate sealing crack identification configuration data. Based on the configuration data for identifying sealing cracks, the image of the sealing area that is not subjected to micro-pressure excitation is benchmarked and the linear area is initially screened to generate benchmark image data of the sealing area. Based on the configuration data for sealing crack identification, a pseudo-random coded micro-pressure excitation sequence is generated and applied to the sealing structure to be detected. During the micro-pressure excitation process, an image sensor is simultaneously triggered to perform exposure acquisition and generate a coded response image sequence. Based on the reference image data of the sealed area and the encoded response image sequence, the images of the same sealed detection area under different encoded pressure states are aligned to generate pressure response aligned image data. Based on pressure response aligned image data, crack opening displacement field, relative edge displacement, morphological extension state, and residual opening response of suspected linear regions during the coded pressure change process are extracted to generate crack response data. Based on crack response data and pseudo-random coded micro-pressure excitation sequences, synchronous response screening and non-crack interference elimination are performed on suspected linear regions to generate crack candidate region data. Based on the crack candidate region data and crack response data, the results of the candidate crack regions are integrated to generate the sealing crack identification results.

[0007] Optionally, the sealing area information includes the sealing structure type, sealing detection area, sealing contact boundary, and the surface range to be detected; the image acquisition conditions include the image sensor identifier, acquisition resolution, exposure time, acquisition frame rate, and imaging distance; the micro-pressure excitation conditions include pressure amplitude, pressure duration, number of pressure loading cycles, and pressure release interval; and the spatial calibration includes coordinate mapping, boundary matching, and calibration result recording of the field of view of the image sensor and the effective area of ​​the micro-pressure excitation unit.

[0008] Optionally, the generation of the reference image data for the sealed area specifically includes: Based on the configuration data for sealing crack identification, the sealing detection area is extracted from the image of the sealing area that has never been subjected to micro-pressure excitation and its coordinates are marked to generate an image of the sealing detection area. Based on the image of the sealed detection area, image brightness correction and noise suppression are performed on the image brightness distribution and noise distribution to generate a brightness-stable sealed area image; Based on the brightness-stable sealed area image, texture smoothing and contour standardization are performed on the background texture, sealing contact boundary and boundary contour to generate the sealing area reference feature data. Based on the reference feature data of the sealing area, initial screening is performed on linear gray-level abrupt changes, linear edge continuity and texture breaks in the sealing detection area to generate reference image data of the sealing area. The reference image data of the sealing area includes reference gray-level distribution, reference texture distribution, reference boundary contour and initial distribution of suspected linear areas.

[0009] Optionally, the generation of the encoded response image sequence specifically includes: Based on the micro-pressure excitation conditions in the sealing crack identification configuration data, the micro-pressure excitation process of the sealing structure to be detected is divided into pressure action processes, and basic data of pressure nodes are generated. Different pressure nodes in the basic data of pressure nodes are encoded and arranged in a non-equal interval order, and the micro-pressure excitation conditions of each pressure node are associated with the corresponding encoding position to generate a pseudo-random encoded micro-pressure excitation sequence. Based on the pseudo-random coded micro-pressure excitation sequence, the micro-pressure excitation unit is driven to apply micro-pressure excitation to the sealing structure to be tested according to the node arrangement order, generating coded pressure execution data. The coded pressure execution data includes pressure node type, node start time, node end time, pressure amplitude, pressure duration, pressure release interval, and pressure status identifier. Based on the coded pressure execution data and image acquisition conditions, the execution time of each pressure node is synchronously bound to the exposure time of the image sensor to generate pressure node synchronization trigger data. The pressure node synchronization trigger data includes pressure node type, node execution time, exposure time, image acquisition frame number, pressure status identifier, and synchronization binding relationship. Based on the pressure node synchronous trigger data, images of the sealing area are synchronously acquired under different coded pressure states, and the sequence is organized according to the acquisition time and pressure state identifier to generate a coded response image sequence. The coded response image sequence includes images of the sealing area under different coded pressure states, image acquisition time, image acquisition frame number, pressure state identifier, and pressure node type.

[0010] Optionally, the generation of the pressure response aligned image data specifically includes: Based on the reference image data of the sealed area, the corresponding image blocks of the same sealed detection area under different coding pressure states are located from the encoded response image sequence, and the corresponding image data of the area is generated. Based on the region-corresponding image data, boundary coordinates are compared and region positions are registered for image blocks of the sealed region under different coded pressure conditions to generate sealed region position registration data. Based on the registration data of the sealed area position, the image blocks of the registered sealed area are arranged in the acquisition time sequence according to the image acquisition time and pressure status identifier to generate image acquisition time sequence alignment data. Based on the image acquisition time-series aligned data, taking the initial distribution of suspected linear regions as the tracking entry point, local region tracking is performed on the same suspected linear region under continuous encoding pressure to generate suspected linear region tracking data. Based on suspected linear region tracking data, image blocks of the same suspected linear region under different coding pressure states are subjected to brightness uniformity processing and pressure state label matching to generate pressure response aligned image data.

[0011] Optionally, the generation of the crack response data specifically includes: Based on pressure response aligned image data, the two edges of the same suspected linear region under different coded pressure states are located to generate linear region edge localization data. Based on the edge localization data of the linear region, the opening displacement of the two sides of the edge under different coded pressure states is extracted to generate the crack opening displacement field. Based on the coordinates of the two sides of the linear region edge localization data and the edge sampling position in the crack opening displacement field, the position changes of the two sides of the same suspected linear region under different coded pressure states are compared to generate relative edge displacement data. Based on edge relative displacement data and pressure response aligned image data, the contour extension and endpoint changes of suspected linear regions are extracted to generate morphological extension states. Based on the crack opening displacement field and morphological extension state, the opening state after pressure release is correlated with the initial opening state under the condition of no micro-pressure excitation, and the residual opening response is generated. Based on the pressure state identifier, the crack opening displacement field, edge relative displacement data, morphological extension state, and residual opening response are collected to generate crack response data.

