A Defect Identification System for the Production of New Ceramic Materials

Abstract

This invention relates to the field of defect detection technology, specifically a defect identification system for the production of new ceramic materials. The system includes a three-dimensional morphology reconstruction module, a curvature adaptive illumination module, a multi-angle light field acquisition module, a defect pixel initial screening module, a defect region extraction module, and a defect intelligent identification module. The system performs three-dimensional structured light scanning on the surface and reconstructs triangular meshes to obtain digital morphology data; it performs curvature adaptive mapping on the light source array to obtain a differentiated illumination strategy; it sequentially triggers multi-angle structured light projection on the surface and performs exposure capture to obtain a multi-angle reflected light field image sequence; it performs temporal feature analysis on the same pixel coordinate positions and performs deviation comparison to obtain defect candidate pixels; it performs morphological closure analysis on the spatiotemporal response spectrum to obtain defect candidate regions; and it measures the similarity between feature parameters and historical defect samples to obtain the defect identification result. This invention can improve the efficiency of defect identification.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a defect identification system for the production of new ceramic materials. Background Technology

[0002] In the field of defect identification technology in the production of new ceramic materials, existing technologies do not combine the surface structural characteristics of new ceramic materials with targeted processing when collecting and reconstructing the surface morphology. They only obtain basic surface data through conventional scanning methods and do not perform professional triangular mesh reconstruction processing on the collected point cloud data. As a result, they cannot accurately obtain digital morphology data of new ceramic materials, nor can they obtain accurate normal vector distribution based on morphology characteristics. This makes the subsequent lighting methods configured for the ceramic surface lack adaptability and make it difficult to clearly capture the detailed information of the ceramic surface.

[0003] Existing defect identification technologies for new ceramic materials have significant technical shortcomings in the defect detection and identification stages. When screening defect pixels, pixel coordinate system registration and temporal feature analysis of depth are not performed on the light field image sequence. Only simple feature comparison is used to complete the initial pixel screening, resulting in low accuracy of the obtained defect candidate pixels. At the same time, when extracting defect candidate regions, scientific morphological closure analysis of the spatiotemporal response map of candidate pixels is not performed, resulting in inaccurate defect region labeling. Furthermore, in the final defect identification, multi-dimensional similarity measurement is not combined with historical defect samples. Defect judgment is completed by relying on a single feature, resulting in the overall defect identification accuracy and efficiency failing to meet the actual testing needs of new ceramic material production. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a defect identification system for the production of new ceramic materials, characterized in that the system includes a three-dimensional morphology reconstruction module, a curvature adaptive illumination module, a multi-angle light field acquisition module, a defect pixel initial screening module, a defect region extraction module, and a defect intelligent identification module, wherein: The three-dimensional morphology reconstruction module is used to perform structured light three-dimensional scanning on the surface of the new ceramic material and to reconstruct the collected surface point cloud data into a triangular mesh to obtain the digital morphology data of the new ceramic material. The curvature adaptive lighting module is used to perform curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digital topography data, so as to obtain a differentiated lighting strategy for the new ceramic material. The multi-angle light field acquisition module is used to sequentially trigger multi-angle structured light projection on the surface of the new ceramic material based on the differentiated lighting strategy, and simultaneously perform exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material. The defect pixel screening module is used to perform temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence, and to compare the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain the defect candidate pixels of the new ceramic material. The defect region extraction module is used to perform morphological closure analysis on the spatiotemporal response map of the defect candidate pixels to obtain the defect candidate regions of the new ceramic material. The defect intelligent identification module is used to measure the similarity between the feature parameters of the defect candidate region and the historical defect samples of the new ceramic material, so as to obtain the defect identification result of the new ceramic material.

[0005] In a preferred embodiment, when the three-dimensional topography reconstruction module performs structured light three-dimensional scanning on the surface of the new ceramic material and reconstructs triangular meshes from the collected surface point cloud data to obtain the digital topography data of the new ceramic material, it is specifically used for: Based on a coded structured light generation strategy, a high-frequency structured light field is projected onto the surface of a new ceramic material to obtain a composite grating pattern of the new ceramic material. High-frequency image sequence acquisition is performed on the distorted structured light field of the composite grating pattern to obtain the distorted structured light image sequence of the new ceramic material; Based on the phase wrapping value in the distorted structured light image sequence, the distorted structured light image sequence is phase demodulated to obtain the continuous phase distribution of the new ceramic material; The continuous phase distribution is mapped to the three-dimensional spatial coordinates of the new ceramic material, and the three-dimensional spatial coordinates are discretely organized into point cloud to obtain the surface point cloud data of the new ceramic material. The surface point cloud data is subjected to Delaunay triangulation to obtain the digital morphology data of the new ceramic material.

[0006] In a preferred embodiment, when the curvature adaptive lighting module performs curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digitized topography data to obtain a differentiated lighting strategy for the new ceramic material, it is specifically used for: The surface normal vector field inversion is performed on the digital morphology data to obtain the normal vector distribution data of the new ceramic material; Based on the normal vector distribution data, the surface of the new ceramic material is segmented into curvature feature regions to obtain the curvature feature regions of the new ceramic material. Based on the principal direction of the normal vector of the triangular facet in the curvature feature region, the incident angle of the light source array of the new ceramic material is calibrated to obtain the incident angle configuration parameters of the new ceramic material. Based on the incident angle configuration parameters, the triggering timing and incident angle of the light sources in the light source array are jointly arranged to obtain the differentiated lighting strategy for the new ceramic material.

[0007] In a preferred embodiment, when the multi-angle light field acquisition module executes the differential illumination strategy to sequentially trigger multi-angle structured light projection onto the surface of the new ceramic material, and simultaneously performs exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material, it is specifically used for: Based on the incident angle calibration parameters in the differentiated lighting strategy, the incident angle servo positioning of the light source group in the light source array is performed to obtain the directional light source array of the new ceramic material. Based on the trigger timing arrangement in the differentiated lighting strategy, the directional light source array is sequentially excited to obtain the light source group identification code of the new ceramic material; The reflected light field of the new ceramic material is sampled to obtain the original reflected light field image frame sequence of the new ceramic material; Based on the light source group identifier encoding, the original reflected light field image frame sequence is decoded by the light source group identifier to obtain the reflected light field image frame of the new ceramic material. The reflected light field image frame is mapped to pixel coordinate space to obtain the multi-angle reflected light field image frame of the new ceramic material; The multi-angle reflected light field image frames are serialized and grouped to generate the multi-angle reflected light field image sequence.

[0008] In a preferred embodiment, when the defect pixel screening module performs temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence and compares the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain candidate defect pixels of the new ceramic material, it is specifically used for: Pixel coordinate system registration is performed on the image frames in the multi-angle reflected light field image sequence to obtain the spatially registered image sequence of the new ceramic material; The radiance response amplitude at the spatial pixel coordinates in the spatially registered image sequence is extracted in time sequence to obtain the radiance response curve of the new ceramic material; By deconstructing the trajectory morphology features of the radiance response curve, the spatiotemporal fingerprint spectrum of the new ceramic material is obtained; Based on the normal vector distribution field of the digital morphology data, spatial proximity analysis is performed on the spatial pixel coordinates in the spatiotemporal feature fingerprint map to construct the dynamic reference feature fingerprint of the new ceramic material. Based on the dynamic reference feature fingerprint, the spatiotemporal feature fingerprint spectrum is offset and compared to obtain the candidate pixels for defects in the new ceramic material.

[0009] In a preferred embodiment, when the defect pixel screening module performs spatial proximity analysis on the spatial pixel coordinates in the spatiotemporal feature fingerprint map based on the normal vector distribution field of the digitized topography data to construct the dynamic reference feature fingerprint of the new ceramic material, it is specifically used for: The spatial search radius is determined based on the surface normal vector of the spatial pixel coordinate position, and a neighborhood window is delineated with the spatial pixel coordinate position as the center and the spatial search radius as the constraint to construct the spatial neighborhood pixel set of the new ceramic material. The normal vector deviation of the pixel coordinates in the spatial neighborhood pixel set is analyzed, and the pixel coordinates are sorted in ascending order according to the obtained vector angle to obtain the candidate reference pixel set of the new ceramic material. The spatiotemporal feature fingerprints of the candidate reference pixel set are fused to obtain the dynamic reference feature fingerprints of the new ceramic material.

