Silica sol shell production defect intelligent detection method and system

Abstract

The application provides a silica sol shell production defect intelligent detection method and system, and then extracts the curvature distribution characteristics and local geometric deviation of the shell from the point cloud data of the shell; the suspected defect area of the shell is identified according to the curvature distribution characteristics, and the material uniformity index of the shell is calculated according to the echo intensity sequence of the suspected defect area; the multi-scale roughness characteristics of the shell in the suspected defect area are determined through the material uniformity index and the local geometric deviation, the defect confidence atlas of the shell is constructed based on the multi-scale roughness characteristics, and the plurality of defect positions of the shell are located; the defect depth grade of each defect position is determined according to the attenuation degree of the echo intensity sequence at each defect position, and then the defect score of the shell is output, and a three-dimensional defect distribution visualization report of the shell production is generated. Based on the above scheme, the joint discrimination of geometric morphology and material attenuation characteristics can be realized.

Description

Technical Field

[0001] This application relates to the field of silica sol manufacturing technology, and more specifically, to a method and system for intelligent detection of defects in finished silica sol shell products. Background Technology

[0002] Silica sol shell casting is a precision casting process that uses silica sol as a binder to form a shell by coating multiple layers of refractory material. The finished product made by silica sol shell casting has the characteristics of high surface smoothness, good dimensional stability and high temperature resistance, and is used in the mass production of complex thin-walled castings for aerospace and medical devices.

[0003] Silica sol-formed shells possess rich free-form surface features and dramatic local curvature gradients. Their surfaces simultaneously exhibit various defects, including open cracks, subsurface shrinkage, and subsurface porosity. Traditional inspection methods, such as contact coordinate measuring machines (CMMs), rely on discrete point-by-point acquisition, taking tens of minutes per piece and failing to cover the entire surface area. Manual caliper measurements are only suitable for regular areas and are largely inaccessible to depressions or protrusions in areas of abrupt curvature changes. Ordinary visual inspection, limited by visible light imaging mechanisms, can only detect open defects with significant grayscale differences from the background. Shrinkage and internal pores located below the shell surface and covered by a thin shell layer cannot be detected due to the lack of significant change in reflectivity. This results in numerous subsurface defects becoming blind spots, directly impacting the yield and structural safety of subsequent casting processes. Therefore, achieving joint discrimination based on geometric morphology and material attenuation characteristics to improve the detection rate of surface defects in silica sol-formed shells has become a challenging problem for the industry. Summary of the Invention

[0004] This application provides an intelligent detection method and system for defects in silica sol shell-made products, which can achieve joint discrimination of geometric morphology and material attenuation characteristics, thereby improving the detection rate of surface defects in silica sol shell-made products.

[0005] In a first aspect, this application provides an intelligent detection method for defects in finished products made from silica sol shells, including: A three-dimensional laser scan is performed on the surface of the finished shell made of silica sol to obtain point cloud data of the shell, and then the curvature distribution characteristics and local geometric deviations of the shell are extracted from the point cloud data. Based on the curvature distribution characteristics, suspected defect areas of the shell are identified, and echo intensity sequences of the shell are collected within the suspected defect areas. The material uniformity index of the shell is calculated based on the echo intensity sequences. The multi-scale roughness characteristics of the shell in the suspected defect area are determined by the material uniformity index and the local geometric deviation. Based on the multi-scale roughness characteristics, a defect confidence map of the shell is constructed, and the locations of multiple defects in the shell are located. The defect depth level at each defect location is determined based on the attenuation degree of the echo intensity sequence at each defect location, and then the defect score of the shell is output, and a three-dimensional defect distribution visualization report of the finished shell is generated.

[0006] In some embodiments, extracting the curvature distribution features and local geometric deviations of the shell from the point cloud data specifically includes: The point cloud data is subjected to voxel filtering for noise reduction, and then the normal vector and curvature value of each point in the point cloud data are calculated to obtain the curvature distribution characteristics of all points. The standard 3D model of the shell and the point cloud data are registered, and the normal distance from each point cloud point to the surface of the standard model is calculated to obtain the local geometric deviation.