[0012] Optionally, the generation of the crack candidate region data specifically includes: Based on crack response data and pseudo-random encoded micro-pressure excitation sequence, the crack opening displacement change in the crack opening displacement field is matched with the pressure loading node in time sequence and state to generate loading opening matching data. Based on crack response data and pseudo-random encoded micro-pressure excitation sequence, the residual opening response is matched with the pressure release node in time sequence and response to generate release residual matching data. Based on crack response data and pseudo-random coded micro-pressure excitation sequence, the morphological extension state is continuously corresponded and matched with the continuously coded pressure state to generate continuous extension matching data. Based on the loading opening matching data, releasing residual matching data, and continuous extension matching data, suspected linear regions are synchronously screened and non-crack interference is eliminated to generate crack candidate region data. The crack candidate region data includes candidate crack location, candidate crack outline, candidate crack response identifier, and non-crack interference elimination identifier.

[0013] Optionally, the generation of the sealing crack identification result specifically includes: Based on crack candidate region data and crack response data, interference screening, similar grouping and contour correction are performed on the candidate crack regions to generate crack location and crack contour. Based on the morphological extension state in the crack response data, the extension direction and endpoint change direction of the crack profile are aggregated to generate the crack extension direction. Based on the crack opening displacement field and edge relative displacement in the crack response data, the opening changes of the same candidate crack region under different coded pressure states are collected to generate crack opening states. Based on the residual opening response in the crack response data, the residual opening state after pressure release and the initial opening state under the condition of no micro-pressure excitation are residually collected to generate the crack residual response state. The crack location, crack profile, crack extension direction, crack opening state and crack residual response state are integrated to generate the sealing crack identification result.

[0014] The beneficial effects of this invention are: This invention introduces a pseudo-random coded micro-pressure excitation sequence during the identification of sealing cracks. This causes the sealed structure under test to generate collectable image response changes under different coded pressure states, such as pressure loading, pressure holding, pressure release, and pressure intervals. The coded response image sequence is then simultaneously captured by an image sensor. Compared to identification methods that rely solely on static grayscale, edge, or texture features under no-stress conditions, this invention utilizes the opening changes, relative edge displacements, and residual opening responses of real cracks under micro-pressure to transform previously inconspicuous shallow cracks into objects with dynamic response characteristics, thereby improving the detection capability of low-contrast, shallow-opening sealing cracks.

[0015] This invention generates baseline image data of the sealed area by benchmarking images of unexcited sealing regions and initially screening linear regions. Furthermore, it performs response alignment on images under different coded pressure states, reducing the impact of image brightness fluctuations, sensor noise, surface reflection, and slight positional shifts on crack identification. Based on this, it extracts the crack opening displacement field, relative edge displacement, morphological extension state, and residual opening response from suspected linear regions. This allows crack identification to move beyond a single linear appearance and incorporate multi-dimensional response characteristics under pressure, improving the stability and reliability of sealing crack identification results.

[0016] This invention also performs synchronous response screening of crack response data and pseudo-random coded micro-pressure excitation sequences, and eliminates non-crack interference such as scratches, stains, processing textures, and lighting shadows that do not form candidate crack response identifiers. Through interference screening, similar merging, and contour correction of candidate crack regions, the final generated sealing crack identification result can simultaneously include crack location, crack contour, crack extension direction, crack opening state, and crack residual response state, avoiding misjudging ordinary surface linear marks as cracks, and improving the identification accuracy in factory inspection, online inspection, and rework determination of sealing structures. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a sealing crack identification method based on an image sensor proposed in this invention; Figure 2 This is a flowchart illustrating the generation of crack candidate region data in a sealing crack identification method based on an image sensor proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-2 A method for identifying sealing cracks based on an image sensor includes the following steps: The system acquires information on the sealing area of ​​the sealing structure to be tested, image acquisition conditions, and micro-pressure excitation conditions. It also spatially calibrates the field of view of the image sensor and the effective area of ​​the micro-pressure excitation unit to generate sealing crack identification configuration data. Based on the configuration data for identifying sealing cracks, the image of the sealing area that is not subjected to micro-pressure excitation is benchmarked and the linear area is initially screened to generate benchmark image data of the sealing area. Based on the configuration data for sealing crack identification, a pseudo-random coded micro-pressure excitation sequence is generated and applied to the sealing structure to be detected. During the micro-pressure excitation process, an image sensor is simultaneously triggered to perform exposure acquisition and generate a coded response image sequence. Based on the reference image data of the sealed area and the encoded response image sequence, the images of the same sealed detection area under different encoded pressure states are aligned to generate pressure response aligned image data. Based on pressure response aligned image data, crack opening displacement field, relative edge displacement, morphological extension state, and residual opening response of suspected linear regions during the coded pressure change process are extracted to generate crack response data. Based on crack response data and pseudo-random coded micro-pressure excitation sequences, synchronous response screening and non-crack interference elimination are performed on suspected linear regions to generate crack candidate region data. Based on the crack candidate region data and crack response data, the results of the candidate crack regions are integrated to generate the sealing crack identification results.

[0020] In this embodiment, the sealing area information includes the sealing structure type, sealing detection area, sealing contact boundary, and the range of the surface to be detected. The image acquisition conditions include the image sensor identifier, acquisition resolution, exposure time, acquisition frame rate, and imaging distance. The micro-pressure excitation conditions include the pressure amplitude, pressure duration, number of pressure loading cycles, and pressure release interval. The spatial calibration includes mapping the field of view of the image sensor to the effective area of ​​the micro-pressure excitation unit, matching the area boundaries, and recording the calibration results.