[0010] In a preferred embodiment, when the defect pixel screening module performs the following steps to construct the spatial neighborhood pixel set of the new ceramic material: It determines the spatial search radius based on the surface normal vector of the spatial pixel coordinate position, and delineates a neighborhood window centered on the spatial pixel coordinate position and constrained by the spatial search radius: Obtain the vector angle between the surface normal vector at the spatial pixel coordinate position and the mean normal vector of the global normal vector distribution field in the digitized topography data; Based on the vector angle, the spatial search radius of the spatial pixel coordinate position is dynamically calibrated to obtain the dynamic search radius of the new ceramic material; Centered on the spatial pixel coordinates and constrained by the dynamic search radius, a neighborhood window is delineated in the pixel coordinate space of the spatiotemporal feature fingerprint to obtain the initial neighborhood pixel set of the new ceramic material; The initial neighborhood pixel set is calibrated by a distance weighting factor to obtain the spatial neighborhood pixel set of the new ceramic material.

[0011] In a preferred embodiment, when the defect pixel screening module performs deviation comparison on the spatiotemporal feature fingerprint spectrum based on the dynamic reference feature fingerprint to obtain candidate defect pixels of the new ceramic material, it is specifically used for: The comprehensive deviation of the spatial pixel coordinates in the spatiotemporal feature fingerprint is calibrated, and the formula for calculating the comprehensive deviation is as follows: ; In the formula, The spatial pixel coordinates in the spatiotemporal feature fingerprint map are... The comprehensive deviation of the spatial pixel coordinate position. The rising edge steepness feature value is the spatial pixel coordinate position. This refers to the rising edge steepness reference value in the dynamic reference feature fingerprint. The peak response angle feature value is the location of the spatial pixel coordinates. This refers to the peak response angle reference value in the dynamic reference feature fingerprint. The decay rate feature value is the spatial pixel coordinate position. This is the attenuation rate reference value in the dynamic reference feature fingerprint. The oscillation period feature value of the spatial pixel coordinate position is... The oscillation period reference value in the dynamic reference feature fingerprint is used. The preset stability factor, The preset rising edge steepness weighting coefficient, The preset peak response angle weighting coefficient, The preset decay rate weighting coefficient, The weighting coefficients are the preset oscillation period dimension. The preset rising edge steepness index adjustment factor, The preset peak response angle exponential adjustment factor, The preset decay rate exponential adjustment factor, This is the preset oscillation period dimension index adjustment factor.

[0012] In a preferred embodiment, when the defect region extraction module performs morphological closure analysis on the spatiotemporal response maps of the defect candidate pixels to obtain the defect candidate regions of the new ceramic material, it is specifically used for: The spatial distribution field of the spatiotemporal response map of the defect candidate pixels is reconstructed to obtain the spatial distribution field data of the new ceramic material; Based on the spatial distribution field data, the spatiotemporal response spectrum of the defect candidate pixels is dynamically constructed using structural elements to obtain the dynamic structural elements of the new ceramic material. Based on the dynamic structural element, morphological dilation analysis is performed on the pixel coordinate positions in the spatiotemporal response map. When there are defect candidate pixels within the coverage area of ​​the dynamic structural element, the pixel coordinate positions are marked as dilated domain pixels to obtain the dilated domain connectivity map of the new ceramic material. Based on the dynamic structural element, morphological erosion analysis is performed on the expansion domain connectivity graph. When the entire coverage area of ​​the dynamic structural element is an expansion domain pixel, the expansion domain pixel label at the pixel coordinate position is retained to obtain the closed domain connectivity graph of the new ceramic material. The connected domain boundaries of the closed domain connected graph are extracted, and the pixel coordinate range is calibrated according to the obtained edge contour to obtain the defect candidate region of the new ceramic material.

[0013] In a preferred embodiment, when the defect intelligent recognition module performs a similarity measurement between the feature parameters of the defect candidate region and historical defect samples of the new ceramic material to obtain the defect recognition result of the new ceramic material, it is specifically used for: Based on the geometric morphological feature parameters of the feature parameter map in the defect candidate region, the historical defect sample library of the new ceramic material is selected to obtain the dynamic historical defect sample set of the new ceramic material. The feature parameters of the historical defect samples in the dynamic historical defect sample set are calibrated to obtain the dynamic defect type feature map of the new ceramic material; Based on the dynamic defect type feature map, the similarity of the feature parameter map is measured in terms of rising edge steepness, peak response angle, decay rate, oscillation period and geometric shape, so as to obtain the similarity index distribution of the new ceramic material. By comprehensively evaluating the distribution of the similarity index, the defect identification results of the new ceramic material are obtained.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention uses structured light 3D scanning combined with triangular mesh reconstruction to accurately acquire digital morphological data of new ceramic materials. Based on the normal vector distribution of this data, the curvature of the light source array is adaptively mapped to form a differentiated lighting strategy that fits the surface characteristics of the new ceramic material. The multi-angle structured light projection and exposure capture completed by this strategy can obtain a multi-angle reflected light field image sequence with complete details and clear features, which greatly improves the accuracy and adaptability of light field data acquisition on the surface of new ceramic materials, and lays a solid data foundation for subsequent defect identification.

[0015] 2. This invention performs refined temporal feature analysis and deviation comparison on multi-angle reflected light field image sequences, which can accurately screen out defect candidate pixels. Then, by processing the spatiotemporal response spectrum of the defect candidate pixels through morphological closure analysis, the defect candidate region can be accurately extracted. Finally, the feature parameters of the defect candidate region are compared with historical defect samples in multiple dimensions to achieve accurate determination of defects in new ceramic materials. This comprehensively improves the overall accuracy of defect identification in new ceramic materials, while making the defect identification process more systematic and efficient, and effectively ensuring the quality control effect of production and testing of new ceramic materials. Attached Figure Description

[0016] Figure 1 A system architecture diagram of a defect identification system for the production of new ceramic materials provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0019] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0020] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0021] In practice, the server-side equipment deployed in a defect identification system for the production of new ceramic materials may consist of one or more devices. This defect identification system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node, providing a defect identification system for the production of new ceramic materials to various users. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various users. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide a defect identification system for the production of new ceramic materials to various users.

[0022] In terms of implementation, the defect identification system for the production of new ceramic materials and the user terminal are mutually compatible. That is, if the defect identification system for the production of new ceramic materials is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the defect identification system for the production of new ceramic materials is implemented as a website, then the user terminal is implemented as a webpage; or if the defect identification system for the production of new ceramic materials is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0023] like Figure 1 The diagram shown is a system architecture diagram of a defect identification system for the production of new ceramic materials according to an embodiment of the present invention.

[0024] The defect identification system for the production of new ceramic materials described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the defect identification system for the production of new ceramic materials may include a three-dimensional shape reconstruction module, a curvature adaptive illumination module, a multi-angle light field acquisition module, a defect pixel initial screening module, a defect region extraction module, and a defect intelligent identification module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0025] In this embodiment of the invention, in a defect identification system for the production of new ceramic materials, each of the above-mentioned modules can be implemented independently and can be called by other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the defect identification system for the production of new ceramic materials provided by this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the defect identification system for the production of new ceramic materials. In practical applications, the above-mentioned modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0026] The following describes, with reference to specific embodiments, each component and its specific workflow of a defect identification system for the production of new ceramic materials: The three-dimensional morphology reconstruction module is used to perform structured light three-dimensional scanning on the surface of the new ceramic material and to reconstruct the collected surface point cloud data into a triangular mesh to obtain the digital morphology data of the new ceramic material. In this embodiment of the invention, when the three-dimensional topography reconstruction module performs structured light three-dimensional scanning on the surface of the new ceramic material and reconstructs the collected surface point cloud data into a triangular mesh to obtain the digital topography data of the new ceramic material, it is specifically used for: Based on a coded structured light generation strategy, a high-frequency structured light field is projected onto the surface of a new ceramic material to obtain a composite grating pattern of the new ceramic material. High-frequency image sequence acquisition is performed on the distorted structured light field of the composite grating pattern to obtain the distorted structured light image sequence of the new ceramic material; Based on the phase wrapping value in the distorted structured light image sequence, the distorted structured light image sequence is phase demodulated to obtain the continuous phase distribution of the new ceramic material; The continuous phase distribution is mapped to the three-dimensional spatial coordinates of the new ceramic material, and the three-dimensional spatial coordinates are discretely organized into point cloud to obtain the surface point cloud data of the new ceramic material. The surface point cloud data is subjected to Delaunay triangulation to obtain the digital morphology data of the new ceramic material.

[0027] According to the pre-set coded structured light generation strategy that generates structured light according to fixed coding rules, the high-frequency structured light field formed by the preset high-frequency structured light is uniformly projected onto the complete surface of the new ceramic material in a direction perpendicular to the surface of the new ceramic material through a structured light projection device. During the projection process, the light field is guaranteed to cover all physical areas of the surface of the new ceramic material, and finally a complete and continuous composite grating pattern is formed on the surface of the new ceramic material.

