A method and apparatus for detecting cracks in large structural components

Abstract

This invention discloses a method and apparatus for detecting cracks in large components, specifically relating to the field of crack detection technology. It involves planning a round-trip scanning path based on the component size and the resolution of the detection camera, simultaneously acquiring the fatigue loading signal phase, and collecting multi-phase reference and detection images. When a load phase difference exists in the images, a phase correction reference image is generated. Abnormal regions are extracted through grayscale correlation analysis, distinguishing between crack opening / closing regions and crack propagation regions, thereby accurately determining the crack profile, length, and crack propagation rate, thus improving the accuracy and reliability of crack detection.

Description

Technical Field

[0001] This invention relates to the field of crack detection technology, and more specifically, to a method and apparatus for crack detection in large components. Background Technology

[0002] During fatigue and damage tolerance tests of large components, continuous detection of cracks on the component surface is usually required. In existing technologies, a detection camera is typically used to scan the surface of the large component back and forth along a preset path. By comparing the detection image at the same shooting point with a reference image, it is determined whether any crack anomalies exist in the corresponding area.

[0003] However, during the cyclic loading of large components, cracks are affected by changes in the load phase, exhibiting different opening and closing states. Existing technologies typically use only the first image as a fixed reference and compare subsequent images based on the captured images. This lack of identification and constraint of load phase changes easily leads to misjudging reversible apparent changes caused by crack opening and closing as crack anomalies or crack propagation, thus affecting the reliability of crack detection results for large components. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and apparatus for detecting cracks in large components to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting cracks in large structural components includes the following steps: S1: Plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and acquire the load phase of the fatigue loading signal; S2: During the first round of operation of the detection camera, multiple reference images under the load phase are collected at each shooting point to form a reference image group; S3: Acquire detection images during cyclic operation, and form a detection image group according to the shooting point, running direction and load phase; S4: When there is a load phase difference between the detected image and the reference image, select an adjacent phase reference image at the same shooting point, register the adjacent phase reference image and perform grayscale interpolation to generate a phase correction reference image; S5: Compare the correlation between the detected image and the phase correction reference image, extract the abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions; S6: Based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle, the crack profile, crack length and crack propagation rate are determined.

[0006] In a preferred embodiment, S1 specifically refers to: Determine the boundary of the detection area based on the size of large components; The actual size of a single pixel is determined based on the detection camera resolution and detection accuracy, and the coverage area of ​​a single frame image is calculated according to the actual size of the single pixel. A sequence of shooting points is generated based on the detection area boundary and the coverage of a single frame image, and a round-trip scanning path is generated based on the sequence of shooting points. The fatigue loading signal is acquired synchronously, and the load phase is divided according to the periodic change of the fatigue loading signal.

[0007] In a preferred embodiment, S2 specifically refers to: The detection camera is controlled to run in the first round along the reciprocating scanning path. When the detection camera reaches the shooting point in the shooting point sequence, the load phase obtained by dividing the periodic change of the fatigue loading signal triggers image acquisition. Use images corresponding to different load phases at the same shooting point as reference images; The reference images are associated with the shooting location, direction of travel, and load phase to form a reference image group.

[0008] In a preferred embodiment, S3 specifically refers to: After the reference image group is formed, the detection camera is controlled to run cyclically along the round-trip scanning path; When the camera reaches a shooting point in the shooting point sequence, the load phase acquisition and detection image is obtained based on the periodic change of the fatigue loading signal. The detected images are associated with the shooting location, running direction, load phase, and running cycle, and are collected in the order of running cycle and shooting location to form a detection image group.

[0009] In a preferred embodiment, S4 specifically refers to: Read the shooting point, running direction and load phase associated with the detection image from the detection image group, and find the reference image with the same shooting point and running direction in the reference image group; The load phase associated with the detected image is compared with the load phase associated with the reference image. If the load phases are inconsistent, the reference image corresponding to the load phases adjacent to the load phase associated with the detected image is selected as the adjacent phase reference image. Image registration is performed on adjacent phase reference images, and grayscale interpolation is performed based on the position of the load phase associated with the detection image between the preceding and following adjacent load phases to generate a phase correction reference image.

[0010] In a preferred embodiment, image registration is performed on adjacent phase reference images, and grayscale interpolation is performed based on the position of the load phase associated with the detection image between adjacent load phases to generate a phase correction reference image, specifically: The adjacent phase reference image whose load phase is earlier than the load phase associated with the detection image is used as the previous phase reference image; The adjacent phase reference image whose load phase is later than the load phase associated with the detection image is used as the subsequent phase reference image; Register the rear phase reference image to the image coordinates of the front phase reference image; Based on the position of the load phase associated with the detected image between the load phase associated with the front phase reference image and the load phase associated with the rear phase reference image, grayscale interpolation is performed on the front phase reference image and the rear phase reference image to generate a phase correction reference image.

[0011] In a preferred embodiment, S5 specifically refers to: The detected image and the phase correction reference image are divided into image sub-regions corresponding to the same shooting point according to the image coordinates of the same shooting point, and the gray-level correlation between the image sub-regions corresponding to the same position is calculated. Image sub-regions with gray-level correlation below the abnormal threshold and adjacent to each other are merged to obtain abnormal connected regions. Extract the edge pixels on both sides of the abnormal connected region, and determine the edge normal of the abnormal connected region based on the local tangent direction of the edge pixels on both sides; Match relative edge pixels along the edge normal of the abnormal connected region, form an opening spacing sequence based on the distance between the relative edge pixels, and form an endpoint coordinate sequence based on the edge pixel coordinates at both ends of the abnormal connected region.