[0021] In this embodiment, the generation of the reference image data of the sealed area specifically includes: Based on the configuration data for identifying sealing cracks, the sealing detection area is extracted from the image of the sealing area that has never been subjected to micro-pressure excitation and the coordinates are marked to generate a sealing detection area image. The sealing detection area image includes a sealing area image block, sealing detection area coordinates, sealing contact boundary coordinates, the coordinates of the surface range to be detected, and the original grayscale distribution. The original grayscale distribution is formed by arranging the grayscale values ​​of each pixel in the sealing area image block according to the pixel coordinates. Based on the image of the sealing detection area, image brightness correction and noise suppression are performed on the image brightness distribution and noise distribution to generate a brightness-stable sealing area image; image brightness correction is based on the difference between the gray value at each position in the sealing detection area and the average gray value of the area, and noise suppression is used to reduce dark noise of the image sensor, supplementary light fluctuation noise and surface reflection noise. In practice, the average grayscale value of the sealed detection area is 120, and the grayscale value of the pixel to be corrected is 132. Subtracting 120 from 132 yields a grayscale difference of 12, and the grayscale value of the pixel to be corrected is adjusted from 132 to 120. When the grayscale value of a noise pixel is 170 and the average grayscale value of its surrounding pixels is 122, the grayscale value of the noise pixel is adjusted to 122, thus forming a stable brightness sealed area image. Based on the image of the brightness-stable sealed area, texture smoothing and contour standardization are performed on the background texture, sealing contact boundary, and boundary contour to generate reference feature data of the sealed area. Specifically, neighborhood gray-level equalization is performed on the background texture area to reduce the gray-level fluctuations of non-crack textures, and a reference gray-level distribution and reference texture distribution are formed based on the pixel gray-level values ​​after neighborhood gray-level equalization. Contour points are sorted, breakpoints are connected, and coordinate scales are unified for the sealing contact boundary and the edge of the surface to be detected to generate the reference boundary contour. In specific implementation, the grayscale value of a pixel in the background texture area is 126, and the average grayscale value of the neighboring areas of this pixel is 118. The grayscale value of this pixel is corrected from 126 to 118 to form the pixel grayscale value after neighborhood grayscale equalization. The horizontal coordinates of the contour points on the sealed contact boundary are 120, 180 and 240 respectively, and the vertical coordinates are all 80. The contour points are sorted and connected in ascending order of horizontal coordinates to generate the baseline boundary contour. Based on the reference feature data of the sealing area, initial screening is performed on linear gray-level abrupt changes, linear edge continuity, and texture breaks in the sealing detection area to generate reference image data of the sealing area. The reference image data of the sealing area includes reference gray-level distribution, reference texture distribution, reference boundary contour, and initial distribution of suspected linear areas. The initial distribution of suspected linear areas is formed by merging areas that overlap in position or have the same extension direction in the areas of slender gray-level abrupt changes, edge continuity, and texture breaks. The initial distribution of suspected linear areas includes the coordinates of suspected linear areas, the contour of suspected linear areas, and the linear extension direction. The initial screening includes screening for elongated gray-level abrupt change regions, screening for continuous edge regions, and screening for texture break regions. Elongated gray-level abrupt change regions are pixel regions in the baseline gray-level distribution where the gray-level values ​​change continuously and are arranged in a linear pattern along the same extension direction. Continuous edge regions are linear edge regions formed by connecting adjacent edge points in sequence according to pixel coordinates. Texture break regions are regions in the baseline texture distribution where the surface texture is interrupted, misaligned, or changes direction.

[0022] In this embodiment, the generation of the encoded response image sequence specifically includes: Based on the micro-pressure excitation conditions in the sealing crack identification configuration data, the micro-pressure excitation process of the sealing structure to be detected is divided into pressure action processes to generate basic pressure node data. The basic pressure node data includes pressure loading nodes, pressure holding nodes, pressure release nodes, and pressure interval nodes. The pressure loading node is the process by which the micro-pressure excitation unit increases from a state of no pressure action to the pressure amplitude. The pressure holding node is the process by which the micro-pressure excitation unit maintains the pressure amplitude and continues to act according to the pressure duration. The pressure release node is the process by which the micro-pressure excitation unit releases the pressure amplitude and returns to a state of no pressure action. The pressure interval node is the no-pressure action period formed according to the pressure release interval from the end of one pressure release node to the beginning of the next pressure loading node. Different pressure nodes in the basic pressure node data are encoded and arranged in a non-equal interval order, and the micro-pressure excitation conditions of each pressure node are associated with the corresponding encoding position to generate a pseudo-random encoded micro-pressure excitation sequence. The pseudo-random encoded micro-pressure excitation sequence includes the node arrangement order, pressure amplitude, pressure duration, number of pressure loading times, pressure release interval, and pressure status identifier. The non-equal interval sequence generates a corresponding number of pressure code positions through the number of pressure loadings. Then, according to the order of the pressure code positions, the pressure release interval is added to the pressure duration that is a multiple of the corresponding number to obtain the time interval between adjacent pressure nodes. Finally, the pressure nodes are arranged in order according to the time interval. In specific implementation, the micro-pressure excitation conditions are: pressure amplitude of 2 kPa, pressure duration of 0.3 seconds, pressure loading number of 6 times, and pressure release interval of 0.2 seconds. Based on the 6 pressure loading times, 6 pressure coding positions are generated. According to the sequential number of the pressure coding positions, the 0.2-second pressure release interval is added to 1, 2, 3, 4, 5, and 6 times the 0.3-second pressure duration, respectively, to obtain non-equidistant time intervals between adjacent pressure nodes of 0.5 seconds, 0.8 seconds, 1.1 seconds, 1.4 seconds, 1.7 seconds, and 2.0 seconds, respectively. Based on the pseudo-random coded micro-pressure excitation sequence, the micro-pressure excitation unit is driven to apply micro-pressure excitation to the sealing structure to be tested according to the node arrangement order, generating coded pressure execution data; the coded pressure execution data includes pressure node type, node start time, node end time, pressure amplitude, pressure duration, pressure release interval, and pressure status identifier; Micro-pressure excitation refers to the application of low-amplitude pressure by a micro-pressure excitation unit to the sealing structure to be tested, according to the pressure amplitude, pressure duration, number of pressure loading cycles, and pressure release intervals. This is used to make the actual cracks in the sealing area produce opening changes, relative edge displacements, and residual opening responses that can be captured by an image sensor during pressure loading, pressure holding, pressure release, and pressure intervals. Based on the coded pressure execution data and image acquisition conditions, the execution time of each pressure node is synchronously bound to the exposure time of the image sensor to generate pressure node synchronization trigger data. The pressure node synchronization trigger data includes pressure node type, node execution time, exposure time, image acquisition frame number, pressure status identifier, and synchronization binding relationship. The exposure time of the image sensor is calculated by adding the product of the synchronization trigger start time and the image acquisition frame number and the reciprocal of the acquisition frame rate. The synchronization trigger start time is the moment when the pseudo-random coded micro-pressure excitation sequence starts executing and sends a synchronization trigger signal to the image sensor, which corresponds to the node start time of the first pressure node. In specific implementation, the acquisition frame rate in the image acquisition conditions is set to 100 frames per second, and the synchronization trigger start time is set to 0 seconds. When the image acquisition frame number is 25, the corresponding image sensor exposure time is 0.25 seconds. The synchronization binding relationship is obtained by matching the time interval between the exposure time of the image sensor and the start time and end time of the node. When the exposure time of the image sensor falls within the execution time period of the corresponding pressure node, a synchronization binding relationship is formed between the corresponding image acquisition frame number and the pressure node. Based on the pressure node synchronous trigger data, images of the sealed area are synchronously acquired under different coded pressure states, and the sequence is organized according to the acquisition time and pressure state identifier to generate a coded response image sequence. The coded response image sequence includes images of the sealed area under different coded pressure states, image acquisition time, image acquisition frame number, pressure state identifier, and pressure node type.