[0028] The distorted structured light field is a structured light field formed by the morphological distortion of the composite grating pattern due to the undulations of the surface of the new ceramic material. The distorted structured light field is continuously imaged by an industrial area array camera at a preset fixed acquisition frame rate adapted to the changes of the high-frequency structured light field. During the acquisition process, the camera shooting angle is kept at a fixed angle of 90 degrees with the surface of the new ceramic material. The continuously acquired images are arranged in the order of actual acquisition time to form a distorted structured light image sequence of the new ceramic material.

[0029] Phase wrapping is the value of the phase value of each pixel in each frame of the distorted structured light image sequence within the fixed phase range of 0-2π, forming a wrapped state. Phase demodulation is to perform point-by-point phase unwrapping processing on the phase wrapping value of each pixel in each frame of the distorted structured light image sequence, and restore the true phase value of each pixel in sequence according to the two-dimensional spatial coordinates of the pixels. During the unwrapping process, a fixed phase threshold of π is used as the judgment condition for the phase difference between adjacent pixels. After the phase unwrapping operation of all pixels is completed, the true phase values ​​of all pixels are arranged in an orderly manner according to their corresponding two-dimensional spatial coordinates, and finally the continuous phase distribution of the new ceramic material is obtained.

[0030] Based on the fixed optical path parameters preset during structured light projection, each pixel corresponding to a phase value in the continuous phase distribution is converted into a corresponding three-dimensional coordinate point in the three-dimensional space of the ceramic new material. During the conversion process, the spatial relative position between each pixel remains unchanged. The discrete point cloud organization involves classifying and arranging all the converted three-dimensional coordinate points according to the actual spatial distribution pattern of the ceramic new material surface. All three-dimensional coordinate points are meshed at a fixed spatial interval of 0.01mm. After the meshing is completed, all valid three-dimensional coordinate points are retained and an ordered point cloud set is formed, finally obtaining the surface point cloud data of the ceramic new material.

[0031] Delaunay triangulation involves constructing triangular facets from all 3D coordinate points in the surface point cloud data. The sole criterion for construction is that the circumcircle of any triangular facet does not contain any other 3D coordinate points. Based on the spatial distribution of the 3D coordinate points, triangular facets are sequentially constructed for each 3D coordinate point and its adjacent points, ensuring seamless splicing and no overlap. After all triangular facets are constructed, they are spliced ​​together according to the actual spatial morphology of the ceramic material surface, forming a triangular mesh model that perfectly matches the actual surface morphology of the ceramic material. This triangular mesh model is the digital morphological data of the ceramic material.

[0032] The curvature adaptive lighting module is used to perform curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digital topography data, so as to obtain a differentiated lighting strategy for the new ceramic material. In this embodiment of the invention, when the curvature adaptive lighting module performs curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digitized topography data to obtain a differentiated lighting strategy for the new ceramic material, it is specifically used for: The surface normal vector field inversion is performed on the digital morphology data to obtain the normal vector distribution data of the new ceramic material; Based on the normal vector distribution data, the surface of the new ceramic material is segmented into curvature feature regions to obtain the curvature feature regions of the new ceramic material. Based on the principal direction of the normal vector of the triangular facet in the curvature feature region, the incident angle of the light source array of the new ceramic material is calibrated to obtain the incident angle configuration parameters of the new ceramic material. Based on the incident angle configuration parameters, the triggering timing and incident angle of the light sources in the light source array are jointly arranged to obtain the differentiated lighting strategy for the new ceramic material.

[0033] The surface normal vector field inversion is performed on the triangular mesh model in the digital topography data. The normal vector of each triangular facet is calculated based on the spatial position relationship of its three vertices. After the calculation, each normal vector is normalized so that the magnitude of all normal vectors is unified to 1. Then, the normalized normal vectors corresponding to all triangular facets on the surface of the ceramic new material are arranged in an orderly manner according to their spatial coordinate positions in the digital topography data to form a normal vector field data set covering the complete surface of the ceramic new material. This set is the normal vector distribution data of the ceramic new material.

[0034] When segmenting the curvature feature region of a new ceramic material surface based on normal vector distribution data, the curvature value is first calculated for each triangular facet using its normal vector distribution data. The preset curvature value threshold range is 0-0.05mm. -1 0.05-0.2mm -1 Greater than 0.2mm -1 Triangular facets with curvature values ​​within the same threshold range and spatially continuous on the surface of a new ceramic material are grouped into one region. During the segmentation process, it is ensured that there is no overlap between regions and that all triangular facets are assigned to their corresponding regions. Each continuous region that has been segmented is the curvature feature region of the new ceramic material.

[0035] For each curvature feature region, the directional distribution of the normal vectors of all triangular facets within the region is statistically analyzed. The normal vector direction with an occurrence frequency of 90% or higher is taken as the principal direction of the triangular facet normal vectors in that curvature feature region. The incident angle of the light source is calibrated to a fixed angle of 45 degrees with this principal direction. Each curvature feature region corresponds to a unique calibrated incident angle. The position information of all curvature feature regions is matched one-to-one with the corresponding calibrated incident angle and recorded in an orderly manner. The resulting record set is the incident angle configuration parameter of the new ceramic material.

[0036] Based on the fixed order of spatial distribution on the surface of the new ceramic material from left to right and from top to bottom, a unique trigger time node is assigned to the light source group in the light source array corresponding to each curvature feature region. The interval between the trigger time nodes of two adjacent light source groups is set to a fixed duration of 5ms. At the same time, each time node is matched one by one with the corresponding calibrated incident angle in the incident angle configuration parameters. The single exposure duration of each light source group is set to 20ms. The matching information of the trigger time nodes, calibrated incident angles, and exposure durations of all light source groups is integrated and arranged in an orderly manner to form a complete and directly executable lighting configuration scheme, which is the differentiated lighting strategy for the new ceramic material.

[0037] The multi-angle light field acquisition module is used to sequentially trigger multi-angle structured light projection on the surface of the new ceramic material based on the differentiated lighting strategy, and simultaneously perform exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material. In this embodiment of the invention, when the multi-angle light field acquisition module executes the differentiated illumination strategy to sequentially trigger multi-angle structured light projection onto the surface of the new ceramic material, and simultaneously performs exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material, it is specifically used for: Based on the incident angle calibration parameters in the differentiated lighting strategy, the incident angle servo positioning of the light source group in the light source array is performed to obtain the directional light source array of the new ceramic material. Based on the trigger timing arrangement in the differentiated lighting strategy, the directional light source array is sequentially excited to obtain the light source group identification code of the new ceramic material; The reflected light field of the new ceramic material is sampled to obtain the original reflected light field image frame sequence of the new ceramic material; Based on the light source group identifier encoding, the original reflected light field image frame sequence is decoded by the light source group identifier to obtain the reflected light field image frame of the new ceramic material. The reflected light field image frame is mapped to pixel coordinate space to obtain the multi-angle reflected light field image frame of the new ceramic material; The multi-angle reflected light field image frames are serialized and grouped to generate the multi-angle reflected light field image sequence.

[0038] Incident angle servo positioning is based on the preset incident angle calibration parameters in the differentiated lighting strategy. It matches a unique calibration incident angle for each light source group in the light source array. The emission angle of each light source group is adjusted group by group through the servo drive device. During the adjustment process, the deviation between the actual emission angle and the calibration incident angle does not exceed 0.1 degrees as the judgment condition. After the deviation meets the condition, the spatial emission position of the corresponding light source group is locked. After all light source groups in the light source array have completed the angle adjustment and achieved position locking, the light source array formed by the fixed arrangement according to the calibration incident angle is the directional light source array of the new ceramic material.

[0039] According to the fixed order of the trigger timing arrangement set in the differentiated lighting strategy, electrical signal excitation commands are sent sequentially to each light source group in the directional light source array. The interval between the sending of the excitation commands corresponds exactly to the time nodes in the trigger timing arrangement. At the same time as exciting each light source group, a unique 6-digit digital identifier is assigned to it. The digital identifier is continuously incremented according to the excitation order of the light source group. The excitation order of all light source groups is bound to the corresponding 6-digit digital identifier to form an ordered record set. This record set is the light source group identification code of the new ceramic material.

[0040] An industrial area array camera is used to sample the reflected light field formed on the surface of a new ceramic material after structured light projection. The camera's sampling frame rate is set to 200 frames / second to match the trigger interval of the light source group. The single exposure duration of the camera is consistent with the exposure duration of the light source group set in the differentiated illumination strategy. When a single light source group is excited and structured light is projected, the camera synchronously starts a single exposure to complete one light field sampling and generate a single frame image. All the single frame images obtained by continuous sampling are arranged in order according to the actual excitation order of the light source groups in the directional light source array. The resulting image frame set is the original reflected light field image frame sequence of the new ceramic material.