[0012] In a preferred embodiment, S6 specifically refers to: The aperture spacing sequence at the same shooting point and in the same running cycle is sorted according to the load phase to obtain the change of aperture spacing sequence with the load phase. The endpoint coordinate sequence under the same shooting point and the same load phase is sorted according to the running round to obtain the continuous change of the endpoint coordinate sequence with the running round; An abnormally connected region with reciprocating changes but without a continuously changing sequence of endpoint coordinates is designated as the crack opening and closing region, while an abnormally connected region with a continuously changing sequence of endpoint coordinates is designated as the crack propagation region. The crack profile is determined based on the crack propagation area, and the crack length and crack propagation rate are determined by combining the actual size corresponding to a single pixel.

[0013] In a preferred embodiment, the determination of the crack initiation / closing region and the crack propagation region specifically involves: By sorting the sequence of opening spacing at the same shooting point, the same running cycle, and different load phases, the direction of change of opening spacing can be obtained. By sorting the endpoint coordinate sequences under the same shooting point, the same load phase, and adjacent operating cycles, the endpoint displacement direction can be obtained. When the direction of the opening spacing changes alternately with the load phase and the direction of the endpoint displacement does not continue along the running cycle, the abnormal connected region is classified as the crack opening and closing region. When the endpoint displacement direction continues along the running cycle, the abnormally connected region is included in the crack propagation region.

[0014] On the other hand, the present invention provides a large component crack detection device, comprising: The preprocessing module is used to plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and to acquire the load phase of the fatigue loading signal; The reference image acquisition module is used to detect reference images under multiple load phases acquired at each shooting point during the first round of camera operation, forming a reference image group; The detection image acquisition module is used to acquire detection images during cyclic operation and form detection image groups according to the shooting point, running direction and load phase. The phase correction module is used to select adjacent phase reference images at the same shooting point when there is a load phase difference between the detected image and the reference image, register the adjacent phase reference images and perform grayscale interpolation to generate a phase correction reference image. The crack detection module is used to compare the correlation between the detected image and the phase correction reference image, extract abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions. The crack assessment module is used to determine the crack profile, crack length, and crack propagation rate based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle.

[0015] The technical effects and advantages of the present invention regarding a method and device for detecting cracks in large structural components are as follows: By planning the reciprocating scanning path based on the size of large components and the resolution of the inspection camera, the inspection camera can systematically cover the surface of large components according to the shooting points. The load phase of fatigue loading signals is acquired synchronously, and a set of reference images containing different load phases is formed in the first run, providing a phase reference for image comparison. In the cyclic run, a set of inspection images containing shooting points, running direction, and load phase is formed, establishing a clear correspondence between the inspection images and the reference images. When load phase differences exist, adjacent phase reference images are selected and a phase-corrected reference image is generated to reduce the impact of load phase differences on image comparison. Abnormal connected regions are extracted through correlation comparison, and the opening spacing sequence and endpoint coordinate sequence are calculated, allowing for separate characterization of crack opening and closing changes and crack endpoint changes. Combining the reciprocating changes of the opening spacing sequence and the continuous changes of the endpoint coordinate sequence determines the crack profile, crack length, and crack propagation rate, improving the accuracy, continuity, and comparability of crack detection results for large components. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a method for detecting cracks in large components according to the present invention; Figure 2 This is a schematic diagram of the camera layout for detecting large curved surface components according to the present invention; Figure 3 This is a schematic diagram illustrating the field-of-view coverage relationship of the camera for detecting arc components according to the present invention; Figure 4 This is a structural block diagram of a large component crack detection device according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1

[0019] Figure 1 This invention provides a method for detecting cracks in large structural components, comprising the following steps: S1: Plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and acquire the load phase of the fatigue loading signal; S2: During the first round of operation of the detection camera, multiple reference images under the load phase are collected at each shooting point to form a reference image group; S3: Acquire detection images during cyclic operation, and form a detection image group according to the shooting point, running direction and load phase; S4: When there is a load phase difference between the detected image and the reference image, select an adjacent phase reference image at the same shooting point, register the adjacent phase reference image and perform grayscale interpolation to generate a phase correction reference image; S5: Compare the correlation between the detected image and the phase correction reference image, extract the abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions; S6: Based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle, the crack profile, crack length and crack propagation rate are determined.

[0020] S1: Plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and acquire the load phase of the fatigue loading signal, including: First, the dimensions of the large component are determined to define the geometric dimensional range of the surface of the large component to be measured, such as its length, width, height, and diameter. Then, a 3D laser scanner or laser rangefinder is used to measure the actual dimensions of the surface of the large component to be measured, thus generating the component's dimensional data.

[0021] Based on the obtained dimensional data of large components, the boundary of the inspection area on the surface of the large components is determined. Specifically, according to the dimensional data of the large components, the area to be inspected for cracks is delineated on the surface of the component to be tested. The boundary of the inspection area can be the entire surface of the component, or it can focus on covering special areas that are greatly affected by fatigue loads, such as weld areas, areas around holes, or stress concentration areas.