[0023] In this embodiment, the generation of pressure response aligned image data specifically includes: Based on the reference image data of the sealed area, the corresponding image blocks of the same sealed detection area under different coding pressure states are located from the coding response image sequence, and the corresponding image data of the area is generated. The corresponding image data of the area includes the reference grayscale distribution, the reference boundary contour, the initial distribution of the suspected linear area, the image blocks of the sealed area under different coding pressure states, the image acquisition time and the pressure state identifier. In practice, the horizontal coordinate range of the sealing detection area is 120 to 720, and the vertical coordinate range is 80 to 400. The corresponding sealing area image blocks in the pressure interval state, pressure loading state, pressure holding state and pressure release state are located from the coded response image sequence. The image acquisition times are 0 seconds, 0.01 seconds, 0.02 seconds and 0.03 seconds respectively. Based on region-corresponding image data, boundary coordinate comparison and region position registration are performed on sealing region image blocks under different coded pressure states to generate sealing region position registration data. Specifically, using the reference boundary contour as the registration reference, the sealing contact boundary coordinates are extracted from the sealing region image blocks under different coded pressure states. The sealing contact boundary coordinates under different coded pressure states are compared point by point with the corresponding boundary coordinates in the reference boundary contour to obtain the lateral coordinate difference and longitudinal coordinate difference of each boundary point. The sealing region image blocks under the corresponding pressure states are then translated according to the lateral coordinate difference and longitudinal coordinate difference to ensure that the sealing detection area under different coded pressure states maintains coordinate correspondence with the reference boundary contour. The sealing region position registration data includes the registered sealing region image blocks, registration coordinates, lateral coordinate difference, longitudinal coordinate difference, and pressure state identifier. In specific implementation, the horizontal coordinate of a boundary point in the baseline boundary contour is 150 and the vertical coordinate is 100. The horizontal coordinate of the corresponding sealing contact boundary point under the coded pressure state is 153 and the vertical coordinate is 98. Thus, the difference in horizontal coordinates is 3 pixels and the difference in vertical coordinates is -2 pixels. The image block of the sealing area under the corresponding pressure state is then translated as a whole, so that the translated horizontal coordinate is equal to the original horizontal coordinate minus 3 pixels, and the translated vertical coordinate is equal to the original vertical coordinate plus 2 pixels. Based on the registration data of the sealed area, the registered image blocks of the sealed area are arranged in the acquisition time sequence according to the image acquisition time and pressure status identifier to generate image acquisition time sequence alignment data. The image acquisition time sequence alignment data includes the image acquisition time sequence, the adjacent acquisition time interval, the pressure status identifier sequence, and the time sequence aligned image blocks. In practice, the image acquisition times of the image blocks in the registered sealed area are 0 seconds, 0.01 seconds, 0.02 seconds and 0.03 seconds respectively, and the adjacent acquisition time intervals are 0.01 seconds, 0.01 seconds and 0.01 seconds respectively. The image acquisition time sequence and pressure status identifier sequence are formed according to the corresponding pressure status identifier. Based on temporal alignment data from image acquisition, and using the initial distribution of suspected linear regions as the tracking entry point, local region tracking is performed on the same suspected linear region under continuous encoding pressure, generating suspected linear region tracking data. Specifically, the coordinates, contours, and extension directions of the suspected linear regions in the initial distribution are used as tracking references. Tracking markers are configured for the same suspected linear region. The initial region center position is extracted based on the coordinates and contours of the suspected linear region. The corresponding image block is extracted from the temporally aligned image block under continuous encoding pressure, and the region center position under the current pressure state is used as the tracking starting position under the next pressure state. Corresponding matching is performed on the linear gray-scale abrupt change regions, edge continuity regions, and texture break regions in the corresponding image block to obtain the region contour, region center position, and extension direction of the same suspected linear region under different pressure state markers. The suspected linear region tracking data includes the tracking marker, region contour, region center position, extension direction, corresponding image block, and pressure state marker for the same suspected linear region. In practice, the horizontal coordinate range of the suspected linear region is 260 to 410, and the vertical coordinate range is 180 to 188. Based on the coordinate range, the horizontal coordinate of the initial region center position is 335 and the vertical coordinate is 184. Under the next pressure state, the horizontal coordinate of the corresponding region center position is 337 and the vertical coordinate is 185, and this region center position is used as the tracking starting position under subsequent pressure states. Based on tracking data of suspected linear regions, image blocks of the same suspected linear region under different coding pressure states are subjected to brightness uniformity processing and pressure state identifier matching to generate pressure response aligned image data. Specifically, the image blocks of suspected linear regions in the sealed region reference image data corresponding to the same tracking identifier under the unexcited state are used as reference image blocks. The average grayscale value of the image blocks corresponding to different coding pressure states under the same tracking identifier is extracted. The average grayscale value of the reference image block is subtracted from the average grayscale value of the image block under the corresponding pressure state to obtain the brightness correction difference. The brightness correction difference is then superimposed on the pixel grayscale value of the corresponding image block. The corrected image blocks are then arranged according to the pressure state identifier and the image acquisition time to form aligned image blocks of the same suspected linear region under different coding pressure states. The pressure response aligned image data includes image blocks of the same suspected linear region under different coding pressure states, image acquisition time, and pressure state identifier. In practice, the average grayscale value of the reference image block is 118, and the average grayscale value of the image block under pressure is 126. Subtracting 126 from 118 yields a brightness correction difference of -8 grayscale levels. The -8 grayscale level is then superimposed on the pixel grayscale value of the corresponding image block to ensure that the corrected average grayscale value of the image block is consistent with the average grayscale value of the reference image block.