[0041] The 6-digit identifier in the light source group identification code is matched with each frame of the original reflected light field image frame sequence. The matching criteria are that the actual acquisition time of the image and the excitation time of the corresponding light source group are completely coincident. Information processing is performed on each frame of the matched image to remove invalid and redundant identification information generated during the acquisition process, and only the valid pixel data of the image and the corresponding 6-digit identifier of the light source group are retained. After completing the matching and invalid information removal operations of all image frames, the single frame of valid processed image is the reflected light field image frame of the new ceramic material.

[0042] Using the three-dimensional spatial coordinates in the digital morphology data of new ceramic materials as a reference, a unified pixel coordinate spatial mapping coordinate system covering the entire surface of the new ceramic materials is established. The pixel unit of the coordinate system is set to 0.01 mm / pixel. All pixels in each reflected light field image frame are mapped to the unified coordinate system one by one according to the incident angle of their corresponding light source group. During the mapping process, the spatial deviation between the actual mapped position and the theoretical position of the pixel is no more than 1 pixel unit as a judgment condition to ensure that each pixel has a unique corresponding spatial position in the coordinate system. After completing the spatial mapping operation of all pixels, the resulting image frame with unified pixel coordinate spatial information is the multi-angle reflected light field image frame of the new ceramic materials.

[0043] According to the increasing order of the incident angle of the light source group in the directional light source array from 0 degrees to 90 degrees, all multi-angle reflected light field image frames are classified and sorted. During the sorting process, it is ensured that each incident angle interval corresponds to a unique multi-angle reflected light field image frame. All image frames after angle sorting are continuously grouped and numbered, starting from 1 and increasing sequentially, with each number corresponding to a unique image frame. All multi-angle reflected light field image frames that have been numbered are systematically integrated and collected according to the numbering order. The resulting continuous set of image frames with unique numbers is the multi-angle reflected light field image sequence of the new ceramic material.

[0044] The defect pixel screening module is used to perform temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence, and to compare the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain the defect candidate pixels of the new ceramic material. In this embodiment of the invention, when the defect pixel screening module performs temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence and compares the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain the defect candidate pixels of the new ceramic material, it is specifically used for: Pixel coordinate system registration is performed on the image frames in the multi-angle reflected light field image sequence to obtain the spatially registered image sequence of the new ceramic material; The radiance response amplitude at the spatial pixel coordinates in the spatially registered image sequence is extracted in time sequence to obtain the radiance response curve of the new ceramic material; By deconstructing the trajectory morphology features of the radiance response curve, the spatiotemporal fingerprint spectrum of the new ceramic material is obtained; Based on the normal vector distribution field of the digital morphology data, spatial proximity analysis is performed on the spatial pixel coordinates in the spatiotemporal feature fingerprint map to construct the dynamic reference feature fingerprint of the new ceramic material. Based on the dynamic reference feature fingerprint, the spatiotemporal feature fingerprint spectrum is offset and compared to obtain the candidate pixels for defects in the new ceramic material.

[0045] When the defect pixel screening module performs spatial proximity analysis on the spatial pixel coordinates in the spatiotemporal feature fingerprint map based on the normal vector distribution field of the digitized topography data to construct the dynamic reference feature fingerprint of the new ceramic material, it is specifically used for: The spatial search radius is determined based on the surface normal vector of the spatial pixel coordinate position, and a neighborhood window is delineated with the spatial pixel coordinate position as the center and the spatial search radius as the constraint to construct the spatial neighborhood pixel set of the new ceramic material. The normal vector deviation of the pixel coordinates in the spatial neighborhood pixel set is analyzed, and the pixel coordinates are sorted in ascending order according to the obtained vector angle to obtain the candidate reference pixel set of the new ceramic material. The spatiotemporal feature fingerprints of the candidate reference pixel set are fused to obtain the dynamic reference feature fingerprints of the new ceramic material.

[0046] When the defect pixel initial screening module performs the following operations to determine the spatial search radius based on the surface normal vector of the spatial pixel coordinate position, and delineates the neighborhood window with the spatial pixel coordinate position as the center and the spatial search radius as a constraint to construct the spatial neighborhood pixel set of the new ceramic material, it is specifically used for: Obtain the vector angle between the surface normal vector at the spatial pixel coordinate position and the mean normal vector of the global normal vector distribution field in the digitized topography data; Based on the vector angle, the spatial search radius of the spatial pixel coordinate position is dynamically calibrated to obtain the dynamic search radius of the new ceramic material; Centered on the spatial pixel coordinates and constrained by the dynamic search radius, a neighborhood window is delineated in the pixel coordinate space of the spatiotemporal feature fingerprint to obtain the initial neighborhood pixel set of the new ceramic material; The initial neighborhood pixel set is calibrated by a distance weighting factor to obtain the spatial neighborhood pixel set of the new ceramic material.

[0047] When the defect pixel screening module performs deviation comparison of the spatiotemporal feature fingerprint spectrum based on the dynamic reference feature fingerprint to obtain candidate defect pixels of the new ceramic material, it is specifically used for: The comprehensive deviation of the spatial pixel coordinates in the spatiotemporal feature fingerprint is calibrated, and the formula for calculating the comprehensive deviation is as follows: ; In the formula, The spatial pixel coordinates in the spatiotemporal feature fingerprint map are... The comprehensive deviation of the spatial pixel coordinate position. The rising edge steepness feature value is the spatial pixel coordinate position. This refers to the rising edge steepness reference value in the dynamic reference feature fingerprint. The peak response angle feature value is the location of the spatial pixel coordinates. This refers to the peak response angle reference value in the dynamic reference feature fingerprint. The decay rate feature value is the spatial pixel coordinate position. This is the attenuation rate reference value in the dynamic reference feature fingerprint. The oscillation period feature value of the spatial pixel coordinate position is... The oscillation period reference value in the dynamic reference feature fingerprint is used. The preset stability factor, The preset rising edge steepness weighting coefficient, The preset peak response angle weighting coefficient, The preset decay rate weighting coefficient, The weighting coefficients are the preset oscillation period dimension. The preset rising edge steepness index adjustment factor, The preset peak response angle exponential adjustment factor, The preset decay rate exponential adjustment factor, This is the preset oscillation period dimension index adjustment factor.

[0048] Pixel coordinate system registration uses a unified pixel coordinate space mapping coordinate system corresponding to the digital morphology data of the new ceramic material as the sole reference. The pixel unit of this coordinate system is 0.01 mm / pixel. Pixel coordinate correction is performed frame by frame for each image in the multi-angle reflection light field image sequence. During the correction process, the deviation of the pixel coordinate origin, coordinate axis direction, and pixel scale of the image frame from the reference coordinate system is determined by the condition that the deviation value does not exceed 1 pixel unit. After the deviation value meets the condition, the pixel coordinate configuration of the image frame is locked. After all image frames have completed pixel coordinate correction and are completely matched with the reference coordinate system, the resulting image frame sequence is the spatial registration image sequence of the new ceramic material.

[0049] Using the excitation sequence of the light source groups in the directional light source array as the time axis, with the minimum time scale set to 5ms, each unique spatial pixel coordinate position in the spatially registered image sequence is identified one by one. According to the chronological order of the time axis, the radiance response amplitude corresponding to each spatial pixel coordinate position at different time nodes is extracted. During the extraction process, the correlation information between the spatial pixel coordinate and the time node corresponding to each amplitude is retained. All radiance response amplitudes of each spatial pixel coordinate position are arranged continuously according to the time axis and form a continuous line. This line is the radiance response curve of the new ceramic material, with each spatial pixel coordinate position corresponding to a unique radiance response curve.

[0050] Trajectory morphology feature deconstruction involves extracting morphological features from the entire trajectory of each radiance response curve. The extracted morphological features include four core components: steepness of the rising edge, peak response angle, decay rate, and oscillation period. During extraction, the starting acquisition point of the radiance response curve is used as a fixed feature analysis reference point. The rising, peak, decay, and oscillation stages of the curve are calibrated segment by segment, and the specific morphological feature information corresponding to each stage is recorded. The pixel coordinate information of all spatial pixel coordinate positions in the spatially registered image sequence is bound one by one with the corresponding four core morphological feature information. All the bound information is arranged in an orderly manner according to the spatial distribution law of the ceramic new material surface. The complete information set formed is the spatiotemporal feature fingerprint spectrum of the ceramic new material.