[0022] Based on the effective pixel count of the detection camera and the preset crack detection accuracy requirements, the actual size corresponding to a single pixel is determined. Specifically, crack detection accuracy typically depends on the minimum length requirement of the crack to be detected. For example, if the required crack length is 0.5 mm, then the actual size corresponding to a single pixel must be less than or equal to half of the required crack length, i.e., 0.25 mm, to ensure that at least two pixels correspond to the minimum crack length requirement. Simultaneously, the actual size corresponding to a single pixel is also constrained by the installation distance of the detection camera, i.e., calculated based on the installation distance and then divided by the effective pixel count of the detection camera.

[0023] Based on the determined actual size of each pixel, the actual coverage area of ​​a single frame image is calculated. Specifically, the effective pixel count of the detection camera is multiplied by the actual size of each pixel. For example, if the effective pixel count of the detection camera is 4000 pixels × 3000 pixels, and the actual size of each pixel is 0.25 mm, then the calculated actual coverage area of ​​a single frame image is 1 meter horizontally (4000 pixels × 0.25 mm / pixel) and 0.75 meters vertically (3000 pixels × 0.25 mm / pixel).

[0024] Based on the actual coverage area of ​​the obtained single-frame image, a sequence of shooting points is generated according to the boundary of the detection area. Specifically, the determination of the shooting point sequence should ensure that there is a certain overlap area between adjacent single-frame images to facilitate image registration accuracy and crack continuity detection. The overlap area ratio is determined, for example, through pre-experimentation. That is, images are acquired under different overlap ratios, and the crack contour matching accuracy is evaluated. When the crack contour registration error at the image edge is minimized, the corresponding overlap area ratio is the suitable overlap area ratio. For example, the overlap area ratio is determined to be 15% through pre-experimentation.

[0025] Based on the generated sequence of shooting points, the round-trip scanning path of the detection camera is planned. Specifically, the sequence of shooting points is arranged sequentially according to the spatial order of the component surface to form a continuous scanning path, for example, arranged row by row from one end of the component to the other, ensuring that the detection camera can completely cover the entire area defined by the boundary of the detection region in each run. For arc components or complex curved surface components, the scanning path is planned as a curve based on the surface topology of the arc component or complex curved surface component to ensure complete coverage without omissions.

[0026] like Figure 2 As shown, when the large component is an arc-shaped component or a semi-enclosed curved surface component, the detection camera is arranged with corresponding shooting points along the length and arc direction of the large component. The shooting direction of the detection camera is towards the surface of the large component to be measured, and the sequence of shooting points is arranged according to the spatial unfolding order of the surface of the large component to be measured. The detection camera sequentially reaches the shooting points in the shooting point sequence on the reciprocating scanning path, and acquires an image covering the corresponding surface area at each shooting point, so that the reciprocating scanning path is adapted to the curved shape of the surface of the large component to be measured.

[0027] like Figure 3 As shown, when the large component is an arc-shaped component, the field of view coverage angle of the detection camera in the cross-sectional direction of the arc-shaped component is determined based on the cross-sectional radius of the arc-shaped component, the shooting distance between the detection camera and the surface to be measured of the arc-shaped component, and the actual coverage area of ​​a single frame image. The field of view coverage angle is used to convert the actual coverage area of ​​a single frame image into the coverage area in the cross-sectional direction of the arc-shaped component, and accordingly determine the number of shooting points in the cross-sectional direction of the arc-shaped component. Overlapping coverage areas are retained between adjacent shooting points, forming a continuous coverage relationship between adjacent single frames. The number of shooting points in the cross-sectional direction of the arc-shaped component and the number of shooting points in the length direction of the arc-shaped component together form a shooting point sequence, which is used to generate the round-trip scanning path of the detection camera.

[0028] Simultaneously, during the fatigue loading process of large components, fatigue loading signals are collected in real time. Specifically, load signal data is collected in real time through load sensors, such as strain gauge load sensors, force sensors, or piezoelectric sensors, to record load signal data in real time and form continuous load signal data.

[0029] The method for dividing the load phase is as follows: Numerical analysis is performed on the real-time load signal data, such as using Fourier transform analysis to obtain the periodic variation law and load amplitude of the load signal. Based on the periodic characteristics of the load signal, such as the maximum and minimum load values, the slope of change, or the inflection point, each complete load cycle is divided into several load phases, including, for example, the initial loading phase, the intermediate loading phase, the maximum load loading phase, the unloading phase, and the minimum load phase. Threshold parameters for dividing the load phases can be determined experimentally, for example, by identifying the point where the load change slope abruptly changes as the load phase boundary.

[0030] To ensure synchronization between the image acquired by the detection camera and the load phase, a load phase triggering mechanism is adopted. Specifically, the fatigue loading signal is continuously sampled to obtain a continuous load value sequence; a complete load cycle is identified based on the continuous load value sequence, and the maximum and minimum load values ​​corresponding to the complete load cycle are determined; the load phase is divided according to the load change interval between the maximum and minimum load values, and a load threshold interval is set for each load phase based on the load change interval; the real-time load value corresponding to the load signal is acquired in real time, and it is determined whether the real-time load value enters the load threshold interval associated with the corresponding load phase; when the real-time load value enters the load threshold interval associated with the corresponding load phase, the detection camera is triggered to acquire an image, and the acquired image is associated with the corresponding load phase; the load threshold interval is determined based on the load change interval between the maximum and minimum load values, for example, by dividing the load change interval proportionally or by a preset proportional division method to obtain multiple continuous load intervals, and each of the multiple continuous load intervals corresponds to a different load phase.