[0024] In this embodiment, the generation of crack response data specifically includes: Based on pressure response aligned image data, the two edges of the same suspected linear region under different coded pressure states are located to generate linear region edge localization data; the linear region edge localization data includes the coordinates of the two edges of the same suspected linear region, the direction of linear extension, the image acquisition time, and the pressure state identifier; Based on the edge localization data of the linear region, the opening displacement of the two sides of the edge is extracted under different coded pressure states to generate a crack opening displacement field. Specifically, the corresponding edge sampling positions are selected along the linear extension direction of the same suspected linear region. The distance between the two sides of the edge coordinates at the corresponding edge sampling positions under different coded pressure states is calculated. The distance between the two sides of the edge under the corresponding pressure state is subtracted from the distance between the two sides of the edge under the state without micro-pressure excitation to obtain the opening displacement of each edge sampling position. Then, the opening displacements are arranged according to the edge sampling positions and pressure state indicators to form a crack opening displacement field. The crack opening displacement field includes the edge sampling position, the distance between the two sides of the edge, the opening displacement, the image acquisition time, and the pressure state indicator. In specific implementation, under the condition of no micro-pressure excitation, the distance between the two edges at the three edge sampling positions of the same suspected linear region is 4 pixels; under the pressure loading state, the distance between the two edges at the three edge sampling positions is 7 pixels. Subtracting 4 pixels from 7 pixels respectively, the opening displacement of the three edge sampling positions is 3 pixels. The crack opening displacement field under the pressure loading state is formed according to the edge sampling position and pressure state label. Based on the coordinates of the two sides of the linear region edge localization data and the edge sampling position in the crack opening displacement field, the position changes of the two sides of the same suspected linear region under different coded pressure states are compared to generate relative edge displacement data. The relative edge displacement data includes the relative displacement of the first side edge, the relative displacement of the second side edge, the difference in displacement between the two sides edge, the image acquisition time, and the pressure state identifier. The relative displacement of the first side edge and the relative displacement of the second side edge are calculated by subtracting the corresponding edge coordinates under the corresponding pressure state from the corresponding edge coordinates under the state without micro-pressure excitation. In practice, at the same edge sampling position, the horizontal coordinate of the first side edge is 300 and the horizontal coordinate of the second side edge is 304 when there is no micro-pressure excitation; when pressure is applied, the horizontal coordinate of the first side edge is 299 and the horizontal coordinate of the second side edge is 306. Thus, the relative displacement of the first side edge is -1 pixel, the relative displacement of the second side edge is 2 pixels, and the displacement difference between the two sides is 3 pixels. Based on edge relative displacement data and pressure response aligned image data, the contour extension and endpoint changes of suspected linear regions are extracted to generate morphological extension states. The morphological extension states include the linear extension direction, contour extension length, endpoint offset position, and extension continuity state. The contour extension length is calculated by subtracting the distance between the endpoints of the suspected linear regions under the corresponding pressure state from the distance between the endpoints of the suspected linear regions under the state without micro-pressure excitation. The endpoint offset position is obtained by subtracting the coordinates of the corresponding endpoints under the corresponding pressure state from the coordinates of the endpoints of the suspected linear regions under the state without micro-pressure excitation. The extension continuity state is formed by continuously arranging the contour extension lengths of adjacent edge sampling positions according to the linear extension direction. In specific implementation, under the unexcited state, the coordinates of the first endpoint of the suspected linear region are 260, 184, the coordinates of the second endpoint are 410, 184, and the distance between the endpoints is 150 pixels; under the pressure loading state, the coordinates of the first endpoint are 258, 184, the coordinates of the second endpoint are 414, 184, and the distance between the endpoints is 156 pixels. Subtracting 150 pixels from 156 pixels yields a contour extension length of 6 pixels; the offset positions of the first endpoint are -2 pixels and 0 pixels, and the offset positions of the second endpoint are 4 pixels and 0 pixels. The contour extension length and the endpoint offset positions are continuously arranged according to the linear extension direction to form a continuous extension state. Based on the crack opening displacement field and morphological extension state, the opening state after pressure release is correlated with the initial opening state under the condition of no micro-pressure excitation, and a residual opening response is generated. The residual opening response includes the distance between the two sides after release, the initial distance between the two sides, the contour position after release, and the residual opening difference. The residual opening difference is calculated by subtracting the initial distance between the two sides under the condition of no micro-pressure excitation from the distance between the two sides after pressure release. In practice, after the pressure is released, the distance between the two edges at the same edge sampling position is 5 pixels, and the initial distance between the two edges in the state without micro-pressure excitation is 4 pixels. Subtracting 4 pixels from 5 pixels gives a residual opening difference of 1 pixel. The contour position after pressure release is matched with the initial contour position in the state without micro-pressure excitation to form a residual opening response. Based on the pressure state identifier, the crack opening displacement field, edge relative displacement data, morphological extension state, and residual opening response are collected to generate crack response data.

