A method and system for detecting defects of complex precision parts based on optical detection

Abstract

The application provides a complex precision part defect detection method and system based on optical detection, and belongs to the technical field of defect detection. The method comprises the following steps: dividing the detection surface of a complex precision part to be detected into a plurality of sub-aperture regions, and obtaining a corresponding backscattering image sequence; converting each backscattering image into a corresponding Stokes parameter image, and extracting an optical rotation angle parameter and a depolarization parameter; constructing a polarization state feature map with the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and identifying a polarization state consistency region; spatially registering and splicing the polarization state consistency regions corresponding to adjacent sub-aperture regions to identify the polarization state evolution path corresponding to the spatial position points; and identifying the defect region of the complex precision part to be detected according to the polarization state evolution path of each spatial position point. The technical scheme provided by the application can realize accurate detection of defects of the complex precision part to be detected under the influence of abnormal polarization disturbance.

Description

Technical Field

[0001] This application relates to the field of defect detection technology, and more specifically, to a method and system for detecting defects in complex and precision components based on optical detection. Background Technology

[0002] In high-end manufacturing, the manufacturing process of complex and precision parts is prone to producing minute defects such as microcracks, pores, scratches, and dimensional deviations. Although these defects are difficult to detect with the naked eye, they can seriously affect the reliability, fatigue life, and safety of products. Traditional manual visual inspection is inefficient and highly subjective, making it difficult to meet the inspection requirements of micron-level precision and full surface coverage. Conventional automated optical inspection is often affected by factors such as large curvature of parts, strong reflection, or structural obstruction, resulting in high rates of missed and false detections. Therefore, researching high-precision and intelligent defect detection technologies is of great engineering significance for ensuring the quality of precision parts and reducing safety risks.

[0003] In existing precision component defect detection, physical principles such as optics, electromagnetics, and acoustics are typically integrated. In optical inspection, high-resolution cameras acquire images of the component surface, and algorithms such as threshold segmentation and edge detection are used to identify visible defects such as cracks and scratches that have significant differences in grayscale from the background. For small or hidden defects, technologies such as laser holography and infrared thermal imaging are used for defect identification. However, in visual defect detection of complex precision components (such as components with curved surfaces, specular highlights, micro-textures, or multi-layered media structures), traditional imaging methods based on intensity or a single polarization parameter cannot effectively distinguish between the inherent polarization response caused by the anisotropy of the material itself and changes in surface curvature and the abnormal polarization disturbance caused by small defects. This results in a large number of false defects in non-flat, multi-material interfaces, or high-roughness areas, leading to false detections of defects in the complex precision components under test. Therefore, how to achieve accurate detection of defects in complex precision components under the influence of abnormal polarization disturbances has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for detecting defects in complex precision components based on optical detection, which can achieve accurate detection of defects in complex precision components under the influence of abnormal polarization disturbances.

[0005] In a first aspect, this application provides a method for detecting defects in complex precision components based on optical inspection, comprising the following steps:

[0006] The surface of the complex and precision component to be tested is divided into multiple adjacent sub-aperture regions, and backscattering image sequences of the corresponding sub-aperture regions at multiple different polarization angles are obtained.

[0007] Each backscattered image in the backscattered image sequence is converted into a corresponding Stokes parametric image, and the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation are extracted from the Stokes parametric image.

[0008] Using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, a polarization state feature map is constructed for each sub-aperture region, and a polarization state consistency region is identified in the polarization state feature map.

[0009] Spatial registration and stitching are performed on the corresponding polarization state consistency regions in adjacent sub-aperture regions, and the polarization state evolution path of the same spatial location point in the stitched overlapping region in the original Stokes parametric image is calculated.

[0010] The defect region of the complex and precision component under test can be identified based on the polarization state evolution path of each spatial location point.

[0011] In some embodiments, dividing the surface of the complex precision component to be tested into multiple adjacent sub-aperture regions specifically includes:

[0012] The curvature distribution characteristics of the surface to be inspected are extracted from the three-dimensional digital model of the complex and precision component to be tested. Based on the curvature distribution characteristics, the surface to be inspected is divided into regions by growth to obtain multiple surface sheets.

[0013] An initial set of center points is generated by equidistant sampling of each surface piece, and a Veno map is constructed on the surface to be inspected based on the initial set of center points. The initial set of center points is then subjected to Lloyd relaxation iteration with the boundary of the Veno map polygon as a constraint to generate an optimized set of center points.

[0014] A sub-aperture grid covering the surface to be inspected is generated with each point in the optimized center point set as the center. Adjacent sub-apertures in the sub-aperture grid are connected to each other according to a preset overlap rate, thus completing the division of the sub-aperture region.

[0015] In some embodiments, obtaining a sequence of backscattered images of the corresponding sub-aperture region at multiple different polarization angles specifically includes:

[0016] The incident beam is modulated into linearly polarized light with a known polarization state by a polarizer, and then focused by an objective lens and projected into a sub-aperture region. The surface micro-morphology of the sub-aperture region modulates the incident polarization state and forms backscattered light carrying polarization information.

[0017] The backscattered light is collected by the objective lens and passes through the analyzer arranged in the same optical path as the polarizer. By rotating the analyzer to multiple preset polarization angles, the backscattered light components with different polarization azimuth angles pass through in sequence and are imaged on the photosensitive surface of the detector.

[0018] The detector performs one exposure at each preset polarization angle to acquire the backscattered image corresponding to the polarization angle. After traversing all preset polarization angles, a backscattered image sequence of the corresponding sub-aperture region is obtained.

[0019] In some embodiments, converting each backscattered image in the backscattered image sequence into a corresponding Stokes parametric image specifically includes:

[0020] Fourier series fitting is performed on the intensity values ​​of each pixel in the backscattered image sequence at different polarization detection angles to extract Fourier coefficients, which represent all polarization information of the backscattered light intensity modulated by the polarization detection angle.

[0021] The Stokes parameters of each spatial location point in the sub-aperture region are reconstructed from the Fourier coefficients according to the Stokes-Muller polarization algebraic relationship;

[0022] Arrange the values ​​of the same Stokes parameter at all spatial locations in pixel coordinates to form a Stokes parameter image for the corresponding sub-aperture region.

[0023] In some embodiments, extracting the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation from the Stokes parametric image specifically includes:

[0024] Using the difference between the horizontal and vertical linear polarization components and the difference between the 45° and 135° linear polarization components at each spatial location point in the Stokes parametric image as input, the optical rotation angle parameter at each spatial location point is extracted through arctangent operation;

[0025] Using the total light intensity component, the difference between horizontal and vertical linear polarization components, the difference between 45° and 135° linear polarization components, and the circular polarization component as inputs, the degree of polarization is calculated, and the depolarization parameter of each spatial location point is obtained by taking the complement of the degree of polarization.

[0026] In some embodiments, constructing a polarization state feature map for each sub-aperture region, using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, specifically includes:

[0027] Extract the optical rotation angle parameter and depolarization parameter of each spatial location point within the sub-aperture region to form a two-parameter spatial sample set with the optical rotation angle parameter and depolarization parameter as the dimensions;

[0028] A two-dimensional scatter coordinate system is established with the optical rotation angle parameter as the horizontal axis and the depolarization parameter as the vertical axis. Each spatial position point in the two-parameter spatial sample set is mapped to the two-dimensional scatter coordinate system according to the values ​​of its optical rotation angle parameter and depolarization parameter, thereby generating a polarization state feature map of the sub-aperture region.