[0031] S2: During the first round of operation, the detection camera acquires reference images under multiple load phases at each shooting point, forming a reference image group, including: The inspection camera sequentially passes through each shooting point according to the round-trip scanning path. The spatial location of the shooting points is a sequence generated by combining the boundary of the inspection area determined by the size of the large component and the actual coverage area of ​​a single frame image determined by the resolution of the inspection camera. The process of the inspection camera moving to each shooting point is automatically controlled by a high-precision linear guide or robotic arm, such as by stepper motor drive or servo motor drive, to ensure that the inspection camera can reach each predetermined shooting point.

[0032] When the detection camera reaches a certain shooting point in the shooting point sequence, the load phase corresponding to the current real-time load value is determined in real time according to the load phase triggering mechanism. When the real-time load value enters the load threshold range corresponding to a specific load phase, the detection camera is immediately triggered to perform image acquisition operation to ensure that the image acquired by the detection camera accurately corresponds to the specific load phase of the real-time load value.

[0033] Each time the camera captures an image, the resulting image data becomes a reference image. Each reference image includes the image's own grayscale or color pixel information, and the image resolution is consistent with the effective pixel count of the camera, for example, 4000 pixels × 3000 pixels. The image data is stored in a digital image format (such as BMP or JPEG format).

[0034] The coordinates of the shooting points are represented using two-dimensional or three-dimensional coordinates. For example, in two-dimensional coordinates, the origin is a fixed boundary point on the component surface, with the length direction as the X-axis and the width direction as the Y-axis; in three-dimensional coordinates, the origin is a fixed point on the component surface, with the length direction as the X-axis, the width direction as the Y-axis, and the height direction as the Z-axis. The camera's running direction information is recorded, indicating the direction of the scanning path, such as from left to right or from right to left. Load phase information is recorded using a marking method, such as marking it as the initial loading phase, the intermediate loading phase, the maximum loading phase, the unloading phase, and the minimum loading phase.

[0035] After each detection camera completes the acquisition of multiple load phase images at a shooting point, it gathers the reference images corresponding to different load phases at the same shooting point, the shooting point coordinate information, the running direction information, and the load phase information to obtain a reference image group.

[0036] S3: Acquire detection images during cyclic operation, and form detection image groups according to the shooting point, running direction, and load phase, including: Based on the round-trip scanning path and the sequence of shooting points, the inspection camera is controlled to repeatedly perform multiple round-trip scans. Each round-trip scan follows the shooting point sequence sequentially, ensuring that the inspection camera reaches each shooting point in the sequence in turn. The spatial coordinates of the shooting point sequence are recorded, for example, in two-dimensional or three-dimensional coordinates. The two-dimensional coordinate system, for example, uses a fixed boundary point on the component surface as the origin, with the length direction as the X-axis and the width direction as the Y-axis; the three-dimensional coordinate system, for example, uses a fixed point on the component surface as the origin, with the length direction as the X-axis, the width direction as the Y-axis, and the height direction as the Z-axis. The movement of the inspection camera is achieved through a high-precision linear guide or robotic arm, driven by, for example, a servo motor or stepper motor, ensuring that the positioning accuracy of the inspection camera reaches the millimeter or even sub-millimeter level.

[0037] When the detection camera accurately reaches the shooting point during each cycle, the load phase corresponding to the real-time load value is determined in real time according to the load phase triggering mechanism. Continuous load signal data sequences are acquired in real time, and the load threshold range for each load phase is determined. When the real-time load value first enters the load threshold range corresponding to a certain load phase, a trigger signal is automatically generated and transmitted in real time to the detection camera controller. Upon receiving the trigger signal, the controller immediately starts image acquisition to ensure that each acquired detection image corresponds to the load phase of the real-time load value, avoiding crack opening and closing errors caused by load changes.

[0038] After each image acquisition is completed, the acquired image is defined as a detection image. The data content of each detection image includes the grayscale or color pixel matrix of the image. The spatial resolution of the image is consistent with the effective pixel count of the detection camera, such as 4000 pixels × 3000 pixels. The storage format is a common digital image file format, such as BMP or JPEG.

[0039] After each frame of the detection image is acquired, the coordinates of the shooting point corresponding to the current detection image, the camera's running direction, the load phase, and the cycle number are recorded. The shooting point coordinates are recorded in a coordinate system; the camera's running direction is marked with the direction of each scan path, such as from left to right or from right to left; the load phase is marked with text, such as initial load phase, intermediate load phase, maximum load phase, unload phase, and minimum load phase; the cycle number is marked with the number of scans performed by the detection camera, such as the first cycle, the second cycle, etc., to distinguish the detection images from different cycles.

[0040] All detected images are collected in an ordered manner according to the order of the running rounds and the shooting locations. The first level of sorting is based on the running round as the primary sorting index, and the second level of sorting is based on the shooting location coordinates within each running round. For example, all detected images from the first running round are sorted and collected first, and then all detected images from the second running round are collected.