[0025] In this embodiment, the generation of crack candidate region data specifically includes: Based on crack response data and pseudo-random coded micro-pressure excitation sequences, the crack opening displacement changes in the crack opening displacement field are temporally correlated and state-matched with pressure loading nodes to generate loading opening matching data. Specifically, the opening displacement at the image acquisition time corresponding to the pressure loading node is extracted, and the opening displacement is subtracted from the opening displacement corresponding to the previous pressure state identifier to obtain the opening change amount corresponding to the pressure loading node. Then, the opening change amount is associated with the corresponding pressure loading node according to the same tracking identifier. The loading opening matching data includes tracking identifier, pressure loading node, opening change amount, edge sampling position, image acquisition time, and pressure state identifier. In specific implementation, the first suspected linear region is configured as tracking identifier 1, and the second suspected linear region is configured as tracking identifier 2. The opening displacement of the suspected linear region corresponding to tracking identifier 1 is 3 pixels at the time of image acquisition at the pressure loading node, and the opening displacement is 0 pixels at the previous pressure state identifier. Subtracting 0 pixels from 3 pixels gives the opening change amount corresponding to the pressure loading node as 3 pixels. Based on crack response data and pseudo-random coded micro-pressure excitation sequences, the residual opening response is matched with the pressure release node in terms of time sequence and response to generate release residual matching data. Specifically, the residual opening difference value under the pressure state identifier after the pressure release node ends is extracted, and the residual opening difference value is associated with the node end time of the pressure release node according to the same tracking identifier to form the residual opening matching relationship after pressure release. The release residual matching data includes tracking identifier, pressure release node, node end time, residual opening difference value, post-release contour position, and pressure state identifier. Based on crack response data and pseudo-random coded micro-pressure excitation sequences, the morphological extension state is continuously correlated and directionally matched with the continuously coded pressure state to generate continuous extension matching data. Specifically, the contour extension length, endpoint offset position, and extension continuity state under adjacent pressure state identifiers are extracted according to the node arrangement order, and the morphological extension state continuously arranged along the linear extension direction is correlated with the continuously coded pressure state. The continuous extension matching data includes the continuous coded pressure state, linear extension direction, contour extension length, endpoint offset position, and extension continuity state. Based on loading opening matching data, releasing residual matching data, and continuous extension matching data, suspected linear regions are synchronously screened for response and excluded from non-crack interference to generate crack candidate region data. Specifically, suspected linear regions with loading opening matching data, releasing residual matching data, and continuous extension matching data under the same tracking identifier are marked as candidate crack regions. Edge sampling positions and endpoint offset positions under the same tracking identifier are aggregated to form candidate crack positions. Edge sampling positions under the same tracking identifier are connected along the linear extension direction and combined with endpoint offset positions to form candidate crack contours. Loading opening matching data, releasing residual matching data, and continuous extension matching data are combined to generate candidate crack response identifiers. Scratches, stains, processing textures, and lighting shadows that do not simultaneously possess loading opening matching data, releasing residual matching data, and continuous extension matching data are marked as non-crack interference exclusion identifiers. The crack candidate region data includes candidate crack positions, candidate crack contours, candidate crack response identifiers, and non-crack interference exclusion identifiers.

[0026] In this embodiment, the generation of the sealing crack identification result specifically includes: Based on crack candidate region data and crack response data, interference screening, similar grouping and contour correction are performed on the candidate crack regions to generate crack location and crack contour. Interference removal includes removing suspected linear areas with non-crack interference exclusion indicators, and removing scratches, stains, processing textures, and lighting shadows that do not form candidate crack response indicators. Grouping of similar types involves merging candidate crack locations that are continuously distributed along the linear extension direction within the same candidate crack region and are not marked as non-crack interference exclusion markers. Contour correction involves adding the edge coordinates corresponding to each edge sampling position in the candidate crack contour to the relative displacement of the edge under the same pressure state identifier to obtain the corrected edge coordinates, and connecting the corrected edge coordinates according to the arrangement order of the edge sampling positions along the linear extension direction to form the crack contour. Based on the morphological extension state in the crack response data, the extension direction and endpoint change direction of the crack profile are aggregated to generate the crack extension direction. The profile extension direction is formed by the coordinate arrangement direction of the sampling positions of adjacent edges in the crack profile, and the endpoint change direction is formed by the movement direction of the endpoint offset position after pressure is applied relative to the endpoint position under the state without micro-pressure excitation. When the endpoint change direction is consistent with the profile extension direction, the profile extension direction is taken as the crack extension direction. When the endpoint change direction is inconsistent with the profile extension direction, the profile extension direction is corrected according to the movement direction corresponding to the endpoint offset position to form the crack extension direction. Based on the crack opening displacement field and edge relative displacement in the crack response data, the opening changes of the same candidate crack region under different coded pressure states are collected to generate crack opening states. The crack opening state is formed by the opening displacement and relative edge displacement under different coded pressure states, including the pressure loading opening increase state, the pressure holding opening maintenance state, the pressure release opening retraction state, and the pressure interval opening stability state. Based on the residual opening response in the crack response data, the residual opening state after pressure release and the initial opening state under the condition of no micro-pressure excitation are residually aggregated to generate the crack residual response state. The crack location, crack profile, crack extension direction, crack opening state and crack residual response state are integrated to generate the sealing crack identification result. The crack residual response state is jointly formed by the residual opening difference after pressure release, the profile position after release and the initial opening state under the condition of no micro-pressure excitation.

[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to the scenario of identifying factory-sealed cracks in a sealing cover of an automotive battery pack. The sealing cover is formed by pressing an aluminum alloy shell with a rubber sealing ring. The sealing edge is 320 mm long and 18 mm wide, and its surface exhibits processing textures, a slight oil film, pressing marks, and localized reflections. Because some cracks have small openings when not under stress, their appearance is similar to scratches, texture lines, and oil film reflections. Manual visual inspection and ordinary static image recognition are prone to missing shallow cracks or misjudging non-crack linear traces.