[0051] Spatial proximity analysis is based on the normal vector distribution field of digital morphological data. For each independent spatial pixel coordinate position in the spatiotemporal feature fingerprint map, the surface normal vector corresponding to that position in the normal vector distribution field is first obtained. Then, a fixed range of 3×3 pixels is defined as the spatial neighborhood search area centered on the spatial pixel coordinate position. The spatiotemporal feature fingerprint information corresponding to all valid spatial pixel coordinate positions in the search area is extracted. Feature fusion processing is performed on all extracted spatiotemporal feature fingerprint information. During fusion, the mean value of the same morphological feature at each spatial pixel coordinate position is taken as the fused feature value. The mean value of the four core morphological features obtained by fusion is bound to the target spatial pixel coordinate position. The fused feature information set corresponding to each spatial pixel coordinate position is the dynamic reference feature fingerprint of the new ceramic material.

[0052] The four morphological feature values ​​corresponding to each spatial pixel coordinate position in the spatiotemporal feature fingerprint are compared one by one with the average value of the four features fused in the dynamic reference feature fingerprint corresponding to that position. The deviation judgment threshold for each morphological feature is preset to ±5% of the corresponding fused feature average value. If any morphological feature value of a certain spatial pixel coordinate position exceeds the corresponding deviation judgment threshold range, the spatial pixel coordinate position is determined to be a defect candidate pixel. All spatial pixel coordinate positions that meet the defect candidate pixel conditions are recorded in an orderly manner. The resulting set of pixel coordinate positions is the defect candidate pixel of the new ceramic material.

[0053] The surface normal vector corresponding to the spatial pixel coordinate position of the target in the spatiotemporal feature fingerprint spectrum is extracted from the normal vector distribution field of the digitized topography data. At the same time, the vector angle between the surface normal vector and the mean of the global normal vector in the normal vector distribution field is calculated. The vector angle is preset to correspond to a spatial search radius of 3×3 pixels for 0-10 degrees, 5×5 pixels for 10-20 degrees, and 7×7 pixels for greater than 20 degrees. The fixed spatial search radius of the target pixel is determined according to this correspondence rule. Then, a square neighborhood window is delineated with the target spatial pixel coordinate position as the geometric center and the determined spatial search radius as the boundary. The boundary of the window is completely aligned with the pixel grid line of the pixel coordinate system. All valid spatial pixel coordinate positions within the surface range of the ceramic new material are extracted from the window. Invalid pixel coordinates outside the surface range of the ceramic new material are removed. All valid pixel coordinate positions are arranged in a row-major order. The set of pixel coordinate positions formed is the spatial neighborhood pixel set of the ceramic new material.

[0054] The normalized surface normal vector corresponding to each pixel coordinate position in the spatial neighborhood pixel set is extracted from the normal vector distribution field of the digital topography data. At the same time, the normalized surface normal vector corresponding to the target spatial pixel coordinate position is also extracted. The vector angle between the normalized surface normal vector of each neighboring pixel and the normalized surface normal vector of the target pixel is calculated. The calculation result ranges from 0 to 180 degrees. Each pixel coordinate position in the spatial neighborhood pixel set is bound to its corresponding vector angle value. Then, all bound pixel coordinate positions are sorted in ascending order of vector angle values ​​from 0 to 180 degrees. If there are pixel coordinate positions with the same vector angle value, they are sorted in ascending order of row number. If the row numbers are the same, they are sorted in ascending order of column number. The ordered set of pixel coordinate positions formed after the full sorting is the candidate reference pixel set for the new ceramic material.

[0055] The spatiotemporal fingerprint information corresponding to each pixel coordinate position in the candidate reference pixel set is completely extracted from the spatiotemporal fingerprint map. This information includes the specific values ​​of four core morphological features: rising edge steepness, peak response angle, decay rate, and oscillation period. During the extraction process, pixel coordinate positions without valid feature values ​​are removed to ensure that all feature values ​​participating in the fusion are valid data. The arithmetic mean of each of the four core morphological features is calculated independently. All valid values ​​under each feature are summed and divided by the number of values ​​participating in the calculation to obtain the fusion mean corresponding to each feature. The fusion mean of the four core morphological features is integrated and uniquely bound to the pixel coordinate position in the target space. The complete information set containing the pixel coordinate position in the target space and the mean of the four fusion features is the dynamic reference feature fingerprint of the new ceramic material.

[0056] The original surface normal vector corresponding to the target spatial pixel coordinate position is extracted from the normal vector distribution field of the digital topography data. The original surface normal vector is normalized to unify the modulus to 1, resulting in the target normalized normal vector. Then, the original surface normal vectors corresponding to all pixels on the surface of the ceramic new material are extracted from the global normal vector distribution field. After normalizing each original surface normal vector, the arithmetic mean is calculated to obtain the global normalized normal vector mean. The vector angle between the target normalized normal vector and the global normal vector mean is calculated by vector dot product. The angle calculation result is limited to a fixed numerical range of 0 to 180 degrees, thus completing the acquisition of the vector angle corresponding to the target spatial pixel coordinate position.

[0057] The preset vector angle is divided into numerical ranges of 0 to 10 degrees, 10 to 20 degrees, and greater than 20 degrees. A unique pixel-level spatial search radius value is matched to each numerical range. Specifically, 0 to 10 degrees corresponds to a spatial search radius of 3 pixels, 10 to 20 degrees corresponds to a spatial search radius of 5 pixels, and greater than 20 degrees corresponds to a spatial search radius of 7 pixels. The vector angle of the obtained target spatial pixel coordinate position is matched to the corresponding numerical range, and the pixel-level spatial search radius value corresponding to that range is extracted. This value is used as the exclusive search radius of the target spatial pixel coordinate position. This exclusive search radius is the dynamic search radius of the new ceramic material.

[0058] Using the target spatial pixel coordinates as the geometric center and the dynamic search radius as the pixel numerical constraint, a square neighborhood window is delineated in the unified pixel coordinate space corresponding to the spatiotemporal feature fingerprint. The horizontal and vertical side lengths of the window are both 2 times the dynamic search radius plus 1 pixel unit. The four boundaries of the window are perfectly aligned with the pixel grid lines of the pixel coordinate system. After delineation, all spatial pixel coordinates within the window are extracted. Each pixel coordinate is then determined to be within the effective range of the ceramic new material surface. Invalid spatial pixel coordinates that exceed the effective range are removed. The remaining valid spatial pixel coordinates are arranged in a fixed order: in ascending order of row number, and if the row numbers are the same, in ascending order of column number. The resulting set of valid pixel coordinates is the initial neighborhood pixel set of the ceramic new material.

[0059] The Euclidean pixel distance between each valid spatial pixel coordinate position in the initial neighborhood pixel set and the target spatial pixel coordinate position is calculated. This distance is the integer result of the square root of the sum of the squares of the row difference and column difference of the pixel coordinates. A fixed one-to-one correspondence rule between pixel distance and weighting factor is preset, where pixel distance 0 corresponds to weighting factor 1.0, pixel distance 1 corresponds to weighting factor 0.8, pixel distance 2 corresponds to weighting factor 0.6, pixel distance 3 corresponds to weighting factor 0.4, and pixel distance greater than 3 corresponds to weighting factor 0.2. According to this rule, a unique weighting factor is matched for each valid pixel coordinate position in the initial neighborhood pixel set, and the binding operation between the two is completed. The effective judgment threshold of the weighting factor is set to 0.2. Invalid pixel coordinate positions with weighting factor values ​​lower than this threshold are removed. The remaining spatial pixel coordinate positions with bound valid weighting factors are rearranged in the original row priority order. The resulting ordered set containing pixel coordinate positions and corresponding valid weighting factors is the spatial neighborhood pixel set of the ceramic new material.

[0060] Relevant feature data at the corresponding location is directly extracted from the spatiotemporal feature fingerprint map. This location is the spatial pixel coordinate position in the spatiotemporal feature fingerprint map that actually participates in the comprehensive deviation calibration, and it is the corresponding matching benchmark between all feature data and reference data.

[0061] The rising edge steepness feature value is directly extracted from the four core morphological feature data corresponding to the spatial pixel coordinate position in the spatiotemporal feature fingerprint map. It is the morphological feature quantification result of the rising stage of the radiance response curve at the pixel position.

[0062] The rising edge steepness reference value is directly extracted from the dynamic reference feature fingerprint corresponding to the spatial pixel coordinate position, and is the fusion mean quantization result of the rising edge steepness in the dynamic reference feature fingerprint.

[0063] The peak response angle feature value is directly extracted from the four core morphological feature data corresponding to the spatial pixel coordinate position in the spatiotemporal feature fingerprint map. It is the morphological feature quantification result of the peak stage of the radiation brightness response curve at the pixel position.

[0064] The peak response angle reference value is directly extracted from the dynamic reference feature fingerprint corresponding to the spatial pixel coordinate position. It is the fusion mean quantization result of the peak response angle in the dynamic reference feature fingerprint.