[0041] S4: When there is a load phase difference between the detected image and the reference image, select adjacent phase reference images at the same shooting point, register the adjacent phase reference images and perform grayscale interpolation to generate a phase-corrected reference image, including: The system reads the associated image coordinates, camera movement direction, and load phase information from each image in the detection image set. The image coordinates are represented in a two-dimensional or three-dimensional coordinate system. For example, in two-dimensional coordinates, the origin is a fixed boundary point on the component surface, with the length direction of the component as the X-axis and the width direction as the Y-axis. In three-dimensional coordinates, the origin is a fixed point on the component surface, with the length direction as the X-axis, the width direction as the Y-axis, and the height direction as the Z-axis. The camera movement direction information indicates the actual movement direction of each scan path, such as from left to right or from right to left. The load phase information includes clearly marked initial loading phase, intermediate loading phase, maximum load phase, unloading phase, and minimum load phase.

[0042] Based on the information read from the detected image, a reference image with the exact same shooting point coordinates and camera movement direction information as the current detected image is found from the reference image group.

[0043] The load phase associated with the currently processed detection image is compared one by one with the load phase associated with reference images in the same shooting point coordinates and the same running direction within the reference image group. When the load phase of the detection image is completely consistent with the load phase of a certain reference image, the reference image is selected as the reference image for crack identification and comparison analysis. When the load phase associated with the detection image is not completely consistent with the load phase of the reference image, phase correction reference image generation processing is performed: within the reference image group, two adjacent phase reference images that are adjacent to the load phase associated with the detection image in terms of load change cycle are selected. Among them, the adjacent phase reference image whose load phase is earlier than the load phase associated with the detection image in terms of load change cycle is defined as the previous phase reference image, and the adjacent phase reference image whose load phase is later than the load phase associated with the detection image is defined as the subsequent phase reference image. The selection of the above two adjacent phase reference images depends on the time series relationship of the load phase to ensure that the load phase of the preceding and subsequent phase reference images sandwich the load phase associated with the detection image.

[0044] After selecting the preceding and following phase reference images, image registration is performed. Feature point sets are extracted from both the preceding and following phase reference images, for example, using a scale-invariant feature transform algorithm to extract salient and robust feature points. Using the matched feature point pairs, a spatial transformation model between the images is calculated. An affine transformation model or a perspective transformation model is selected as the mathematical model for the spatial transformation, and the transformation model parameters are solved. The following step is to register the following phase reference image onto the image coordinate system of the preceding phase reference image, ensuring that the same physical location in two adjacent phase reference images corresponds to the same image coordinate location.

[0045] After image registration is completed, grayscale interpolation is performed based on the position of the load phase associated with the detected image between the load phases of the preceding and following phase reference images to generate a phase correction reference image. The relative position between the load phase associated with the detected image and the load phases of the preceding and following phase reference images is calculated and defined as the load phase interpolation coefficient. The calculation formula is as follows: ; in, This represents the load phase interpolation coefficient, with a value ranging from 0 to 1; To detect the real-time load value corresponding to the phase of the image-associated load; The real-time load value corresponding to the load phase is associated with the previous phase reference image; This refers to the real-time load value corresponding to the load phase associated with the back-phase reference image. All load values ​​are derived from real-time acquired fatigue loading signal data, with units such as Newtons (N).

[0046] Using the load phase interpolation coefficients, pixel-by-pixel grayscale interpolation is performed to generate the final phase-corrected reference image. The formula for pixel-by-pixel grayscale interpolation is: ; in, The gray value at pixel coordinates (x, y) in the phase-corrected reference image to be generated; This represents the grayscale value at coordinates (x, y) in the previous phase reference image; This refers to the grayscale value at coordinates (x, y) in the registered post-phase reference image.

[0047] The above formula is used to calculate pixel by pixel until the grayscale interpolation of the entire image is completed, thereby obtaining a phase correction reference image. This ensures that the generated phase correction reference image accurately represents the true crack state of the load phase associated with the detected image.

[0048] S5: Compare the detection image with the phase-corrected reference image to extract anomalous connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normals of the anomalous connected regions, including: The currently processed detection image and its corresponding phase-corrected reference image are divided into several image sub-regions at corresponding locations using the same image coordinate system. The method for dividing the image sub-regions is to determine the appropriate sub-region size based on the spatial resolution of the detection image and the required crack detection accuracy. For example, when the effective pixel count of the detection camera is 4000 pixels × 3000 pixels and the actual pixel size is 0.25 mm, the image sub-region size is set to 100 pixels × 100 pixels, covering an actual area of ​​25 mm × 25 mm. The determination of these dimensions can be based on the minimum crack length requirement and the computational efficiency of image processing, with the optimal size selected through preliminary experiments.

[0049] For each image sub-region that has been divided, the gray-level correlation between the detected image and the phase-corrected reference image at the corresponding positions in the sub-region is calculated. For example, a two-dimensional gray-level correlation coefficient calculation method can be used, and the calculation expression is as follows: ; in, This represents the gray-level correlation coefficient, which is dimensionless and ranges from -1 to 1. A value close to 1 indicates a high degree of gray-level correlation in image sub-regions. This represents the gray value of the detected image sub-region at pixel coordinates (x, y), in gray levels (e.g., 0 to 255). This represents the gray value of the phase-corrected reference image sub-region at pixel coordinates (x, y), expressed in gray levels. and These are the average gray values ​​of the corresponding image sub-regions.