[0029] In some embodiments, identifying the defect region of the complex precision component under test based on the polarization state evolution path at each spatial location point specifically includes:

[0030] Spatial domain difference is performed on the polarization state evolution path of each spatial location point in the overlapping region to extract the local jump features of the Stokes parameter between adjacent sub-apertures. The local jump features characterize the polarization state discontinuity response caused by the abrupt change in surface micromorphology at the spatial location point.

[0031] The local jump feature is matched with a preset defect polarization state response template, which contains Stokes parametric jump modes of known defect types. The spatial location point is then determined to be a defect candidate point based on the matching result.

[0032] Spatial connectivity analysis is performed on the candidate defect points, and they are aggregated to form defect regions.

[0033] Secondly, this application provides a defect detection system for complex precision parts based on optical inspection, used to perform a defect detection method for complex precision parts based on optical inspection. The system includes:

[0034] The acquisition module is used to divide the surface of the complex precision component under test into multiple adjacent sub-aperture regions and acquire the backscatter image sequence of the corresponding sub-aperture regions at multiple different polarization angles.

[0035] The processing module is used to convert each backscattered image in the backscattered image sequence into a corresponding Stokes parametric image, and extract the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation from the Stokes parametric image;

[0036] The processing module is also used to construct a polarization state feature map for each sub-aperture region with the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and to identify polarization state consistency regions in the polarization state feature map.

[0037] The processing module is also used to spatially register and stitch together the polarization state consistency regions corresponding to each other in adjacent sub-aperture regions, and to calculate the polarization state evolution path of the same spatial location point in the original Stokes parametric image in the stitched overlapping region.

[0038] The execution module is used to identify the defect area of ​​the complex and precision parts to be tested based on the polarization state evolution path of each spatial location point.

[0039] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described method for detecting defects in complex and precision components based on optical detection.

[0040] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for detecting defects in complex and precision components based on optical detection.

[0041] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0042] The method and system for detecting defects in complex precision components based on optical detection provided in this application first divides the surface of the complex precision component under test into multiple adjacent sub-aperture regions and acquires backscatter image sequences of the corresponding sub-aperture regions at multiple different polarization detection angles. Second, each backscatter image in the backscatter image sequence is converted into a corresponding Stokes parametric image, and the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation are extracted from the Stokes parametric image. Further, a polarization state feature map is constructed for each sub-aperture region using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and polarization state consistency regions are identified in the polarization state feature map. Then, the corresponding polarization state consistency regions in adjacent sub-aperture regions are spatially registered and stitched together, and the polarization state evolution path of the same spatial location point in the original Stokes parametric image is calculated in the stitched overlapping region. Finally, the defect region of the complex precision component under test is identified based on the polarization state evolution path of each spatial location point.

[0043] Therefore, this application demonstrates that it can accurately detect defects in complex precision components under the influence of abnormal polarization disturbances. Firstly, by dividing the complex curvature of the surface under test into multiple adjacent sub-aperture regions according to the field-of-view constraints of the optical system and acquiring a multi-analytical angle backscattering image sequence, the defocusing and resolution degradation problems caused by insufficient depth of field and curvature changes in full-field detection are effectively avoided. Secondly, the backscattering image sequence is converted into a Stokes parametric image, from which the optical rotation angle parameter and depolarization parameter are extracted. The implicit polarization information of light intensity modulated by the analytical angle is decoupled into quantitative parameters with clear physical meaning, where the optical rotation angle parameter reflects the local surface... The rotational effect of tilt and material anisotropy on the polarization direction, and the depolarization parameter reflecting the degree of damage to the polarization state retention ability caused by surface roughness and microcracks, enable different types of defects to obtain distinguishable numerical expressions in terms of polarization state properties. Furthermore, a polarization state feature map is constructed using the optical rotation angle parameter and the depolarization parameter as coordinate axes, and regions with consistent polarization states are identified. This maps polarization state analysis from the image spatial domain to a two-dimensional feature domain, allowing spatially similar points with similar polarization state responses to naturally cluster into high-density clusters in the feature domain, while anomalous defect regions are separated into outlier clusters due to polarization state deviations. This clustering method in the feature domain then automatically distinguishes between normal surface regions and various defect regions. Unsupervised separation avoids the limitations of traditional threshold segmentation methods that rely on experience for threshold selection. Then, spatial registration and stitching are performed on the polarization state consistency regions of adjacent sub-apertures. These polarization state consistency regions are used as structured matching primitives to replace pixel-by-pixel grayscale registration, significantly reducing registration ambiguities caused by local polarization state fluctuations and uneven illumination. Simultaneously, by calculating the polarization state evolution path, the static polarization state information of a single field of view is expanded into dynamic polarization state change information across the field of view, allowing for explicit capture of the polarization state response differences of defects under different illumination-observation geometries. Finally, defect regions are identified based on the polarization state evolution path, utilizing the polarization state of normal surface regions... The vibrational response exhibits a smooth continuity in cross-field observation, while the defect region presents an abrupt change in polarization state. By extracting local jump features and matching them with known defect polarization state response templates, accurate discrimination of fine-grained defects such as microcracks, surface scratches, and subsurface damage can be achieved. This effectively avoids the inability of traditional imaging methods based on intensity or a single polarization parameter to effectively distinguish between the inherent polarization response caused by the anisotropy of the material itself and changes in surface curvature, and the defect identification error caused by abnormal polarization disturbances due to minute defects. In summary, the technical solution provided in this application can achieve accurate detection of defects in complex precision parts under the influence of abnormal polarization disturbances. Attached Figure Description

[0044] Figure 1 This is an exemplary flowchart of a method for detecting defects in complex and precision components based on optical inspection, according to some embodiments of this application.

[0045] Figure 2This is an exemplary flowchart illustrating the determination of Stokes parametric images according to some embodiments of this application;

[0046] Figure 3 This is an exemplary flowchart illustrating a schematic diagram of identifying polarization state consistency regions according to some embodiments of this application;

[0047] Figure 4 This is a schematic diagram of the structure of a defect detection system for complex and precision components based on optical inspection, according to some embodiments of this application;

[0048] Figure 5 This is a schematic diagram of the structure of a computer device that implements a method for detecting defects in complex and precision components based on optical inspection, according to some embodiments of this application. Detailed Implementation

[0049] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] refer to Figure 1 The figure is an exemplary flowchart of a defect detection method for complex precision parts based on optical inspection, according to some embodiments of this application. The figure mainly includes the following steps:

[0051] In step S101, the surface of the complex precision component to be tested is divided into multiple adjacent sub-aperture regions, and backscattering image sequences of the corresponding sub-aperture regions at multiple different polarization angles are obtained.

[0052] In some embodiments, dividing the surface of the complex precision component to be tested into multiple adjacent sub-aperture regions is achieved by the following steps:

[0053] The curvature distribution characteristics of the surface to be inspected are extracted from the three-dimensional digital model of the complex and precision component to be tested. Based on the curvature distribution characteristics, the surface to be inspected is divided into regions by growth to obtain multiple surface sheets.