[0065] The decay rate feature value is directly extracted from the four core morphological feature data corresponding to the spatial pixel coordinate position in the spatiotemporal feature fingerprint map. It is the morphological feature quantification result of the decay stage of the radiance response curve at that pixel position.

[0066] The decay rate reference value is directly extracted from the dynamic reference feature fingerprint corresponding to the spatial pixel coordinate position. It is the fusion mean quantization result of the decay rate in the dynamic reference feature fingerprint.

[0067] The oscillation period feature value is directly extracted from the four core morphological feature data corresponding to the spatial pixel coordinate position in the spatiotemporal feature fingerprint spectrum. It is the morphological feature quantification result of the oscillation stage of the radiation brightness response curve at that pixel position.

[0068] The oscillation period reference value is directly extracted from the dynamic reference feature fingerprint corresponding to the spatial pixel coordinate position. It is the fusion mean quantization result of the oscillation period in the dynamic reference feature fingerprint.

[0069] The stability factor is a pre-set fixed value. The purpose of setting this value is to avoid meaningless calculations caused by the feature reference value being zero, and to keep it constant throughout the entire process of defect identification in new ceramic materials.

[0070] The rising edge steepness weighting coefficient is a pre-set fixed value. This value is set according to the characteristic importance of the rising edge steepness dimension in the defect identification of new ceramic materials, and remains fixed throughout the entire process of defect identification of new ceramic materials.

[0071] The peak response angle weighting coefficient is a fixed value that is set in advance. This value is set according to the characteristic importance of the peak response angle dimension in the defect identification of new ceramic materials and remains fixed throughout the entire process of defect identification of new ceramic materials.

[0072] The attenuation rate weighting coefficient is a pre-set fixed value. This value is set according to the characteristic importance of the attenuation rate dimension in the defect identification of new ceramic materials and remains fixed throughout the entire process of defect identification of new ceramic materials.

[0073] The weighting coefficient for the oscillation period dimension is a fixed value that is set in advance. This value is set according to the characteristic importance of the oscillation period dimension in the defect identification of new ceramic materials and remains fixed throughout the entire process of defect identification of new ceramic materials.

[0074] The rising edge steepness index adjustment factor is a pre-set fixed value. This value is used to adjust the degree of deviation of the rising edge steepness dimension from the calculation result and remains constant throughout the entire process of defect identification in new ceramic materials.

[0075] The peak response angle index adjustment factor is a pre-set fixed value. This value is used to adjust the degree of deviation of the peak response angle dimension from the calculation result and remains constant throughout the entire process of defect identification of new ceramic materials.

[0076] The attenuation rate index adjustment factor is a pre-set fixed value. This value is used to adjust the degree of deviation of the attenuation rate dimension from the calculation result and remains constant throughout the entire process of defect identification in new ceramic materials.

[0077] The oscillation period dimension index adjustment factor is a pre-set fixed value. This value is used to adjust the degree of change of the oscillation period dimension from the calculation result, and remains constant throughout the entire process of defect identification of new ceramic materials.

[0078] The core significance of this calculation is to accurately calibrate the comprehensive deviation of any spatial pixel coordinate position in the spatiotemporal feature fingerprint spectrum. This is achieved by calculating the deviation of the actual values ​​of the four core morphological features at that position and the reference values ​​of the corresponding dynamic reference feature fingerprints in one dimension. Then, the deviation calculation results of each dimension are weighted and allocated using preset weight coefficients. At the same time, the magnitude of the deviation results after weighting each dimension is adjusted using preset exponential adjustment factors. A stabilization factor participates in the calculation of deviation in each dimension to ensure the effectiveness of the calculation process. Finally, the deviation results of all dimensions after weighting and magnitude adjustment are accumulated to obtain the comprehensive deviation quantification result of the spatial pixel coordinate position.

[0079] The comprehensive deviation quantification result is the core quantitative basis for determining whether the spatial pixel coordinate position is a defect candidate pixel. The magnitude of the comprehensive deviation directly reflects the overall deviation between the feature data of the spatial pixel coordinate position and the reference feature data of the normal ceramic new material surface pixels. The larger the value, the higher the deviation. Combined with the pre-set comprehensive deviation judgment threshold, the corresponding spatial pixel coordinate position can be accurately judged to see whether it meets the judgment conditions of defect candidate pixels. This provides an accurate and quantitative pixel-level judgment basis for the subsequent extraction of defect candidate regions and the acquisition of defect recognition results in ceramic new materials.

[0080] The defect region extraction module is used to perform morphological closure analysis on the spatiotemporal response map of the defect candidate pixels to obtain the defect candidate regions of the new ceramic material. In this embodiment of the invention, when the defect region extraction module performs morphological closure analysis on the spatiotemporal response map of the defect candidate pixels to obtain the defect candidate region of the new ceramic material, it is specifically used for: The spatial distribution field of the spatiotemporal response map of the defect candidate pixels is reconstructed to obtain the spatial distribution field data of the new ceramic material; Based on the spatial distribution field data, the spatiotemporal response spectrum of the defect candidate pixels is dynamically constructed using structural elements to obtain the dynamic structural elements of the new ceramic material. Based on the dynamic structural element, morphological dilation analysis is performed on the pixel coordinate positions in the spatiotemporal response map. When there are defect candidate pixels within the coverage area of ​​the dynamic structural element, the pixel coordinate positions are marked as dilated domain pixels to obtain the dilated domain connectivity map of the new ceramic material. Based on the dynamic structural element, morphological erosion analysis is performed on the expansion domain connectivity graph. When the entire coverage area of ​​the dynamic structural element is an expansion domain pixel, the expansion domain pixel label at the pixel coordinate position is retained to obtain the closed domain connectivity graph of the new ceramic material. The connected domain boundaries of the closed domain connected graph are extracted, and the pixel coordinate range is calibrated according to the obtained edge contour to obtain the defect candidate region of the new ceramic material.

[0081] Using the unified pixel coordinate space of the new ceramic material as a benchmark, with a pixel unit of 0.01 mm / pixel, the spatiotemporal response map of defect candidate pixels is precisely calibrated pixel by pixel. The spatiotemporal response feature data corresponding to each defect candidate pixel is uniquely bound to its pixel coordinates. Then, in a fixed order of ascending row number of pixel coordinates and ascending column number if the row numbers are the same, all bound coordinates and feature data are arranged in a grid. During the arrangement process, blank spatiotemporal response feature data is marked for the coordinate positions of defect-free candidate pixels in the map. The final gridded information set covering the entire surface of the new ceramic material and containing all pixel coordinates and corresponding spatiotemporal response feature data is the spatial distribution field data of the new ceramic material.

[0082] Based on spatial distribution field data, the spatial distribution state of defect candidate pixels is statistically analyzed one by one to accurately obtain the maximum row pixel interval and the maximum column pixel interval of defect candidate pixels in the spatial distribution. Taking the maximum value of the two interval values ​​as the base value, the side length of the structural element is set to the base value plus 2 pixel units, and the structural element is in the form of a square pixel matrix. All pixel nodes in the matrix are nodes that can effectively participate in morphological analysis. Then, according to the spatial distribution density of defect candidate pixels in the spatial distribution field data, the nodes of the square pixel matrix are uniformly arranged to ensure that the distribution of matrix nodes matches the spatial distribution characteristics of defect candidate pixels. The square pixel matrix with fixed side length and uniform node distribution formed after processing is the dynamic structural element of the new ceramic material.

[0083] The geometric center of the dynamic structural element is aligned with each pixel coordinate in the spatiotemporal response map one by one. During the alignment process, the direction of the pixel matrix of the dynamic structural element is kept completely consistent with the direction of the coordinate axis of the pixel coordinate system. For each aligned pixel coordinate, the presence of at least one defect candidate pixel is detected within the overall coverage of the pixel matrix of the dynamic structural element. If the detection result is that it exists, the pixel coordinate of the aligned position is marked as an expansion domain pixel and a unique expansion domain pixel identifier is added to it. If the detection result is that it does not exist, no labeling operation is performed on the pixel coordinate. After the detection and labeling of all pixel coordinate positions in the spatiotemporal response map are completed, all positions marked as expansion domain pixels are arranged in order according to their pixel coordinates. The resulting spatial map containing expansion domain pixel coordinates and corresponding unique identifiers is the expansion domain connectivity map of the new ceramic material.