[0007] In some embodiments, identifying suspected defect areas of the shell based on the curvature distribution characteristics specifically includes: Points with curvature values ​​exceeding a preset threshold are marked as high curvature points; Euclidean distance clustering is performed on the high curvature points to obtain multiple candidate regions; Candidate regions with an area smaller than the minimum defect size are removed, and the remaining regions are retained as suspected defect regions.

[0008] In some embodiments, calculating the material uniformity index of the shell based on the echo intensity sequence specifically includes: Calculate the mean and standard deviation of the echo intensity sequence within the suspected defect area; Using a defect-free area on the shell surface as a reference area, the baseline value of the echo intensity of the reference area is calculated. The material uniformity index of the shell is obtained based on the echo intensity benchmark value, mean, and standard deviation.

[0009] In some embodiments, determining the multi-scale roughness characteristics of the shell in the suspected defect region using the material uniformity index and the local geometric deviation specifically includes: Multiple spherical neighborhoods with different radii are set within the suspected defect area; The change in the normal vector angle of the point cloud within each spherical neighborhood is calculated based on the local geometric deviation. Using the material uniformity index as a weighting coefficient, the changes in the angles of each normal vector are weighted and fused to obtain the multi-scale roughness characteristics of the shell in the suspected defect area.

[0010] In some embodiments, constructing a defect confidence map of the shell based on the multi-scale roughness features and locating multiple defect locations of the shell specifically includes: The multi-scale roughness features are spatially rasterized in the shell point cloud coordinate system, and the average roughness value within each grid is calculated. Rasteres with average roughness values ​​greater than the confidence threshold are marked as high-confidence defect rasters, and adjacent high-confidence defect rasters are merged by region growing to obtain multiple defect locations. Map the confidence values ​​of all defect locations to color labels to generate a defect confidence map.

[0011] In some embodiments, determining the defect depth level at each defect location based on the attenuation degree of the echo intensity sequence at each defect location specifically includes: The attenuation coefficient at each defect location is determined based on the minimum intensity value of the echo intensity sequence at each defect location and the background intensity value. Each attenuation coefficient is compared with multiple preset depth level threshold ranges to obtain the initial depth level of each defect location; The initial depth levels are corrected based on the material uniformity index of the neighborhood of each defect location, and the defect depth level of each defect location is output.

[0012] Secondly, this application provides an intelligent detection system for defects in finished silica sol shell products, comprising: The acquisition module is used to perform three-dimensional laser scanning on the surface of the finished product made of silica sol to obtain point cloud data of the shell, and then extract the curvature distribution characteristics and local geometric deviations of the shell from the point cloud data; The processing module is used to identify suspected defect areas of the shell based on the curvature distribution characteristics, collect the echo intensity sequence of the shell in the suspected defect areas, and calculate the material uniformity index of the shell based on the echo intensity sequence. The processing module is also used to determine the multi-scale roughness characteristics of the shell in the suspected defect area through the material uniformity index and the local geometric deviation, construct the defect confidence map of the shell based on the multi-scale roughness characteristics, and locate multiple defect locations of the shell. The execution module is used to determine the defect depth level of each defect location based on the attenuation degree of the echo intensity sequence at each defect location, and then output the defect score of the shell and generate a three-dimensional defect distribution visualization report of the finished shell.

[0013] Thirdly, this application provides a computer device, which includes a memory and a processor. The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device performs the above-described intelligent detection method for defects in silica sol shell-making products.

[0014] Fourthly, this application provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the aforementioned intelligent detection method for defects in silica sol shell-making products.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: This application provides an intelligent detection method and system for defects in silica sol-formed shells. The method involves performing a three-dimensional laser scan on the surface of the finished silica sol-formed shell to obtain point cloud data of the shell. Curvature distribution features and local geometric deviations of the shell are then extracted from the point cloud data. Suspected defect areas of the shell are identified based on the curvature distribution features. Echo intensity sequences of the shell are collected within these suspected defect areas, and a material uniformity index is calculated based on the echo intensity sequences. Multi-scale roughness features of the shell within the suspected defect areas are determined using the material uniformity index and the local geometric deviations. A defect confidence map of the shell is constructed based on these multi-scale roughness features, and multiple defect locations are located. The defect depth level at each defect location is determined based on the attenuation degree of the echo intensity sequence at each defect location. Finally, a defect score for the shell is output, and a three-dimensional defect distribution visualization report of the finished shell is generated.