[0054] An initial set of center points is generated by equidistant sampling of each surface piece, and a Veno map is constructed on the surface to be inspected based on the initial set of center points. The initial set of center points is then subjected to Lloyd relaxation iteration with the boundary of the Veno map polygon as a constraint to generate an optimized set of center points.

[0055] A sub-aperture grid covering the surface to be inspected is generated with each point in the optimized center point set as the center. Adjacent sub-apertures in the sub-aperture grid are connected to each other according to a preset overlap rate, thus completing the division of the sub-aperture region.

[0056] In practice, firstly, based on the three-dimensional digital model of the complex and precision component to be tested, a local surface fitting method based on the Darboux framework is used to perform quadratic surface fitting on the point set in the neighborhood of each vertex on the triangular mesh. The principal curvature at that vertex is then calculated using the first and second derivatives of the fitted surface equation. , and Gaussian curvature This process forms the curvature distribution characteristics of the surface to be inspected. The principal curvature refers to the extreme value of the normal curvature in two orthogonal directions along the principal direction at a point on the surface. The Gaussian curvature refers to the product of the two principal curvatures. The three-dimensional digital model refers to a three-dimensional digital model that characterizes the geometry and spatial dimensions of the complex precision component to be tested. The curvature distribution characteristics refer to the spatial variation characteristics of the degree and shape of curvature in various local areas of the surface to be inspected. The curvature distribution characteristics are used as the similarity criterion for region growth. The triangular facet with the smallest curvature gradient is selected as the initial seed facet on the triangular mesh. The Gaussian curvature of the initial seed facet is compared with the Gaussian curvature of the adjacent triangular facets. If the absolute value of the curvature difference is smaller... If a preset curvature change threshold is reached, adjacent triangular facets are included in the growth region, and growth continues outward from the neighborhood of the newly included triangular facets until all adjacent triangular facets satisfying the similarity condition are aggregated, resulting in multiple surface patches with approximately constant curvature. The preset curvature change threshold is set according to the manufacturing tolerance and defect detection sensitivity requirements of the surface to be inspected. The surface patch refers to a set of triangular facets whose curvature change is within the set threshold and which are topologically connected. Then, for each surface patch, the number of sampling points is determined by the ratio of the total area of ​​the triangular facets within the surface patch to the square of the preset sampling interval. On the triangular mesh of the surface patch, a Poisson disk based on the centroid coordinates is used to generate an initial set of points that satisfy the minimum distance constraint. The initial center point set refers to discrete sampling points uniformly distributed on the surface sheet. The Poisson disk sampling refers to the distribution of sampling points generated on the curved surface where the distance between any two points is not less than a preset minimum distance. Based on the initial center point set, using the shortest path distance between any two points on the triangular mesh of the surface to be inspected as a metric, the Dijkstra algorithm is used to calculate the geodesic distance from each vertex of the triangular facet to each center point. Each vertex of the triangular facet is assigned to the center point with the smallest geodesic distance, forming geodesic Voronoi units for each center point on the surface to be inspected, thus completing the construction of the Veno map. The Veno map refers to the curved surface region divided according to the principle of shortest geodesic distance, with each center point on the surface to be inspected as a seed point. The domain set is defined by the polygonal boundaries of each Voronoi element in the Vinograph as spatial constraints. The weighted average of the three-dimensional coordinates of all triangular facet vertices within each Voronoi element is calculated using an area-weighted method as the centroid of that element. Each centroid is then updated to the coordinates of this centroid, completing one Lloyd relaxation iteration. This process of centroid calculation and centroid update is repeated until the displacement of each centroid is less than a preset convergence threshold or the preset maximum number of iterations is reached, generating an optimized centroid set. The optimized centroid set refers to the set of centroids where each point is located at the area-weighted centroid of the corresponding Voronoi element after the Lloyd relaxation iteration. The preset convergence threshold is one-thousandth of the sub-aperture size.Finally, the clear imaging range of the objective lens on the surface under test in the inspection optical system is taken as the object-side field of view. Centered on each point of the optimized center point set, planar rectangular windows are generated on the local tangent planes at each center point of the surface under test, using the rectangular boundary size of the object-side field of view. These planar rectangular windows are projected onto the surface under test along the surface normal direction at each center point, forming a sub-aperture grid covering the surface. The center-to-center spacing between adjacent sub-apertures is set according to the object-side field of view size multiplied by a preset overlap rate, ensuring that geometrically adjacent sub-aperture regions overlap. The ratio of this overlapping area to the area of ​​a single sub-aperture is equal to the preset overlap rate, forming an adjacency relationship and completing the division of the sub-aperture regions. The object-side field of view refers to the size of the rectangular area on the object surface that the optical system can clearly image. The preset overlap rate is a fixed value between 20% and 50%.

[0057] It should be noted that, in this application, the sub-aperture region refers to a local curved surface patch on the surface to be inspected that corresponds to the single imaging range. Determining the sub-aperture region can decompose a large-sized or complex curvature surface to be inspected into multiple local detection units that meet the field of view constraints of the optical system and can be stitched together with each other. This ensures that each local region is within the clear imaging depth and field of view of the objective lens in a single acquisition, thereby overcoming the defocus and insufficient resolution problems caused by curvature changes in full-field detection.

[0058] In some embodiments, obtaining a sequence of backscattered images of the corresponding sub-aperture region at multiple different polarization angles is achieved using the following steps:

[0059] The incident beam is modulated into linearly polarized light with a known polarization state by a polarizer, and then focused by an objective lens and projected into a sub-aperture region. The surface micro-morphology of the sub-aperture region modulates the incident polarization state and forms backscattered light carrying polarization information.

[0060] The backscattered light is collected by the objective lens and passes through the analyzer arranged in the same optical path as the polarizer. By rotating the analyzer to multiple preset polarization angles, the backscattered light components with different polarization azimuth angles pass through in sequence and are imaged on the photosensitive surface of the detector.

[0061] The detector performs one exposure at each preset polarization angle to acquire the backscattered image corresponding to the polarization angle. After traversing all preset polarization angles, a backscattered image sequence of the corresponding sub-aperture region is obtained.