[0028] During implementation, the image sensor is fixed 180 mm above the sealing edge, with a resolution of 2448×2048, a frame rate of 100 frames per second, an exposure time of 2 milliseconds, and a constant imaging distance. The micro-pressure excitation unit is arranged inside the sealing cover, ensuring its operating area corresponds to the sealing detection area within the image sensor's field of view. The sealing area to be detected is selected within an image range of 120 to 720 horizontal coordinates and 80 to 400 vertical coordinates. A reference image is acquired before applying micro-pressure excitation, and corrections are made for grayscale fluctuations, noise points, and brightness inconsistencies caused by surface reflection within this area to form a stable reference image of the sealing area.

[0029] In the actual identification process, the micro-pressure excitation conditions were set to a pressure amplitude of 2 kPa, a pressure duration of 0.3 seconds, 6 pressure loading cycles, and a pressure release interval of 0.2 seconds, forming coded pressure states in a non-equidistant sequence. The image sensor simultaneously exposed and acquired images during the micro-pressure excitation process, obtaining images of the sealing area under pressure interval, pressure loading, pressure holding, and pressure release states. For suspected linear regions with horizontal coordinates of 260 to 410 and vertical coordinates of 180 to 188, the system tracked the corresponding changes in the positions and contours of their two edges and endpoints under different coded pressure states. Under pressure loading, this suspected linear region showed an increase in opening and relative edge displacement, retaining a residual opening response after pressure release. Ordinary scratches, oil film reflections, and processing textures did not show opening changes synchronized with the coded pressure states. Therefore, regions with synchronized response characteristics were retained as candidate crack regions, and their location, contour, extension direction, opening state, and residual response state were integrated to generate the sealing crack identification result.

[0030] Table 1. Performance Comparison of the Invention and Traditional Sealing Crack Identification Methods

[0031] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0032] The crack recognition accuracy of the method of this invention is 93.2%, while that of the traditional static image recognition method is 88.5%, representing an improvement of 4.7 percentage points. This improvement mainly stems from the fact that this invention does not rely solely on grayscale abrupt changes or edge morphology in a single image to identify cracks, but rather combines the crack opening displacement field, relative edge displacement, and residual opening response under coded pressure conditions for identification, making the difference between real cracks and ordinary scratches and texture lines more obvious.

[0033] Regarding the detection rate of shallow cracks, the method of this invention achieves 90.1%, while the traditional static image recognition method achieves 84.6%, representing an improvement of 5.5 percentage points. Shallow cracks have small openings when not subjected to micro-pressure excitation, and traditional methods easily ignore them as ordinary surface textures. This invention, through micro-pressure excitation, causes collectable changes in the opening of shallow cracks during pressure loading, pressure holding, and pressure release, thus improving the detection capability of shallow cracks.

[0034] Regarding the non-crack interference exclusion rate, the method of this invention achieves 91.4%, while the traditional static image recognition method achieves 86.8%, representing an improvement of 4.6 percentage points. This improvement stems from the invention's simultaneous response screening of scratches, stains, processing textures, and lighting shadows. Linear areas that do not form candidate crack response markers are excluded, reducing misjudgments caused by similar linear features in static images.

[0035] Regarding the accuracy of dynamic response discrimination, the method of this invention achieves 92.0%, while the traditional static image recognition method achieves 87.2%, representing an improvement of 4.8 percentage points. This is because the present invention matches crack response data with a pseudo-random coded micro-pressure excitation sequence, ensuring that crack opening changes, residual openings, and morphological extension all correspond to pressure nodes, thereby enhancing the temporal consistency of the recognition process.

[0036] In terms of recognition stability, the method of this invention achieves 93.5%, while the traditional static image recognition method achieves 89.1%, representing an improvement of 4.4 percentage points. This improvement mainly stems from the fact that this invention performs brightness correction, noise suppression, and response alignment on the sealed area image before recognition, reducing the impact of supplementary lighting fluctuations, surface reflections, and slight positional shifts on crack detection.

[0037] Regarding the false negative rate, the method of this invention has a false negative rate of 5.6%, while the traditional static image recognition method has a false negative rate of 8.9%, a decrease of 3.3 percentage points. This decrease indicates that the present invention uses the crack opening displacement field, relative edge displacement, morphological extension state, and residual opening response to jointly determine the crack region, thus exhibiting a more stable recognition capability for low-contrast, shallow-opening cracks.

[0038] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for identifying sealing cracks based on an image sensor, characterized in that, Includes the following steps: The system acquires information on the sealing area of ​​the sealing structure to be tested, image acquisition conditions, and micro-pressure excitation conditions. It also spatially calibrates the field of view of the image sensor and the effective area of ​​the micro-pressure excitation unit to generate sealing crack identification configuration data. Based on the configuration data for identifying sealing cracks, the image of the sealing area that is not subjected to micro-pressure excitation is benchmarked and the linear area is initially screened to generate benchmark image data of the sealing area. Based on the configuration data for sealing crack identification, a pseudo-random coded micro-pressure excitation sequence is generated and applied to the sealing structure to be detected. During the micro-pressure excitation process, an image sensor is simultaneously triggered to perform exposure acquisition and generate a coded response image sequence. Based on the reference image data of the sealed area and the encoded response image sequence, the images of the same sealed detection area under different encoded pressure states are aligned to generate pressure response aligned image data. Based on pressure response aligned image data, crack opening displacement field, relative edge displacement, morphological extension state, and residual opening response of suspected linear regions during the coded pressure change process are extracted to generate crack response data. Based on crack response data and pseudo-random coded micro-pressure excitation sequences, synchronous response screening and non-crack interference elimination are performed on suspected linear regions to generate crack candidate region data. Based on the crack candidate region data and crack response data, the results of the candidate crack regions are integrated to generate the sealing crack identification results.

2. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The sealing area information includes the sealing structure type, sealing detection area, sealing contact boundary, and the surface range to be detected. The image acquisition conditions include the image sensor identifier, acquisition resolution, exposure time, acquisition frame rate, and imaging distance. The micro-pressure excitation conditions include pressure amplitude, pressure duration, number of pressure loading cycles, and pressure release interval. The spatial calibration includes coordinate mapping between the field of view of the image sensor and the effective area of ​​the micro-pressure excitation unit, area boundary matching, and recording of calibration results.