[0016] Therefore, in this application, the defect depth level at each defect location is determined based on the attenuation degree of the echo intensity sequence at each defect location, thereby outputting the defect score of the shell and generating a three-dimensional defect distribution visualization report of the finished shell. First, by determining the material uniformity index, the degree to which the material reflection characteristics in the suspected defect area deviate from the normal substrate can be obtained, thus providing a basis for material consistency discrimination for areas with abnormal geometric morphology. In the detection logic that relies solely on curvature distribution, surface stains, residual mold release agents, or process textures are often misjudged as defects due to local curvature abrupt changes. The introduction of the material uniformity index expands the detection dimension from a single geometry to the material property level. This material uniformity index transforms the statistical differences of the echo intensity sequence into weighting factors that can participate in subsequent fusion calculations in the form of normalized numerical values, enabling the two types of heterogeneous data, geometric morphology and material attenuation, to be compared and jointly analyzed within the same framework, eliminating misjudgments caused by surface contamination or casting flow marks, and improving the specificity of defect identification. Then, by determining the defect location, the following can be obtained: The spatial coordinate set on the shell surface that simultaneously satisfies both geometric curvature anomalies and material attenuation anomalies allows for the precise convergence of potential defects from coarse-grained region-level localization to point-level spatial localization. By introducing multi-scale roughness features to construct a defect confidence map, and performing spatial rasterization and region growing and merging on the defect confidence map, the determination of defect location no longer depends on a fixed curvature threshold or a manually set clustering radius, but is based on the spatial continuity of confidence values ​​for segmentation. Each output defect location corresponds to a geometric center of a connected high-confidence region verified by the material uniformity index. This means that the defect location satisfies both the local geometric distortion condition and the material attenuation condition, eliminating purely geometric distortion points and purely material anomaly points, retaining only those jointly identified as real physical defects. This improves the accuracy and repeatability of the localization results and provides a spatial benchmark for depth level determination and 3D visualization reports. In summary, based on the above scheme, joint discrimination of geometric morphology and material attenuation characteristics can be achieved, thereby improving the detection rate of surface defects in silica sol shell products. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an exemplary flowchart of a method for intelligent detection of defects in finished silica sol shell products according to some embodiments of this application; Figure 2This is a flowchart illustrating the process of determining multi-scale roughness features according to some embodiments of this application; Figure 3 This is a schematic diagram of the structure of an intelligent detection system for defects in silica sol shell-made products, as shown in some embodiments of this application. Figure 4 This is a schematic diagram of the structure of a computer device for implementing an intelligent detection method for defects in finished silica sol shell products, according to some embodiments of this application. Detailed Implementation

[0019] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] refer to Figure 1 The figure is an exemplary flowchart of a method for intelligent detection of defects in silica sol shell-made products according to some embodiments of this application. The method for intelligent detection of defects in silica sol shell-made products mainly includes the following steps: In step 101, a three-dimensional laser scan is performed on the surface of the finished silica sol shell to obtain point cloud data of the shell, and then the curvature distribution characteristics and local geometric deviations of the shell are extracted from the point cloud data.

[0021] It should be noted that, in this application, point cloud data is a data set that records the three-dimensional spatial coordinate information of the shell surface. In specific implementation, the finished product of the silica sol shell is placed on the measurement platform of the laser scanner, the distance between the scanner and the shell is adjusted so that the shell is completely within the scanning field of view, the laser scanner is started to perform a 360-degree rotation scan on the shell surface, the scanner emits a laser beam and receives the signal reflected back from the shell surface, and records the spatial three-dimensional coordinates of each sampling point. After the scan is completed, the set of all collected spatial coordinate points is used as the point cloud data of the shell.

[0022] In some embodiments, extracting the curvature distribution features and local geometric deviations of the shell from the point cloud data can be achieved using the following steps: The point cloud data is subjected to voxel filtering for noise reduction, and then the normal vector and curvature value of each point in the point cloud data are calculated to obtain the curvature distribution characteristics of all points. The standard 3D model of the shell and the point cloud data are registered, and the normal distance from each point cloud point to the surface of the standard model is calculated to obtain the local geometric deviation.