[0062] In practice, firstly, in the illumination optical path of the detection optical system, a linear grating polarizer with a working wavelength covering 400nm to 700nm is installed as a polarizer in the parallel optical path between the collimating lens group and the beam splitter. This allows the unpolarized beam emitted by the halogen lamp or LED light source to be perpendicularly incident on the polarizer, transforming the transmitted light into linearly polarized light with an extinction ratio greater than 1000:1, and polarized at a polarization azimuth angle... As a reference for the incident polarization state, the polarizer refers to a polarizing optical element that uses the anisotropic absorption of a metal wire grid to convert incident unpolarized light into linearly polarized light. The extinction ratio refers to the ratio of the intensities of the two orthogonally polarized components in the transmitted light from the polarizing device. After being focused by an objective lens with a numerical aperture of 0.1 to 0.5, the linearly polarized light is projected onto the surface of the sub-aperture region in a near-normal incidence manner. The surface micromorphology at each spatial location within the sub-aperture region modulates the incident polarization state, affecting surface roughness and subsurface surface properties. Damage and material anisotropy alter the polarization state of the backscattered light relative to the incident polarization state, forming backscattered light carrying polarization information at that location. This backscattered light refers to the scattered light that, after interacting with the surface, propagates in the opposite direction along the incident direction and is collected again by the same objective lens. Then, the backscattered light returns along the original optical path, is deflected by the beam splitter, and enters the polarization analyzer path coaxial with the polarizer. In the polarization analyzer path, another linear grating polarizer of the same type is installed as the analyzer at a step resolution of [missing information]. On the electrically driven rotary stage, the analyzer refers to a polarizing optical element used to analyze the polarization state of the backscattered light modulated by the sample. The electric rotary stage is driven by a motion controller, causing the analyzer's transmission axis to rotate sequentially to... , , and Four preset polarization detection angles are used. After each preset polarization detection angle is reached, the stage pauses for 50ms to wait for mechanical stabilization. During this time, the component of the backscattered light whose polarization azimuth is aligned with the transmission axis of the analyzer is transmitted, while the component in the orthogonal direction is suppressed by the extinction ratio. The transmitted component is focused onto the photosensitive surface of the detector by an imaging lens group whose focal length matches that of the objective lens. The detector refers to a complementary metal-oxide-semiconductor image sensor with a pixel size of 3.45μm. Finally, after the stage stabilizes, the detector is controlled by an external trigger signal to perform an exposure with a duration of 10ms to 100ms. The charge accumulated on the photosensitive surface is converted into 12-bit or 16-bit digital grayscale values ​​by an analog-to-digital converter. The grayscale matrix is ​​read out through a USB 3.0 interface and stored as the backscattered image corresponding to the preset polarization detection angle. This process is repeated. , , and After four preset polarization angles, the four backscattered images are arranged in ascending order of polarization angle to form a backscattered image sequence corresponding to the sub-aperture region.

[0063] It should be noted that, in this application, the backscattering image sequence refers to a set of intensity images acquired in ascending order of angle at different polarization angles for the same sub-aperture region. The first image in a single image... Line 1 grayscale value of column pixels This indicates that the backscattered light at this spatial location point has a polarization analysis angle of . Determining the transmitted light intensity at a given time and the backscattering image sequence can provide a complete intensity modulation data foundation for subsequent polarization state calculations.

[0064] In step S102, each backscattered image in the backscattered image sequence is converted into a corresponding Stokes parameter image, and the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation are extracted from the Stokes parameter image.

[0065] In some embodiments, reference Figure 2 As shown, this figure is an exemplary flowchart of determining Stokes parametric images according to some embodiments of this application. In this embodiment, converting each backscattered image in the backscattered image sequence into a corresponding Stokes parametric image can be achieved by the following steps:

[0066] In step S1021, Fourier series fitting is performed on the intensity values ​​of each pixel in the backscattered image sequence at different polarization detection angles to extract Fourier coefficients. The Fourier coefficients characterize all polarization information of the backscattered light intensity modulated by the polarization detection angle.

[0067] In step S1022, the Stokes parameters of each spatial location point in the sub-aperture region are reconstructed from the Fourier coefficients according to the Stokes-Muller polarization algebraic relationship;

[0068] In step S1023, the values ​​of the same Stokes parameter at all spatial locations are arranged in pixel coordinates to form a Stokes parameter image for the corresponding sub-aperture region.

[0069] In practice, firstly, for each pixel corresponding to a spatial location point in the backscattered image sequence, the intensity values ​​at four polarization detection angles are extracted to form an intensity sequence. The least squares fitting method is used to perform Fourier series fitting on this intensity sequence. The Fourier series is truncated to include the DC component, the first cosine component, and the first sine component. After fitting, the DC Fourier coefficients, cosine Fourier coefficients, and sine Fourier coefficients for the corresponding spatial location point are obtained. These Fourier coefficients refer to the coefficients of each term after the intensity-to-polarization angle curve is expanded using a Fourier series. They characterize all polarization information modulated by the backscattered light intensity with the polarization detection angle, corresponding to the total backscattered light intensity, the cosine component of the linear polarization information, and the sine component, respectively. Based on the Stokes-Muller polarization algebraic relationship, the DC Fourier coefficients are subtracted from the cosine Fourier coefficients to obtain the total intensity component in the Stokes parameters. Twice the cosine Fourier coefficients are used as the difference between the horizontal and vertical linear polarization components in the Stokes parameter. Using twice the sine Fourier coefficients as the Stokes parameter and Difference of linear polarization components , and by , , Estimation of circular polarization components using polarization degree constraints Circular polarization component The estimation is constrained by the definition of polarization degree. From the relationship between the intensity of the fully polarized component and the total intensity in the Stokes parameter, it can be seen that fully polarized light satisfies... ,therefore Can be obtained from , , The Stokes parameters of each spatial location point are obtained and then reconstructed. The Stokes parameters refer to a set of four-dimensional vectors consisting of four real numbers. Stokes parameters are used to describe the polarization state and total intensity of backscattered light. Finally, the same Stokes component corresponding to all spatial locations within the sub-aperture region is arranged into a two-dimensional matrix according to the row and column order of the detector pixel coordinate system to form the Stokes parameter image of the sub-aperture region.

[0070] It should be noted that, in this application, the Stokes parameter image refers to the spatial distribution of the complete polarization state and total light intensity of backscattered light at each spatial location point within the sub-aperture region. Specifically, it refers to a collection of multiple component images stored in image form. The pixel value of each component image represents the value of the Stokes component at the corresponding spatial location point. Since each single image in the backscattered image sequence only records the transmitted light intensity at a specific analysis angle, it cannot directly reflect all the information of the polarization state of the backscattered light. However, the Stokes parameter, as the most complete mathematical description of the polarization state, encodes all polarization information modulated by the analysis angle with four orthogonal components, thereby decoupling the polarization information from the light intensity information. This provides the original polarization state data basis for the subsequent extraction of optical rotation angle parameters and depolarization parameters.

[0071] In some embodiments, the extraction of the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation from the Stokes parametric image is achieved by the following steps:

[0072] The difference between the horizontal and vertical linear polarization components at each spatial location point in the Stokes parametric image and the sum and The difference between the linear polarization components is used as input, and the optical rotation angle parameter of each spatial position point is extracted through arctangent operation;

[0073] The total light intensity component and the difference between the horizontal and vertical linear polarization components at each spatial location point in the Stokes parametric image are used as the basis for the analysis. and The difference between linear polarization components and circular polarization components are used as inputs to calculate the degree of polarization, and the polarization degree is complemented to obtain the depolarization parameter for each spatial position point.