[0050] After obtaining the gray-level correlation of all image sub-regions, image sub-regions with gray-level correlation below the anomaly threshold are identified, and multiple spatially adjacent such image sub-regions are merged to form anomaly connected regions. The anomaly threshold is set experimentally. For example, multiple images of normal, crack-free areas are acquired, gray-level correlation is calculated, and the lowest gray-level correlation value of the normal area is used as an initial reference. Then, images of cracked areas are acquired, gray-level correlation is calculated, and the optimal anomaly threshold is determined based on the statistical distribution of gray-level correlation between normal and cracked areas, for example, 0.75. The anomaly threshold is used to distinguish between cracked and normal areas.

[0051] After defining the abnormal connected regions, the edge pixels on both sides of the abnormal connected regions are extracted to determine the boundaries of the crack regions. The edge pixel extraction can be performed using the Canny edge detection algorithm. For example, the Gaussian filter kernel size is set to 3 pixels × 3 pixels, and the upper and lower thresholds of the Canny algorithm are set to 50 gray levels and 150 gray levels respectively to ensure that the edge information of the crack regions is extracted while eliminating background noise. The parameters were selected and optimized through preliminary experiments.

[0052] After extracting the edge pixels of the abnormal connected regions, the local tangent direction of each edge pixel is determined. A local pixel region (e.g., 5 pixels × 5 pixels) is selected centered on each edge pixel, and the local tangent is determined using the least squares line fitting method. Then, the normal direction of the edge of the abnormal connected region is determined based on the local tangent, which is the direction perpendicular to the tangent. The slope of the normal is the negative reciprocal of the slope of the tangent.

[0053] To determine the crack opening width, match the relative edge pixels on both sides of the crack region along the normal direction of the abnormal connected region edge. Search pixel by pixel along the edge normal direction until a matching pair of relative edge pixels is found. Calculate the Euclidean distance between each pair of edge pixels to obtain the crack opening spacing sequence; the crack opening spacing sequence represents the distribution characteristics of the crack opening width along the crack length direction.

[0054] Based on the positional distribution of edge pixels in the abnormally connected region, the endpoint coordinate sequence of the crack region is determined. Principal component analysis is used to determine the principal axis direction of the crack region. Then, the two farthest pixels along the principal axis are found among the edge pixels of the abnormally connected region. These two farthest pixels are defined as the endpoints of the crack region. When multiple pixels have the same maximum distance, the edge pixel with the largest projected distance along the principal axis is selected as the endpoint. The endpoint coordinate sequence represents the coordinate positions of the two farthest pixels.

[0055] S6: Based on the cyclic variation of the opening spacing sequence with the load phase and the continuous variation of the endpoint coordinate sequence with the number of operation cycles, the crack profile, crack length, and crack propagation rate are determined, including: For the same shooting location and the same operating cycle, the crack opening spacing sequence corresponding to different load phases is sorted and analyzed to obtain the trend characteristics of the crack opening spacing sequence changing with the load phase. Based on the temporal order of the load phases, for example, according to the order of the lowest load phase, the initial loading phase, the middle loading phase, the maximum loading phase, and the unloading phase, the crack opening spacing sequence for the same operating cycle is arranged sequentially. By analyzing the sorted opening spacing sequence, the cyclical opening and closing trend of the crack opening spacing under the action of the load phase is determined, that is, the crack opening spacing gradually increases when the load gradually increases, and gradually decreases when the load gradually decreases, thus forming a crack opening and closing trend.

[0056] For the same shooting point and the same load phase, the endpoint coordinate sequences corresponding to different operating cycles are sorted and analyzed to obtain the trend characteristics of crack endpoint coordinate changes with the operating cycle. Starting from the first cycle, the endpoint coordinate sequences are sorted sequentially according to the second, third, and so on operating cycles. The endpoint displacement direction is identified by the positional change trend of the endpoint coordinates between adjacent operating cycles. The determination of the endpoint displacement direction is specifically as follows: based on the image coordinate system, such as a two-dimensional coordinate system with the fixed boundary point on the component surface as the origin, the length direction as the X-axis, and the width direction as the Y-axis, the displacement direction of the crack endpoint coordinates between adjacent operating cycles is analyzed; the principal component analysis method is used to determine the direction of the crack principal axis, and the endpoint displacement trend is judged based on the direction of the crack principal axis to determine whether it shows a continuous outward expansion trend.

[0057] After obtaining two trend characteristics—the direction of change in the crack opening spacing sequence and the direction of displacement of the endpoint coordinate sequence—precise classification of abnormal connected regions is implemented: when the direction of change in the crack opening spacing shows an alternating trend with the load phase, i.e., the crack region periodically opens and closes with the load, and the direction of displacement of the endpoint coordinate sequence does not continue continuously along the running cycle, i.e., the endpoint coordinates remain relatively stable in spatial position or the displacement amplitude does not exceed a preset threshold, such abnormal connected regions are clearly defined as crack opening and closing regions. The displacement amplitude threshold is obtained, for example, through statistical analysis of a large amount of historical experimental data of typical crack opening and closing regions, and is set as the case where the displacement of the crack endpoint coordinates does not exceed 5 pixels between consecutive running cycles.