[0023] It should be noted that, in this application, the normal vector is a three-dimensional direction vector representing the orientation of the local surface where each point in the point cloud data is located; the curvature value is a numerical value that quantifies the degree of curvature of the local surface where each point in the point cloud data is located; the curvature distribution feature is a set of parameters describing the statistical distribution of curvature values ​​of all points on the shell surface; the standard three-dimensional model is an idealized digital model that serves as a benchmark for comparing the shape of the shell surface; the normal distance is a scalar value that measures the vertical distance from each point in the point cloud data to the surface of the standard three-dimensional model; and the local geometric deviation is a set of spatial offsets representing the overall difference between the actual shape of the shell surface and the standard shape.

[0024] In specific implementation, firstly, voxel filtering is performed on the point cloud data for noise reduction. Then, the normal vector and curvature value of each point in the point cloud data are calculated to obtain the curvature distribution characteristics of all points. This can be achieved as follows: the point cloud data is divided into cubic grids with a side length of 1 mm in three-dimensional space, each grid being called a voxel. All points contained in each voxel are counted, and the average spatial coordinate of all points in each voxel is calculated. This average value is used to replace all the original points in the voxel to complete voxel filtering and noise reduction. For the filtered point cloud data, neighborhood points within a radius of 2 mm are searched for each point. A local plane is fitted using the neighborhood points, and the normal direction of this local plane is used as the normal vector of the point. The sum of squared distances from all points in the local plane to the fitted plane is calculated. The square root of the sum of squared distances is divided by the number of neighborhood points, and the result is used as the curvature value of the point. After calculating the curvature values ​​for all points in turn, the maximum, minimum, mean, and standard deviation of the curvature values ​​are counted. The calculation results are used as curvature distribution characteristics. Then, the standard three-dimensional model of the shell and the point cloud data are registered. The normal distance from each point cloud point to the surface of the standard model is calculated. The local geometric deviation can be achieved in the following way: Load the standard three-dimensional model of the shell, which is composed of a series of triangular facets. Use the iterative nearest point algorithm to register the filtered point cloud data with the standard three-dimensional model. The termination condition of the iterative nearest point algorithm is set to the change in registration error between two iterations being less than 0.01 mm. After registration, the point cloud data and the standard three-dimensional model are in the same spatial coordinate system. For each point in the point cloud data, find the triangular facet on the standard three-dimensional model that is closest to the point. Draw a perpendicular line from the point to the plane where the triangular facet is located. The length of the perpendicular line segment is the normal distance of the point. The sign of the normal distance is determined by whether the point is located outside or inside the triangular facet. Organize the normal distances of all points into an array according to the spatial coordinate order. Use this array as the local geometric deviation.

[0025] In step 102, suspected defect areas of the shell are identified based on the curvature distribution characteristics, and echo intensity sequences of the shell are collected within the suspected defect areas. The material uniformity index of the shell is calculated based on the echo intensity sequences.

[0026] In some embodiments, identifying suspected defect areas of the shell based on the curvature distribution characteristics can be achieved using the following steps: Points with curvature values ​​exceeding a preset threshold are marked as high curvature points; Euclidean distance clustering is performed on the high curvature points to obtain multiple candidate regions; Candidate regions with an area smaller than the minimum defect size are removed, and the remaining regions are retained as suspected defect regions.

[0027] It should be noted that in this application, the preset threshold is the curvature value limit for screening high curvature points; high curvature points are candidate defect points that mark the curvature value of the shell surface exceeding the normal range; the candidate region is the spatial range that characterizes the aggregation of high curvature points; the minimum defect size is the lower limit of the area used to filter out candidate regions that are too small; and the suspected defect region is the spatial range on the shell surface that is most likely to contain real defects.

[0028] In specific implementation, firstly, marking points with curvature values ​​exceeding a preset threshold as high curvature points can be achieved as follows: Add twice the standard deviation to the mean curvature value in the curvature distribution feature, and use the sum as the preset threshold; traverse each point in the point cloud data, compare the curvature value of that point with the preset threshold, and mark points with curvature values ​​greater than the preset threshold as high curvature points; then, perform Euclidean distance clustering on the high curvature points to obtain multiple candidate regions, which can be achieved as follows: For all high curvature points, set the clustering distance threshold to 3 mm, starting from any unvisited high curvature point, search for other high curvature points within a 3 mm radius around that point, and group all the searched points into the same cluster set, using the newly grouped points as the basis for the clustering. Starting from a point, continue searching outwards until no new high-curvature points are added to the set. This set is then considered a candidate region. Repeat the above process until all high-curvature points have been visited, resulting in multiple candidate regions. Finally, candidate regions with areas smaller than the minimum defect size are eliminated, and the remaining regions are retained as suspected defect regions. This can be achieved as follows: Set the minimum defect size to 10 square millimeters. For each candidate region, calculate the area of ​​the smallest bounding polygon formed by all points within that region. Use this area as the area of ​​the candidate region. Compare the area of ​​the candidate region with the minimum defect size. Eliminate candidate regions with areas smaller than 10 square millimeters, and retain the remaining candidate regions as suspected defect regions.