[0074] In practical implementation, firstly, for the first sub-aperture region... Line 1 The spatial location point is read from the Stokes parametric image. and Component values, after normalization and Substitute the arctangent function in the four quadrants to calculate the optical rotation angle parameter. The range of values ​​for the optical rotation angle parameter is: The optical rotation angle parameter refers to the rotation of the azimuth angle of the backscattered photoelectric vector vibration relative to the incident polarization azimuth angle, reflecting the degree of modulation of the polarization direction of linearly polarized light by the surface microstructure; then, the same spatial location point is read from the Stokes parametric image. , , and Component value, It refers to the total light intensity component in the Stokes parameter at the spatial location point in the i-th row and j-th column of the sub-aperture region, which represents the total light intensity of the backscattered light at that point. It refers to the difference between the horizontal and vertical linear polarization components at a spatial location point in the i-th row and j-th column within the sub-aperture region, i.e. Intensity of linearly polarized light in the direction of polarization The difference in intensity of linearly polarized light in different directions characterizes the degree of linear polarization dominance of backscattered light in the two orthogonal directions, horizontal and vertical. This refers to the spatial location point in the i-th row and j-th column within the sub-aperture region. and The difference between the linearly polarized components, i.e. Intensity of linearly polarized light in the direction of polarization The difference in intensity of linearly polarized light in different directions characterizes the degree of linear polarization dominance of backscattered light in the diagonal direction. It refers to the circularly polarized component at the spatial location point in the i-th row and j-th column within the sub-aperture region, i.e., the difference between the intensity of right-handed and left-handed circularly polarized light. It characterizes the rotation direction and intensity of the circularly polarized component in the backscattered light, and the degree of polarization is calculated according to the definition of the degree of polarization. ,like If the value is zero, the polarization degree is zero. The polarization degree refers to the ratio of the intensity of the fully polarized component in the backscattered light to the total intensity, and its value ranges from [0,1]. The closer the polarization degree is to 1, the closer the backscattered light is to fully polarized light. The polarization degree is complemented by the depolarization parameter. The depolarization parameter at the spatial location point is calculated. The depolarization parameter refers to the ratio of the intensity of the completely unpolarized component of the backscattered light to the total intensity. The value range is [0,1]. The closer the depolarization parameter is to 1, the more severe the polarization state degradation is, reflecting the degree of disturbance of the incident polarization state by the surface micro-morphology.

[0075] It should be noted that extracting the optical rotation angle parameter and depolarization parameter from the Stokes parametric image can reduce the high-dimensional Stokes four-dimensional vector space into two feature parameters with clear physical meaning and sensitivity to surface micro-morphology anomalies. This allows the construction of a two-dimensional feature space for defect identification. The optical rotation angle parameter reflects the rotation of the azimuth angle of linearly polarized light vibration at various points on the surface due to factors such as local surface tilt, material anisotropy, or subsurface stress birefringence. The depolarization parameter reflects the degree of damage to the ability of surface roughness, microcracks, and subsurface scattering centers to maintain the incident polarization state. Both parameters quantify the surface optical response from the two dimensions of polarization direction modulation and polarization degree degradation, respectively. This allows different types of defects to exhibit distinguishable cluster distributions in the two-dimensional feature space, providing discriminative input features for subsequent identification of polarization state consistency regions through density clustering in the polarization state feature map.

[0076] In step S103, a polarization state feature map is constructed for each sub-aperture region using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and a polarization state consistency region is identified in the polarization state feature map.

[0077] In some embodiments, the following steps are used to construct a polarization state characteristic map for each sub-aperture region, with the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate:

[0078] Extract the optical rotation angle parameter and depolarization parameter of each spatial location point within the sub-aperture region to form a two-parameter spatial sample set with the optical rotation angle parameter and depolarization parameter as the dimensions;

[0079] A two-dimensional scatter coordinate system is established with the optical rotation angle parameter as the horizontal axis and the depolarization parameter as the vertical axis. Each spatial position point in the two-parameter spatial sample set is mapped to the two-dimensional scatter coordinate system according to the values ​​of its optical rotation angle parameter and depolarization parameter, thereby generating a polarization state feature map of the sub-aperture region.

[0080] In practice, firstly, the optical rotation angle parameter values ​​of each spatial location point within the sub-aperture region are extracted. and debias parameter values Each spatial location point is encoded into a two-dimensional feature vector. The two-dimensional feature vectors of all spatial locations are combined to form a two-parameter spatial sample set with the rotation angle parameter and depolarization parameter as dimensions. This two-parameter spatial sample set refers to the set in the two-dimensional feature space spanned by the rotation angle parameter and depolarization parameter, where the feature vector of each spatial location is a sample point, and the total number of sample points equals the total number of pixels in the sub-aperture region. Then, a Cartesian coordinate system is established on the two-dimensional plane, with the horizontal axis labeled as the rotation angle parameter and the range set to... The vertical axis is labeled as the depolarization parameter, and the range is set to [0,1]. The horizontal axis coordinate of each sample point in the dual-parameter spatial sample set is determined according to its optical rotation angle parameter value, and the vertical axis coordinate is determined according to its depolarization parameter value. The points are plotted point by point in the coordinate system in the form of a scatter plot. The scatter plot can be marked with a circle with a diameter of 1 coordinate unit to generate the polarization state characteristic map of the sub-aperture region.

[0081] It should be noted that, in this application, the polarization state feature map refers to the scatter distribution of the polarization state attributes of all spatial locations within the sub-aperture region on a two-dimensional plane of optical rotation angle and depolarization parameter. The spatial distribution density and aggregation morphology of the scatter points reflect the consistency or difference of the polarization state response of different regions of the surface under inspection. Determining the polarization state feature map involves mapping the polarization state attributes of each spatial location within the sub-aperture region from the spatial domain of the original image to a two-dimensional feature domain with optical rotation angle parameter and depolarization parameter as coordinate axes. In the feature domain, spatial locations with similar surface micromorphology and consistent polarization state response will aggregate into high-density scatter clusters. Local areas with defects or material anomalies will appear as outliers or secondary clusters separated from the main clusters due to deviations in their optical rotation angle or depolarization parameter. Thus, the distance and density relationship of the scatter distribution in the feature domain explicitly reveals the degree of polarization state similarity of different regions of the surface under inspection.

[0082] In some embodiments, identifying polarization state consistency regions in a polarization state feature map is achieved using the following steps:

[0083] Kernel density estimation is performed on the two-parameter spatial sample set in the polarization state feature map to obtain the probability density distribution of the two-parameter spatial sample set;

[0084] The local density peaks of the probability density distribution are extracted as cluster centers of the polarization state feature map;

[0085] Using the cluster center as the seed point, perform density-connected region growing in the polarization state feature map to aggregate sample points that meet the preset density connectivity criteria and have similar polarization state features into the same polarization state consistent region, and assign a region identifier to the polarization state consistent region.

[0086] The region identifier of the polarization state consistency region is reverse-mapped to the pixel coordinate system of the sub-aperture region to obtain the polarization state consistency region corresponding to the original spatial position.