[0058] When the endpoint coordinate sequence exhibits a continuous outward expansion trend between consecutive running cycles, i.e., the endpoint coordinate displacement amplitude exceeds a preset threshold (e.g., more than 5 pixels), indicating that the crack length actually increases with the number of running cycles, such an abnormally connected region is defined as a crack propagation region. The displacement amplitude threshold of the crack propagation region is also determined experimentally to ensure accurate differentiation between the crack propagation region and the crack opening and closing region, preventing classification ambiguity or errors.

[0059] After classifying the crack regions, the method for identifying and determining the crack contour in abnormally connected regions identified as crack propagation areas is as follows: Edge detection algorithms, such as the Canny algorithm, are used to extract edge pixels to accurately identify the boundaries of the crack propagation regions. The Canny algorithm parameters include the Gaussian filter kernel size and upper and lower threshold settings. For example, the filter kernel size is set to 3 pixels × 3 pixels, and the upper and lower thresholds are set to 50 gray levels and 150 gray levels, respectively. The optimal parameters are determined through numerous experiments on crack images to improve edge detection accuracy and stability.

[0060] Based on the determined crack profile of the crack propagation region, the crack length is accurately calculated using the actual size corresponding to each pixel. The crack length is calculated by multiplying the pixel length measured in the crack profile image coordinate system by the actual size corresponding to each pixel. For example, when the actual size corresponding to each pixel is 0.25 mm, and the crack profile length measured in the image coordinate system is 400 pixels, then the actual crack length is calculated as 400 pixels × 0.25 mm / pixel = 100 mm. Both the method for determining the crack length and the actual size information of each pixel are derived from the definitions in Example S1.

[0061] Meanwhile, the crack propagation rate is calculated by dividing the actual difference in crack length between adjacent operating cycles by the actual time interval between adjacent operating cycles. For example, if the crack length increases by 10 mm between the first and second operating cycles, and the actual time interval between these two adjacent operating cycles is 100 hours, then the crack propagation rate is calculated as 10 mm ÷ 100 hours = 0.1 mm / hour.

[0062] To ensure the stability and accuracy of crack classification analysis and length measurement, the relevant parameters are set as follows: For example, the optimal range for image sub-region size is determined through extensive crack recognition experiments to be 50 pixels × 50 pixels to 100 pixels × 100 pixels; the anomaly threshold is determined to be 0.75 through crack image experimental analysis; the Gaussian filter kernel size for the Canny edge detection algorithm is selected as 3 pixels × 3 pixels; the displacement threshold for the crack endpoint coordinate sequence is 5 pixels; and the crack propagation rate calculation time interval is the actual recording time interval between adjacent running cycles. The selection methods and ranges of the above parameters and thresholds can all be obtained through extensive pre-experimental data analysis and statistical optimization, ensuring that the parameter selection is reasonable.

[0063] Example 2

[0064] The difference between Embodiment 2 and Embodiment 1 is that this embodiment describes a crack detection device for large components, such as... Figure 4 Specifically, it includes: The preprocessing module is used to plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and to acquire the load phase of the fatigue loading signal; The reference image acquisition module is used to detect reference images under multiple load phases acquired at each shooting point during the first round of camera operation, forming a reference image group; The detection image acquisition module is used to acquire detection images during cyclic operation and form detection image groups according to the shooting point, running direction and load phase. The phase correction module is used to select adjacent phase reference images at the same shooting point when there is a load phase difference between the detected image and the reference image, register the adjacent phase reference images and perform grayscale interpolation to generate a phase correction reference image. The crack detection module is used to compare the correlation between the detected image and the phase correction reference image, extract abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions. The crack assessment module is used to determine the crack profile, crack length, and crack propagation rate based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle.

[0065] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0066] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0067] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0068] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0069] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0070] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0071] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0072] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0073] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting cracks in large structural components, characterized in that, Includes the following steps: S1: Plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and acquire the load phase of the fatigue loading signal; S2: During the first round of operation of the detection camera, multiple reference images under the load phase are collected at each shooting point to form a reference image group; S3: Acquire detection images during cyclic operation, and form a detection image group according to the shooting point, running direction and load phase; S4: When there is a load phase difference between the detected image and the reference image, select an adjacent phase reference image at the same shooting point, register the adjacent phase reference image and perform grayscale interpolation to generate a phase correction reference image; S5: Compare the correlation between the detected image and the phase correction reference image, extract the abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions; S6: Based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle, the crack profile, crack length and crack propagation rate are determined.

2. The method for detecting cracks in large components according to claim 1, characterized in that, S1, specifically: Determine the boundary of the detection area based on the size of large components; The actual size of a single pixel is determined based on the detection camera resolution and detection accuracy, and the coverage area of ​​a single frame image is calculated according to the actual size of the single pixel. A sequence of shooting points is generated based on the detection area boundary and the coverage of a single frame image, and a round-trip scanning path is generated based on the sequence of shooting points. The fatigue loading signal is acquired synchronously, and the load phase is divided according to the periodic change of the fatigue loading signal.