[0084] The geometric center of the dynamic structural element is aligned with the coordinates of each pixel in the expansion domain connectivity graph. The alignment standard is completely consistent with the alignment standard in the morphological expansion analysis. For each aligned pixel coordinate, the coordinates of all pixels within the coverage area of ​​the dynamic structural element pixel matrix are verified to ensure that they are all labeled as expansion domain pixels. During the verification process, the identification information of each pixel position is confirmed one by one. If the verification result shows that all pixels within the coverage area are expansion domain pixels, the expansion domain pixel label and exclusive identifier of the aligned position are retained. If the verification result shows that there are non-expansion domain pixels within the coverage area, the expansion domain pixel label and exclusive identifier of the aligned position are deleted. After the verification and labeling of all expansion domain pixel positions in the expansion domain connectivity graph are completed, all retained labeled expansion domain pixel positions are arranged in order according to their pixel coordinates. The resulting spatial map containing the retained labeled pixel coordinates and corresponding exclusive identifiers is the closed domain connectivity graph of the new ceramic material.

[0085] Connectivity identification is performed on all labeled dilated domain pixels in the closed domain connectivity graph. During identification, the four neighboring pixels (up, down, left, and right) are used as fixed connectivity criteria. That is, if the row number or column number of two adjacent pixels differs by 1 pixel, they are considered connected. All interconnected dilated domain pixels are grouped into an independent connected domain. For each independent connected domain, the pixel coordinates of its edge contour are extracted point by point. During extraction, only the outermost pixel coordinates of the connected domain are retained. Then, extreme value statistics are performed on the extracted edge contour pixel coordinates to obtain the maximum row number, minimum row number, maximum column number, and minimum column number corresponding to each connected domain. Using these four extreme values ​​as boundaries, a rectangular pixel coordinate range that can completely enclose all pixel positions of the corresponding independent connected domain is marked. Each independent connected domain corresponds to a unique marked rectangular pixel coordinate range. The set formed by all marked rectangular pixel coordinate ranges is the candidate defect region of the new ceramic material.

[0086] The defect intelligent identification module is used to measure the similarity between the feature parameters of the defect candidate region and the historical defect samples of the new ceramic material, so as to obtain the defect identification result of the new ceramic material.

[0087] In this embodiment of the invention, when the defect intelligent recognition module performs a similarity measurement between the feature parameters of the defect candidate region and historical defect samples of the new ceramic material to obtain the defect recognition result of the new ceramic material, it is specifically used for: Based on the geometric morphological feature parameters of the feature parameter map in the defect candidate region, the historical defect sample library of the new ceramic material is selected to obtain the dynamic historical defect sample set of the new ceramic material. The feature parameters of the historical defect samples in the dynamic historical defect sample set are calibrated to obtain the dynamic defect type feature map of the new ceramic material; Based on the dynamic defect type feature map, the similarity of the feature parameter map is measured in terms of rising edge steepness, peak response angle, decay rate, oscillation period and geometric shape, so as to obtain the similarity index distribution of the new ceramic material. By comprehensively evaluating the distribution of the similarity index, the defect identification results of the new ceramic material are obtained.

[0088] Geometric morphological feature parameters are extracted from the feature parameter map of the defect candidate region. These parameters include four core components: pixel area, contour perimeter, aspect ratio, and shape factor. Each core component is precisely quantified. Then, the historical defect sample library of ceramic new materials is retrieved, and the quantified value of the geometric morphological feature parameter corresponding to each historical defect sample in the library is extracted. The deviation between the four core parameters of the historical defect sample and the corresponding parameters of the defect candidate region is calculated one by one. The deviation calculation method is the ratio of the absolute value of the parameter difference to the corresponding parameter value of the defect candidate region. The preset deviation judgment threshold is 20%. If the deviation results of the four core parameters of a historical defect sample do not exceed this threshold, the sample is included in the selection range. After the screening of all historical defect samples in the library is completed, all samples included in the selection range are arranged in a fixed order from low to high according to the total geometric morphological parameter deviation. The resulting ordered set of historical defect samples is the dynamic historical defect sample set of ceramic new materials.

[0089] Each historical defect sample in the dynamic historical defect sample set is uniquely labeled with a defect type consistent with common defect types in the production of new ceramic materials. Then, the quantitative values ​​of four optical feature parameters (rise edge steepness, peak response angle, decay rate, and oscillation period) and four geometric feature parameters (pixel area, contour perimeter, aspect ratio, and shape factor) are extracted for each historical defect sample. For all historical defect samples of the same defect type, the arithmetic mean of each optical feature parameter and the arithmetic mean of each geometric feature parameter are calculated. Each defect type is uniquely bound to the mean of the corresponding five dimensions of feature parameters (rise edge steepness, peak response angle, decay rate, oscillation period, and geometric shape). The geometric shape dimension is comprehensively represented by the mean of its four core parameters. All bound information is then arranged in a grid according to the defect type classification order. The resulting atlas of information containing defect types and the mean of the corresponding five dimensions of feature parameters is the dynamic defect type feature atlas of new ceramic materials.

[0090] The feature parameter quantization values ​​of four optical dimensions—rise edge steepness, peak response angle, decay rate, and oscillation period—are extracted from the feature parameter spectrum of the defect candidate region, along with the comprehensive characterization parameter quantization value of the geometric morphology dimension. Then, the mean values ​​of five feature parameters corresponding to each defect type are extracted from the feature spectrum of dynamic defect types. The similarity is calculated for each dimension parameter value of the defect candidate region and the mean value of the corresponding dimension of each defect type. The calculation method is 1 minus the absolute value of the parameter difference and the ratio of the mean value of the corresponding dimension of the defect type. If the calculation result is less than 0, 0 is directly taken as the similarity value of that dimension, ensuring that the similarity value of each dimension is fixed in the range of 0 to 1. After completing the similarity calculation of the five dimensions between the defect candidate region and each defect type, each defect type is bound to the corresponding five dimension similarity values ​​one by one. All the bound similarity information is arranged in an orderly manner according to the order of defect types. The complete set containing defect types and corresponding similarity indicators of each dimension is the similarity index distribution of ceramic new materials.

[0091] The arithmetic mean of the five-dimensional similarity values ​​corresponding to each defect type in the similarity index distribution is calculated to obtain the comprehensive similarity value corresponding to each defect type. The preset comprehensive similarity judgment threshold is 0.8. The comprehensive similarity value of all defect types is compared with this threshold one by one. If the comprehensive similarity value of a certain defect type is not lower than 0.8, the defect type is directly determined as the defect identification result of the new ceramic material. If the comprehensive similarity value of all defect types is lower than 0.8, the unknown defect type is determined as the defect identification result of the new ceramic material. At the same time, the comprehensive similarity values ​​of all defect types and the finally determined defect types are systematically integrated and recorded. The complete information set containing the judgment result and related similarity data is the defect identification result of the new ceramic material.

[0092] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0093] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A defect identification system for the production of new ceramic materials, characterized in that, The system includes a 3D topography reconstruction module, a curvature adaptive illumination module, a multi-angle light field acquisition module, a defect pixel initial screening module, a defect region extraction module, and a defect intelligent recognition module, wherein: The three-dimensional morphology reconstruction module is used to perform structured light three-dimensional scanning on the surface of the new ceramic material and to reconstruct the collected surface point cloud data into a triangular mesh to obtain the digital morphology data of the new ceramic material. The curvature adaptive lighting module is used to perform curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digital topography data, so as to obtain a differentiated lighting strategy for the new ceramic material. The multi-angle light field acquisition module is used to sequentially trigger multi-angle structured light projection on the surface of the new ceramic material based on the differentiated lighting strategy, and simultaneously perform exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material. The defect pixel screening module is used to perform temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence, and to compare the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain the defect candidate pixels of the new ceramic material. The defect region extraction module is used to perform morphological closure analysis on the spatiotemporal response map of the defect candidate pixels to obtain the defect candidate regions of the new ceramic material. The defect intelligent identification module is used to measure the similarity between the feature parameters of the defect candidate region and the historical defect samples of the new ceramic material, so as to obtain the defect identification result of the new ceramic material.

2. The defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the three-dimensional topography reconstruction module performs structured light three-dimensional scanning on the surface of the new ceramic material and reconstructs triangular meshes from the collected surface point cloud data to obtain the digital topography data of the new ceramic material, it is specifically used for: Based on a coded structured light generation strategy, a high-frequency structured light field is projected onto the surface of a new ceramic material to obtain a composite grating pattern of the new ceramic material. High-frequency image sequence acquisition is performed on the distorted structured light field of the composite grating pattern to obtain the distorted structured light image sequence of the new ceramic material; Based on the phase wrapping value in the distorted structured light image sequence, the distorted structured light image sequence is phase demodulated to obtain the continuous phase distribution of the new ceramic material; The continuous phase distribution is mapped to the three-dimensional spatial coordinates of the new ceramic material, and the three-dimensional spatial coordinates are discretely organized into point cloud to obtain the surface point cloud data of the new ceramic material. The surface point cloud data is subjected to Delaunay triangulation to obtain the digital morphology data of the new ceramic material.