3. The sealing crack identification method based on an image sensor according to claim 1, characterized in that, The generation of the reference image data for the sealed area specifically includes: Based on the configuration data for sealing crack identification, the sealing detection area is extracted from the image of the sealing area that has never been subjected to micro-pressure excitation and its coordinates are marked to generate an image of the sealing detection area. Based on the image of the sealed detection area, image brightness correction and noise suppression are performed on the image brightness distribution and noise distribution to generate a brightness-stable sealed area image; Based on the brightness-stable sealed area image, texture smoothing and contour standardization are performed on the background texture, sealing contact boundary and boundary contour to generate the sealing area reference feature data. Based on the reference feature data of the sealing area, initial screening is performed on linear gray-level abrupt changes, linear edge continuity and texture breaks in the sealing detection area to generate reference image data of the sealing area. The reference image data of the sealing area includes reference gray-level distribution, reference texture distribution, reference boundary contour and initial distribution of suspected linear areas.

4. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The generation of the encoded response image sequence specifically includes: Based on the micro-pressure excitation conditions in the sealing crack identification configuration data, the micro-pressure excitation process of the sealing structure to be detected is divided into pressure action processes, and basic data of pressure nodes are generated. Different pressure nodes in the basic data of pressure nodes are encoded and arranged in a non-equal interval order, and the micro-pressure excitation conditions of each pressure node are associated with the corresponding encoding position to generate a pseudo-random encoded micro-pressure excitation sequence. Based on the pseudo-random coded micro-pressure excitation sequence, the micro-pressure excitation unit is driven to apply micro-pressure excitation to the sealing structure to be tested according to the node arrangement order, generating coded pressure execution data. The coded pressure execution data includes pressure node type, node start time, node end time, pressure amplitude, pressure duration, pressure release interval, and pressure status identifier. Based on the coded pressure execution data and image acquisition conditions, the execution time of each pressure node is synchronously bound to the exposure time of the image sensor to generate pressure node synchronization trigger data. The pressure node synchronization trigger data includes pressure node type, node execution time, exposure time, image acquisition frame number, pressure status identifier, and synchronization binding relationship. Based on the pressure node synchronous trigger data, images of the sealing area are synchronously acquired under different coded pressure states, and the sequence is organized according to the acquisition time and pressure state identifier to generate a coded response image sequence. The coded response image sequence includes images of the sealing area under different coded pressure states, image acquisition time, image acquisition frame number, pressure state identifier, and pressure node type.

5. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The generation of the pressure response aligned image data specifically includes: Based on the reference image data of the sealed area, the corresponding image blocks of the same sealed detection area under different coding pressure states are located from the encoded response image sequence, and the corresponding image data of the area is generated. Based on the region-corresponding image data, boundary coordinates are compared and region positions are registered for image blocks of the sealed region under different coded pressure conditions to generate sealed region position registration data. Based on the registration data of the sealed area position, the image blocks of the registered sealed area are arranged in the acquisition time sequence according to the image acquisition time and pressure status identifier to generate image acquisition time sequence alignment data. Based on the image acquisition time-series aligned data, taking the initial distribution of suspected linear regions as the tracking entry point, local region tracking is performed on the same suspected linear region under continuous encoding pressure to generate suspected linear region tracking data. Based on suspected linear region tracking data, image blocks of the same suspected linear region under different coding pressure states are subjected to brightness uniformity processing and pressure state label matching to generate pressure response aligned image data.

6. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The generation of the crack response data specifically includes: Based on pressure response aligned image data, the two edges of the same suspected linear region under different coded pressure states are located to generate linear region edge localization data. Based on the edge localization data of the linear region, the opening displacement of the two sides of the edge under different coded pressure states is extracted to generate the crack opening displacement field. Based on the coordinates of the two sides of the linear region edge localization data and the edge sampling position in the crack opening displacement field, the position changes of the two sides of the same suspected linear region under different coded pressure states are compared to generate relative edge displacement data. Based on edge relative displacement data and pressure response aligned image data, the contour extension and endpoint changes of suspected linear regions are extracted to generate morphological extension states. Based on the crack opening displacement field and morphological extension state, the opening state after pressure release is correlated with the initial opening state under the condition of no micro-pressure excitation, and the residual opening response is generated. Based on the pressure state identifier, the crack opening displacement field, edge relative displacement data, morphological extension state, and residual opening response are collected to generate crack response data.

7. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The generation of the crack candidate region data specifically includes: Based on crack response data and pseudo-random encoded micro-pressure excitation sequence, the crack opening displacement change in the crack opening displacement field is matched with the pressure loading node in time sequence and state to generate loading opening matching data. Based on crack response data and pseudo-random encoded micro-pressure excitation sequence, the residual opening response is matched with the pressure release node in time sequence and response to generate release residual matching data. Based on crack response data and pseudo-random coded micro-pressure excitation sequence, the morphological extension state is continuously corresponded and matched with the continuously coded pressure state to generate continuous extension matching data. Based on the loading opening matching data, releasing residual matching data, and continuous extension matching data, suspected linear regions are synchronously screened and non-crack interference is eliminated to generate crack candidate region data. The crack candidate region data includes candidate crack location, candidate crack outline, candidate crack response identifier, and non-crack interference elimination identifier.

8. The method for identifying sealing cracks based on an image sensor according to claim 1, characterized in that, The generation of the sealing crack identification result specifically includes: Based on crack candidate region data and crack response data, interference screening, similar grouping and contour correction are performed on the candidate crack regions to generate crack location and crack contour. Based on the morphological extension state in the crack response data, the extension direction and endpoint change direction of the crack profile are aggregated to generate the crack extension direction. Based on the crack opening displacement field and edge relative displacement in the crack response data, the opening changes of the same candidate crack region under different coded pressure states are collected to generate crack opening states. Based on the residual opening response in the crack response data, the residual opening state after pressure release and the initial opening state under the condition of no micro-pressure excitation are residually collected to generate the crack residual response state. The crack location, crack profile, crack extension direction, crack opening state and crack residual response state are integrated to generate the sealing crack identification result.

Images