[0087] In practice, firstly, for the two-parameter spatial sample set in the polarization state feature map, a Gaussian kernel function is selected for kernel density estimation. The bandwidth parameter h is calculated based on the number of sample points N and the covariance matrix of the two-dimensional feature space. Then, the normalized optical rotation angle parameter and depolarization parameter of each sample point are substituted into the multivariate Gaussian kernel density estimation formula. The probability density value is calculated point by point on the two-dimensional feature plane to generate a range covering the optical rotation angle. The probability density distribution surface within the debiased range [0,1] is taken as the probability density distribution, where, Position on the two-dimensional feature plane The probability density estimate at point N is the total number of sample points in the two-parameter spatial sample set. The bandwidth parameter is a dimension of the optical rotation angle parameter. The bandwidth parameter is a debiasing parameter. This represents the summation of i from 1 to N. For multidimensional standard Gaussian kernel function, Let be the two-dimensional feature space location vector whose probability density is to be estimated. Let h be the two-dimensional feature vector of the i-th sample point, and h be the bandwidth parameter vector. , Indicates to and The difference is subjected to dimensional bandwidth scaling. Kernel density estimation refers to a non-parametric method that uses a kernel function to smooth discrete sample points to estimate the overall probability density function. The probability density distribution describes the density of samples appearing at various points in the feature space. Secondly, for the probability density distribution surface corresponding to the probability density distribution, the feature plane formed by the optical rotation angle parameter and the depolarization parameter is used as the search space. The mean shift algorithm is used with all sample points as initial search points, iteratively shifting each search point along the rising direction of the probability density gradient until convergence to a local density maximum. The sample points are merged, and the point with the highest probability density in each merged class is extracted as a candidate cluster center. Candidate cluster centers with probability density values ​​greater than a preset density threshold are then selected as cluster centers for the polarization state feature map. The cluster center refers to the representative point in the feature space with the highest local probability density that can represent an independent sample cluster category. Then, each cluster center is used as a seed point, and its coordinates on the two-dimensional feature plane are used as the starting point for growth. The density connectivity criterion is set as follows: the probability density value of the sample point is not less than a preset connectivity threshold, and the feature similarity with the seed point under the Gaussian kernel distance metric is less than a preset distance cutoff value. The preset connectivity threshold is one-tenth of the cluster center probability density value, and the preset distance cutoff value is three times the bandwidth parameter h. Starting from various sub-points, sample points satisfying the density connectivity criterion are searched for in the neighborhood and included in the current cluster. The clustering continues to grow outwards with newly included sample points until no more sample points satisfying the condition can be added. All sample points aggregated by the same seed point are marked as the same polarization state consistency region, and each polarization state consistency region is assigned a unique integer region identifier, generating the clustering results in the polarization state feature map. The polarization state consistency region refers to the region in the two-dimensional feature space that is composed of rotational polarization density and uniform polarization density. A cluster set is formed by sample points with similar values ​​of optical angle parameter and depolarization parameter and connected probability density. Finally, the region identifier assigned to each sample point in the polarization state feature map is filled into the identifier matrix with the same pixel size as the sub-aperture region according to the row and column coordinates i and j of the original spatial position point of the sample point. The region identifier of the polarization state consistency region to which the spatial position point belongs is written at the i-th row and j-th column of the identifier matrix. The result of dividing the polarization state consistency region with the original spatial position is obtained. The region identifier refers to the numerical label assigned to each cluster region in the polarization state feature map.

[0088] It should be noted that, in this application, the polarization state consistency region refers to a set of spatially located points with uniform polarization modulation and depolarization response. Specifically, it is a connected cluster set formed by aggregated sample points with similar polarization state response characteristics in the two-dimensional feature space spanned by the optical rotation angle parameter and the depolarization parameter, after kernel density estimation and density connectivity clustering. By determining the polarization state consistency region, the polarization state attributes originally on a pixel-by-pixel basis within the sub-aperture region can be transformed into homogeneous region divisions with clear physical meaning, enabling the separation of normal surface regions and various defect regions at the regional scale. This provides structured regional primitives for cross-field matching and stitching of polarization state consistency regions between adjacent sub-apertures, avoiding matching ambiguities caused by local fluctuations in polarization state during pixel-by-pixel registration. At the same time, extracting the polarization state evolution path on a regional basis can enhance the statistical robustness of defect signals.

[0089] refer to Figure 3 This figure is a schematic diagram illustrating the identification of polarization state consistency regions according to some embodiments of this application. The horizontal axis of the figure represents the optical rotation angle parameter, and its range is limited to... The vertical axis represents the rotation of the azimuth angle of the linearly polarized light vibration, while the vertical axis represents the depolarization parameter, with a range limited to [0,1]. The vertical axis represents the degree of degradation of the polarization state of the backscattered light. The figure shows the polarization state characteristic distribution of each spatial location point within the sub-aperture region marked by dots, presenting a ring density distribution centered on the origin. The central high-density ring region is marked as the polarization state consistency region, corresponding to the polarization response characteristics of the normal region on the surface of the component under test. That is, the optical rotation angle and depolarization parameter distribution of the normal region are concentrated, and the polarization state consistency is strong. The low-density scattered points diverging outwards correspond to the polarization characteristics of the defect region. That is, the optical rotation angle and depolarization parameter distribution of the defect region are discrete, and the polarization state undergoes abnormal evolution. This figure intuitively quantifies the distribution difference between the normal region and the defect region in the polarization characteristic space, providing a visual characteristic distribution basis for subsequent identification of the polarization state consistency region and accurate location of component defects through density clustering.

[0090] In step S104, the corresponding polarization state consistency regions in adjacent sub-aperture regions are spatially registered and stitched together, and the polarization state evolution path of the same spatial location point in the original Stokes parametric image in the stitched overlapping region is calculated.

[0091] In some embodiments, spatial registration and stitching are performed on the corresponding polarization state consistency regions in adjacent sub-aperture regions, and the polarization state evolution path of the same spatial location point in the original Stokes parametric image in the stitched overlapping region is calculated using the following steps:

[0092] Extract the polarization state consistency region and its region identifier within the overlapping range of adjacent sub-aperture regions. Use the consistency of the region identifier of each spatial location point within the overlapping range in the adjacent sub-aperture regions as the matching cost. Solve the spatial transformation parameters between adjacent sub-aperture regions by minimizing the matching cost to complete the spatial registration of the polarization state consistency region.

[0093] Based on the spatial transformation parameters, the polarization state consistency regions of adjacent sub-aperture regions are transformed to a unified coordinate frame for splicing, and the coordinate mapping relationship of each spatial position point in the overlapping region in the adjacent sub-aperture regions is determined.

[0094] Based on the coordinate mapping relationship, the Stokes parameters of the same spatial location point in the overlapping region are extracted from the original Stokes parameter images corresponding to each of the adjacent sub-aperture regions. The Stokes parameters are then arranged into polarization state evolution paths according to the sub-aperture acquisition time sequence.