3. The method for detecting cracks in large components according to claim 2, characterized in that, S2, specifically: The detection camera is controlled to run the first round along the round-trip scanning path. When the detection camera reaches the shooting point in the shooting point sequence, image acquisition is triggered according to the load phase. Use images corresponding to different load phases at the same shooting point as reference images; The reference images are associated with the shooting location, direction of travel, and load phase to form a reference image group.

4. The method for detecting cracks in large components according to claim 3, characterized in that, S3, specifically: After the reference image group is formed, the detection camera is controlled to run cyclically along the round-trip scanning path; When the detection camera reaches a shooting point in the shooting point sequence, a detection image is acquired based on the load phase. The detected images are associated with the shooting location, running direction, load phase, and running cycle, and are collected in the order of running cycle and shooting location to form a detection image group.

5. The method for detecting cracks in large components according to claim 4, characterized in that, S4, specifically: Read the shooting point, running direction and load phase associated with the detection image from the detection image group, and find the reference image with the same shooting point and running direction in the reference image group; The load phase associated with the detected image is compared with the load phase associated with the reference image. If the load phases are inconsistent, the reference image corresponding to the load phases adjacent to the load phase associated with the detected image is selected as the adjacent phase reference image. Image registration is performed on adjacent phase reference images, and grayscale interpolation is performed based on the position of the load phase associated with the detection image between the preceding and following adjacent load phases to generate a phase correction reference image.

6. The method for detecting cracks in large components according to claim 5, characterized in that, Image registration is performed on adjacent phase reference images. Gray-level interpolation is then performed based on the position of the load phase associated with the detected image between adjacent load phases to generate a phase correction reference image. Specifically: The adjacent phase reference image whose load phase is earlier than the load phase associated with the detection image is used as the previous phase reference image; The adjacent phase reference image whose load phase is later than the load phase associated with the detection image is used as the subsequent phase reference image; Register the rear phase reference image to the image coordinates of the front phase reference image; Based on the position of the load phase associated with the detected image between the load phase associated with the front phase reference image and the load phase associated with the rear phase reference image, grayscale interpolation is performed on the front phase reference image and the rear phase reference image to generate a phase correction reference image.

7. The method for detecting cracks in large components according to claim 6, characterized in that, S5, specifically: The detected image and the phase correction reference image are divided into image sub-regions corresponding to the same shooting point according to the image coordinates of the same shooting point, and the gray-level correlation between the image sub-regions corresponding to the same position is calculated. Image sub-regions with gray-level correlation below the abnormal threshold and adjacent to each other are merged to obtain abnormal connected regions. Extract the edge pixels on both sides of the abnormal connected region, and determine the edge normal of the abnormal connected region based on the local tangent direction of the edge pixels on both sides; Match relative edge pixels along the edge normal of the abnormal connected region, form an opening spacing sequence based on the distance between the relative edge pixels, and form an endpoint coordinate sequence based on the edge pixel coordinates at both ends of the abnormal connected region.

8. The method for detecting cracks in large components according to claim 7, characterized in that, S6, specifically: The aperture spacing sequence at the same shooting point and in the same running cycle is sorted according to the load phase to obtain the change of aperture spacing sequence with the load phase. The endpoint coordinate sequence under the same shooting point and the same load phase is sorted according to the running round to obtain the continuous change of the endpoint coordinate sequence with the running round; An abnormally connected region with reciprocating changes but without a continuously changing sequence of endpoint coordinates is designated as the crack opening and closing region, while an abnormally connected region with a continuously changing sequence of endpoint coordinates is designated as the crack propagation region. The crack profile is determined based on the crack propagation area, and the crack length and crack propagation rate are determined by combining the actual size corresponding to a single pixel.

9. A method for detecting cracks in large components according to claim 8, characterized in that, The determination of the crack initiation and propagation regions is as follows: By sorting the sequence of opening spacing at the same shooting point, the same running cycle, and different load phases, the direction of change of opening spacing can be obtained. By sorting the endpoint coordinate sequences under the same shooting point, the same load phase, and adjacent operating cycles, the direction of endpoint displacement is obtained. When the direction of the opening spacing changes alternately with the load phase and the direction of the endpoint displacement does not continue along the running cycle, the abnormal connected region is classified as the crack opening and closing region. When the endpoint displacement direction continues along the running cycle, the abnormally connected region is included in the crack propagation region.

10. A crack detection device for large components, characterized in that, include: The preprocessing module is used to plan the round-trip scanning path based on the size of the large component and the resolution of the inspection camera, and to acquire the load phase of the fatigue loading signal; The reference image acquisition module is used to detect reference images under multiple load phases acquired at each shooting point during the first round of camera operation, forming a reference image group; The detection image acquisition module is used to acquire detection images during cyclic operation and form detection image groups according to the shooting point, running direction and load phase. The phase correction module is used to select adjacent phase reference images at the same shooting point when there is a load phase difference between the detected image and the reference image, register the adjacent phase reference images and perform grayscale interpolation to generate a phase correction reference image. The crack detection module is used to compare the correlation between the detected image and the phase correction reference image, extract abnormal connected regions, and calculate the opening spacing sequence and endpoint coordinate sequence along the edge normal of the abnormal connected regions. The crack assessment module is used to determine the crack profile, crack length, and crack propagation rate based on the cyclical change of the opening spacing sequence with the load phase and the continuous change of the endpoint coordinate sequence with the running cycle.

Images