3. The defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the curvature adaptive lighting module performs curvature adaptive mapping on the light source array of the new ceramic material based on the normal vector distribution of the digitized topography data to obtain a differentiated lighting strategy for the new ceramic material, it is specifically used for: The surface normal vector field inversion is performed on the digital morphology data to obtain the normal vector distribution data of the new ceramic material; Based on the normal vector distribution data, the surface of the new ceramic material is segmented into curvature feature regions to obtain the curvature feature regions of the new ceramic material. Based on the principal direction of the normal vector of the triangular facet in the curvature feature region, the incident angle of the light source array of the new ceramic material is calibrated to obtain the incident angle configuration parameters of the new ceramic material. Based on the incident angle configuration parameters, the triggering timing and incident angle of the light sources in the light source array are jointly arranged to obtain the differentiated lighting strategy for the new ceramic material.

4. The defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the multi-angle light field acquisition module executes the differentiated illumination strategy to sequentially trigger multi-angle structured light projection onto the surface of the new ceramic material, and simultaneously performs exposure capture after projection to obtain a multi-angle reflected light field image sequence of the new ceramic material, it is specifically used for: Based on the incident angle calibration parameters in the differentiated lighting strategy, the incident angle servo positioning of the light source group in the light source array is performed to obtain the directional light source array of the new ceramic material. Based on the trigger timing arrangement in the differentiated lighting strategy, the directional light source array is sequentially excited to obtain the light source group identification code of the new ceramic material; The reflected light field of the new ceramic material is sampled to obtain the original reflected light field image frame sequence of the new ceramic material; Based on the light source group identifier encoding, the original reflected light field image frame sequence is decoded by the light source group identifier to obtain the reflected light field image frame of the new ceramic material. The reflected light field image frame is mapped to pixel coordinate space to obtain the multi-angle reflected light field image frame of the new ceramic material; The multi-angle reflected light field image frames are serialized and grouped to generate the multi-angle reflected light field image sequence.

5. The defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the defect pixel screening module performs temporal feature analysis on the same pixel coordinate positions in the multi-angle reflected light field image sequence and compares the obtained reflection intensity change trajectory with the standard substrate reflection response benchmark of the normal vector distribution to obtain candidate defect pixels of the new ceramic material, it is specifically used for: Pixel coordinate system registration is performed on the image frames in the multi-angle reflected light field image sequence to obtain the spatially registered image sequence of the new ceramic material; The radiance response amplitude at the spatial pixel coordinates in the spatially registered image sequence is extracted in time sequence to obtain the radiance response curve of the new ceramic material; By deconstructing the trajectory morphology features of the radiance response curve, the spatiotemporal fingerprint spectrum of the new ceramic material is obtained; Based on the normal vector distribution field of the digital morphology data, spatial proximity analysis is performed on the spatial pixel coordinates in the spatiotemporal feature fingerprint map to construct the dynamic reference feature fingerprint of the new ceramic material. Based on the dynamic reference feature fingerprint, the spatiotemporal feature fingerprint spectrum is offset and compared to obtain the candidate pixels for defects in the new ceramic material.

6. The defect identification system for the production of new ceramic materials as described in claim 5, characterized in that, When the defect pixel screening module performs spatial proximity analysis on the spatial pixel coordinates in the spatiotemporal feature fingerprint map based on the normal vector distribution field of the digitized topography data to construct the dynamic reference feature fingerprint of the new ceramic material, it is specifically used for: The spatial search radius is determined based on the surface normal vector of the spatial pixel coordinate position, and a neighborhood window is delineated with the spatial pixel coordinate position as the center and the spatial search radius as the constraint to construct the spatial neighborhood pixel set of the new ceramic material. The normal vector deviation of the pixel coordinates in the spatial neighborhood pixel set is analyzed, and the pixel coordinates are sorted in ascending order according to the obtained vector angle to obtain the candidate reference pixel set of the new ceramic material. The spatiotemporal feature fingerprints of the candidate reference pixel set are fused to obtain the dynamic reference feature fingerprints of the new ceramic material.

7. The defect identification system for the production of new ceramic materials as described in claim 6, characterized in that, When the defect pixel initial screening module performs the following operations to determine the spatial search radius based on the surface normal vector of the spatial pixel coordinate position, and delineates the neighborhood window with the spatial pixel coordinate position as the center and the spatial search radius as a constraint to construct the spatial neighborhood pixel set of the new ceramic material, it is specifically used for: Obtain the vector angle between the surface normal vector at the spatial pixel coordinate position and the mean normal vector of the global normal vector distribution field in the digitized topography data; Based on the vector angle, the spatial search radius of the spatial pixel coordinate position is dynamically calibrated to obtain the dynamic search radius of the new ceramic material; Centered on the spatial pixel coordinates and constrained by the dynamic search radius, a neighborhood window is delineated in the pixel coordinate space of the spatiotemporal feature fingerprint to obtain the initial neighborhood pixel set of the new ceramic material; The initial neighborhood pixel set is calibrated by a distance weighting factor to obtain the spatial neighborhood pixel set of the new ceramic material.

8. The defect identification system for the production of new ceramic materials as described in claim 5, characterized in that, When the defect pixel screening module performs deviation comparison of the spatiotemporal feature fingerprint spectrum based on the dynamic reference feature fingerprint to obtain candidate defect pixels of the new ceramic material, it is specifically used for: The comprehensive deviation of the spatial pixel coordinates in the spatiotemporal feature fingerprint is calibrated, and the formula for calculating the comprehensive deviation is as follows: ； In the formula, The spatial pixel coordinates in the spatiotemporal feature fingerprint map are... The comprehensive deviation of the spatial pixel coordinate position. The rising edge steepness feature value is the spatial pixel coordinate position. This refers to the rising edge steepness reference value in the dynamic reference feature fingerprint. The peak response angle feature value is the location of the spatial pixel coordinates. This refers to the peak response angle reference value in the dynamic reference feature fingerprint. The decay rate feature value is the spatial pixel coordinate position. This is the attenuation rate reference value in the dynamic reference feature fingerprint. The oscillation period feature value of the spatial pixel coordinate position is... The oscillation period reference value in the dynamic reference feature fingerprint is used. The preset stability factor, The preset rising edge steepness weighting coefficient, The preset peak response angle weighting coefficient, The preset decay rate weighting coefficient, The weighting coefficients are the preset oscillation period dimension. The preset rising edge steepness index adjustment factor, The preset peak response angle exponential adjustment factor, The preset decay rate exponential adjustment factor, This is the preset oscillation period dimension index adjustment factor.

9. A defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the defect region extraction module performs morphological closure analysis on the spatiotemporal response map of the defect candidate pixels to obtain the defect candidate regions of the new ceramic material, it is specifically used for: The spatial distribution field of the spatiotemporal response map of the defect candidate pixels is reconstructed to obtain the spatial distribution field data of the new ceramic material; Based on the spatial distribution field data, the spatiotemporal response spectrum of the defect candidate pixels is dynamically constructed using structural elements to obtain the dynamic structural elements of the new ceramic material. Based on the dynamic structural element, morphological dilation analysis is performed on the pixel coordinate positions in the spatiotemporal response map. When there are defect candidate pixels within the coverage area of ​​the dynamic structural element, the pixel coordinate positions are marked as dilated domain pixels to obtain the dilated domain connectivity map of the new ceramic material. Based on the dynamic structural element, morphological erosion analysis is performed on the expansion domain connectivity graph. When the entire coverage area of ​​the dynamic structural element is an expansion domain pixel, the expansion domain pixel label at the pixel coordinate position is retained to obtain the closed domain connectivity graph of the new ceramic material. The connected domain boundaries of the closed domain connected graph are extracted, and the pixel coordinate range is calibrated according to the obtained edge contour to obtain the defect candidate region of the new ceramic material.

10. A defect identification system for the production of new ceramic materials as described in claim 1, characterized in that, When the defect intelligent identification module performs a similarity measurement between the feature parameters of the defect candidate region and historical defect samples of the new ceramic material to obtain the defect identification result of the new ceramic material, it is specifically used for: Based on the geometric morphological feature parameters of the feature parameter map in the defect candidate region, the historical defect sample library of the new ceramic material is selected to obtain the dynamic historical defect sample set of the new ceramic material. The feature parameters of the historical defect samples in the dynamic historical defect sample set are calibrated to obtain the dynamic defect type feature map of the new ceramic material; Based on the dynamic defect type feature map, the similarity of the feature parameter map is measured in terms of rising edge steepness, peak response angle, decay rate, oscillation period and geometric shape, so as to obtain the similarity index distribution of the new ceramic material. By comprehensively evaluating the distribution of the similarity index, the defect identification results of the new ceramic material are obtained.

Images