[0095] In specific implementation, firstly, in the adjacent k-th sub-aperture region and the... Within the overlapping area of ​​two sub-aperture regions, region identifiers assigned to each sub-aperture region after clustering are extracted. A pair of region identifier matching relationships is established for each spatial location point within the overlapping area. The number of spatial location points with inconsistent region identifiers between the two sub-apertures is counted as the matching cost function, and the negative of the matching cost function value is taken as the objective function. The matching cost refers to the relationship between the region identifier of the k-th sub-aperture and the region identifier of the k-th sub-aperture within the overlapping area. The number of pixels with different region identifiers for each sub-aperture is determined by taking the pixel coordinate system of the k-th sub-aperture region as a reference. Using the polarization state consistency region identification matrix of each aperture region as the object to be transformed, the cross-power spectrum is calculated on the two region identification maps of the overlapping region using the phase correlation method. An inverse Fourier transform is performed on the cross-power spectrum to obtain the impulse function. The peak position of the impulse function is taken to determine the initial translation parameters. Based on the initial translation parameters, the translation parameters are set at... Pixels, rotation parameters The search space within the range, with a step size of 1 pixel and a step size of Iterate through all parameter combinations in the search space, and for each parameter combination, perform the following steps: Perform a rigid body transformation on the identifier matrix of each aperture region, calculate the matching cost of the two identifier matrices in the overlapping region after the transformation, and take the translation amount corresponding to the minimum matching cost. , and rotation angle As spatial transformation parameters between adjacent sub-aperture regions, spatial registration of polarization-consistent regions is completed. These spatial transformation parameters refer to the two-dimensional translation vector and in-plane rotation angle describing the geometric transformation relationship between adjacent sub-apertures. Then, based on the spatial transformation parameters, the first... A rigid body transformation is performed on the polarization state consistency region identifier matrix of each sub-aperture region to unify it with the polarization state consistency region identifier matrix of the k-th sub-aperture region under the same coordinate frame. For positions in the overlapping region where both identifier matrices have values ​​at the same coordinate location, their row and column coordinates in sub-aperture region k are defined as follows: In the sub-aperture region The row and column coordinates are defined as follows: establish a one-to-one correspondence coordinate mapping relationship, wherein the coordinate mapping relationship refers to the overlapping area. The coordinate pairs of the same spatial location point within the domain are located in the pixel coordinate systems of adjacent sub-apertures. Finally, based on the coordinate mapping table, for each spatial location point within the overlapping region, the coordinates of that point are read from the Stokes parametric image of the k-th sub-aperture region. , , , Coordinates in component image The pixel value at that location is used to obtain the Stokes parameter vector of that spatial point under the sub-aperture k. Similarly, from the first Coordinates are read from the Stokes parametric image of the sub-aperture region. The component values ​​at each location are used to obtain the spatial location of the point in the sub-aperture. Stokes parameter vector The two Stokes parameter vectors are arranged into an ordered sequence containing the two vectors according to the timing of the sub-aperture acquisition, forming the polarization state evolution path of the spatial location point.

[0096] It should be noted that, in this application, the polarization state evolution path refers to the trajectory of the polarization state response of a spatial location point under different incident angles and observation azimuths. Specifically, it refers to the vector sequence formed by arranging the Stokes parameter vectors measured at the same spatial location point in adjacent sub-aperture regions according to the acquisition time sequence. This is used to reflect the direction sensitivity of surface micro-morphology, local optical anisotropy of materials, and the angle dependence of potential defects on polarization state modulation. Determining the polarization state evolution path can expand the static polarization state information obtained from a single sub-aperture observation into dynamic polarization state change information across the field of view. Since the polarization state response of normal surface regions has smooth continuity under different illumination-observation geometry conditions, while fine-grained defects such as microcracks, surface scratches, and subsurface damage can excite abnormal polarization state abrupt changes at specific incident azimuth angles, the polarization state evolution path of defect regions exhibits distinguishable characteristics such as large jump amplitude and low path smoothness. This provides a data foundation with spatiotemporal discriminative power for subsequent extraction of path morphology features and defect identification.

[0097] In step S105, the defect region of the complex precision component to be tested is identified based on the polarization state evolution path of each spatial location point.

[0098] In some embodiments, the defect region of the complex precision component under test is identified based on the polarization state evolution path at each spatial location point using the following steps:

[0099] Spatial domain difference is performed on the polarization state evolution path of each spatial location point in the overlapping region to extract the local jump features of the Stokes parameter between adjacent sub-apertures. The local jump features characterize the polarization state discontinuity response caused by the abrupt change in surface micromorphology at the spatial location point.

[0100] The local jump feature is matched with a preset defect polarization state response template, which contains Stokes parametric jump modes of known defect types. The spatial location point is then determined to be a defect candidate point based on the matching result.

[0101] Spatial connectivity analysis is performed on the candidate defect points, and they are aggregated to form defect regions.

[0102] In specific implementation, firstly, for each spatial location point in the overlapping region, the Stokes parameter vectors corresponding to adjacent sub-apertures in its polarization state evolution path are subjected to first-order forward difference operations component by component. The jump variables of each Stokes component are calculated, and normalized with the total light intensity to eliminate the influence of illumination intensity fluctuations. The four normalized jump variables are combined to form local jump features. The local jump feature set is obtained by traversing all spatial locations in the overlapping region. The local jump feature refers to a four-dimensional vector that quantifies the degree of polarization state discontinuity response across the field of view using the absolute value of the normalized difference between each Stokes component and adjacent sub-apertures. It is used to characterize the polarization state discontinuity response caused by abrupt changes in surface micromorphology at the spatial location point. Then, a defect polarization state response template library containing three defect types—microcracks, surface scratches, and subsurface damage—is established. The template for each defect type is the expected local jump feature and its tolerance range obtained from prior experimental calibration. The weighted Euclidean distance between the local jump feature of each spatial location point and all defect templates in the defect polarization state response template library is calculated one by one, and the weighting coefficient is taken as the tolerance of each component. If the weighted Euclidean distance is less than 1, the spatial location point is determined to be a candidate point of the corresponding defect type, and the defect type label is assigned to the location point. If the distance to all defect templates is not less than 1, the spatial location point is marked as a normal region. The defect polarization state response template refers to the set of local jump expectation features and tolerance range of Stokes parameters obtained from known defect type samples. The defect candidate point refers to the spatial location point whose local jump feature matches the defect template and is initially determined to be a defect. Finally, for all defect candidate points in the overlapping region that have been assigned defect type labels, the 8-neighborhood connectivity criterion is used to perform connectivity analysis in the pixel coordinate system. Defect candidate points that are spatially adjacent and have the same defect type label are merged into a connected region. The number of pixels contained in each connected region is counted. Connected regions with a number of pixels less than the preset minimum defect area threshold are filtered out to remove isolated noise points. The boundaries of the remaining connected regions are defined by the minimum bounding rectangle to obtain the defect region. The preset minimum defect area threshold is one-thousandth of the total number of pixels in the sub-aperture region.

[0103] It should be noted that, in this application, the defect area refers to the area on the surface of the complex precision component to be tested that is determined to have defects. The defect area corresponds to the projection range of fine-grained defects such as microcracks, surface scratches or subsurface damage on the surface of the complex precision component to be tested in the two-dimensional image space.

[0104] Furthermore, in another aspect of this application, in some embodiments, this application provides a defect detection system for complex and precision components based on optical inspection, referring to... Figure 4The figure is a schematic diagram of the structure of a defect detection system for complex precision parts based on optical detection according to some embodiments of this application. The system includes an acquisition module 201, a processing module 202, and an execution module 203, which are described below:

[0105] The acquisition module 201 in this application is mainly used to divide the surface of the complex precision component to be tested into multiple adjacent sub-aperture regions, and to acquire backscatter image sequences of the corresponding sub-aperture regions at multiple different polarization angles.

[0106] Processing module 202, in this application, is mainly used to convert each backscattered image in the backscattered image sequence into a corresponding Stokes parameter image, and extract the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation from the Stokes parameter image;

[0107] The processing module 202 is also used to construct a polarization state feature map for each sub-aperture region with the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and to identify polarization state consistency regions in the polarization state feature map.

[0108] In addition, the processing module 202 is also used to spatially register and stitch together the polarization state consistency regions corresponding to each other in adjacent sub-aperture regions, and calculate the polarization state evolution path of the same spatial location point in the original Stokes parametric image in the stitched overlapping region.

[0109] The execution module 203 in this application is mainly used to identify the defect area of ​​the complex precision parts to be tested based on the polarization state evolution path of each spatial location point.

[0110] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described method for detecting defects in complex precision parts based on optical detection.

[0111] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device implementing a method for detecting defects in complex precision components based on optical detection, according to some embodiments of this application. The method for detecting defects in complex precision components based on optical detection in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0112] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the optical inspection-based defect detection method for complex and precision components in this application.

[0113] The communication bus 302 can be used to transmit information between the aforementioned components.

[0114] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0115] The memory 303 stores program code for executing the solution of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the determination of the defect detection method for complex precision parts based on optical detection can be achieved by the processor 301 and one or more software modules in the program code in the memory 303.

[0116] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0117] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0118] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0119] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for detecting defects in complex and precision components based on optical detection.

[0120] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0121] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting defects in complex precision components based on optical inspection, characterized in that, The steps include the following: The surface of the complex and precision component to be tested is divided into multiple adjacent sub-aperture regions, and backscattering image sequences of the corresponding sub-aperture regions at multiple different polarization angles are obtained. Each backscattered image in the backscattered image sequence is converted into a corresponding Stokes parametric image, and the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation are extracted from the Stokes parametric image. Using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, a polarization state feature map is constructed for each sub-aperture region, and a polarization state consistency region is identified in the polarization state feature map. Spatial registration and stitching are performed on the corresponding polarization state consistency regions in adjacent sub-aperture regions, and the polarization state evolution path of the same spatial location point in the stitched overlapping region in the original Stokes parametric image is calculated. The defect region of the complex and precision component under test can be identified based on the polarization state evolution path of each spatial location point.

2. The method as described in claim 1, characterized in that, The surface of the complex and precision component to be tested is divided into multiple adjacent sub-aperture regions, specifically including: The curvature distribution characteristics of the surface to be inspected are extracted from the three-dimensional digital model of the complex and precision component to be tested. Based on the curvature distribution characteristics, the surface to be inspected is divided into regions by growth to obtain multiple surface sheets. An initial set of center points is generated by equidistant sampling of each surface piece, and a Veno map is constructed on the surface to be inspected based on the initial set of center points. The initial set of center points is then subjected to Lloyd relaxation iteration with the boundary of the Veno map polygon as a constraint to generate an optimized set of center points. A sub-aperture grid covering the surface to be inspected is generated with each point in the optimized center point set as the center. Adjacent sub-apertures in the sub-aperture grid are connected to each other according to a preset overlap rate, thus completing the division of the sub-aperture region.

3. The method as described in claim 1, characterized in that, Obtaining the backscattering image sequence of the corresponding sub-aperture region at multiple different polarization angles specifically includes: The incident beam is modulated into linearly polarized light with a known polarization state by a polarizer, and then focused by an objective lens and projected into a sub-aperture region. The surface micro-morphology of the sub-aperture region modulates the incident polarization state and forms backscattered light carrying polarization information. The backscattered light is collected by the objective lens and passes through the analyzer arranged in the same optical path as the polarizer. By rotating the analyzer to multiple preset polarization angles, the backscattered light components with different polarization azimuth angles pass through in sequence and are imaged on the photosensitive surface of the detector. The detector performs one exposure at each preset polarization angle to acquire the backscattered image corresponding to the polarization angle. After traversing all preset polarization angles, a backscattered image sequence of the corresponding sub-aperture region is obtained.

4. The method as described in claim 1, characterized in that, Converting each backscattered image in the backscattered image sequence into a corresponding Stokes parametric image specifically includes: Fourier series fitting is performed on the intensity values ​​of each pixel in the backscattered image sequence at different polarization detection angles to extract Fourier coefficients, which represent all polarization information of the backscattered light intensity modulated by the polarization detection angle. The Stokes parameters of each spatial location point in the sub-aperture region are reconstructed from the Fourier coefficients according to the Stokes-Muller polarization algebraic relationship; Arrange the values ​​of the same Stokes parameter at all spatial locations in pixel coordinates to form a Stokes parameter image for the corresponding sub-aperture region.

5. The method as described in claim 1, characterized in that, The optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation are extracted from the Stokes parametric image. Specifically, these parameters include: Using the difference between the horizontal and vertical linear polarization components and the difference between the 45° and 135° linear polarization components at each spatial location point in the Stokes parametric image as input, the optical rotation angle parameter at each spatial location point is extracted through arctangent operation; Using the total light intensity component, the difference between horizontal and vertical linear polarization components, the difference between 45° and 135° linear polarization components, and the circular polarization component as inputs, the degree of polarization is calculated, and the depolarization parameter of each spatial location point is obtained by taking the complement of the degree of polarization.

6. The method as described in claim 1, characterized in that, Using the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, the construction of a polarization state feature map for each sub-aperture region specifically includes: Extract the optical rotation angle parameter and depolarization parameter of each spatial location point within the sub-aperture region to form a two-parameter spatial sample set with the optical rotation angle parameter and depolarization parameter as the dimensions; A two-dimensional scatter coordinate system is established with the optical rotation angle parameter as the horizontal axis and the depolarization parameter as the vertical axis. Each spatial position point in the two-parameter spatial sample set is mapped to the two-dimensional scatter coordinate system according to the values ​​of its optical rotation angle parameter and depolarization parameter, thereby generating a polarization state feature map of the sub-aperture region.

7. The method as described in claim 1, characterized in that, Based on the polarization state evolution path at each spatial location point, the defect regions of the complex and precision parts under test are identified, specifically including: Spatial domain difference is performed on the polarization state evolution path of each spatial location point in the overlapping region to extract the local jump features of the Stokes parameter between adjacent sub-apertures. The local jump features characterize the polarization state discontinuity response caused by the abrupt change in surface micromorphology at the spatial location point. The local jump feature is matched with a preset defect polarization state response template, which contains Stokes parametric jump modes of known defect types. The spatial location point is then determined to be a defect candidate point based on the matching result. Spatial connectivity analysis is performed on the candidate defect points, and they are aggregated to form defect regions.

8. A defect detection system for complex precision components based on optical inspection, used to execute the defect detection method for complex precision components based on optical inspection as described in any one of claims 1 to 7, characterized in that, The system includes: The acquisition module is used to divide the surface of the complex precision component under test into multiple adjacent sub-aperture regions and acquire the backscatter image sequence of the corresponding sub-aperture regions at multiple different polarization angles. The processing module is used to convert each backscattered image in the backscattered image sequence into a corresponding Stokes parametric image, and extract the optical rotation angle parameter reflecting the rotation direction of linearly polarized light and the depolarization parameter reflecting the degree of polarization state preservation from the Stokes parametric image; The processing module is also used to construct a polarization state feature map for each sub-aperture region with the optical rotation angle parameter as the abscissa and the depolarization parameter as the ordinate, and to identify polarization state consistency regions in the polarization state feature map. The processing module is also used to spatially register and stitch together the polarization state consistency regions corresponding to each other in adjacent sub-aperture regions, and to calculate the polarization state evolution path of the same spatial location point in the original Stokes parametric image in the stitched overlapping region. The execution module is used to identify the defect area of ​​the complex and precision parts to be tested based on the polarization state evolution path of each spatial location point.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the optical inspection-based defect detection method for complex precision components as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the optical detection-based defect detection method for complex precision components as described in any one of claims 1 to 7.

Images