[0029] It should be noted that, in this application, the echo intensity sequence is an arrangement of the strength values ​​of the echo signals received by the laser scanner in the suspected defect area according to the scanning order. In specific implementation, all points located in the suspected defect area are extracted from the point cloud data and sorted according to the original time order when these points were acquired by the laser scanner. The echo intensity value corresponding to each sorted point is read, and these values ​​are arranged into a sequence, which is used as the echo intensity sequence.

[0030] In some embodiments, the material uniformity index of the shell can be calculated based on the echo intensity sequence using the following steps: Calculate the mean and standard deviation of the echo intensity sequence within the suspected defect area; Using a defect-free area on the shell surface as a reference area, the baseline value of the echo intensity of the reference area is calculated. The material uniformity index of the shell is obtained based on the echo intensity benchmark value, mean, and standard deviation.

[0031] It should be noted that in this application, the reference area is a defect-free local area on the shell surface that provides a normal echo intensity benchmark; the echo intensity benchmark value is a standard value that characterizes the normal material echo characteristics on the shell surface; and the material uniformity index is a dimensionless parameter that quantifies the degree of difference between the material consistency of the suspected defect area and the normal area.

[0032] In specific implementation, firstly, the mean and standard deviation of the echo intensity sequence within the suspected defect area can be calculated as follows: All values ​​in the echo intensity sequence are summed, and the sum is divided by the total number of values ​​in the sequence; the result is taken as the mean of the echo intensity sequence. For each value in the echo intensity sequence, the square of the difference between that value and the mean is calculated, all squares are summed, and the result is divided by the total number of values ​​in the sequence; the square root of the quotient is then taken as the standard deviation of the echo intensity sequence. Then, the defect-free area on the shell surface is used as a reference area, and the baseline echo intensity value for the reference area can be calculated using... The following method is used to achieve this: Select a region on the shell surface that is visually inspected and confirmed to be free of cracks, pores, and shrinkage. Use this region as a reference region. Extract the echo intensity values ​​of all points within the reference region from the point cloud data. Calculate the arithmetic mean of these echo intensity values ​​and use this mean as the echo intensity benchmark value. Finally, based on the echo intensity benchmark value, mean, and standard deviation, the material uniformity index of the shell can be obtained as follows: Subtract the echo intensity benchmark value from the mean of the echo intensity sequence, divide the difference by the standard deviation of the echo intensity sequence, and use the calculation result as the material uniformity index.

[0033] In step 103, the multi-scale roughness characteristics of the shell in the suspected defect area are determined by the material uniformity index and the local geometric deviation. Based on the multi-scale roughness characteristics, a defect confidence map of the shell is constructed, and the locations of multiple defects in the shell are located.

[0034] In some embodiments, the multi-scale roughness characteristics of the shell in the suspected defect region are determined by the material uniformity index and the local geometric deviation, with reference to... Figure 2 The figure is a flowchart illustrating the determination of multi-scale roughness features in some embodiments of this application. In this embodiment, the determination of multi-scale roughness features can be achieved using the following steps: In step 1031, multiple spherical neighborhoods with different radii are set within the suspected defect area; In step 1032, the change in the normal vector angle of the point cloud within each spherical neighborhood is calculated based on the local geometric deviation. In step 1033, the material uniformity index is used as a weighting coefficient to perform weighted fusion on the changes in the angles of each normal vector, thereby obtaining the multi-scale roughness characteristics of the shell in the suspected defect area.

[0035] It should be noted that, in this application, the spherical neighborhood is a spherical region that defines the spatial range around each point; the change in normal vector angle is a numerical value that quantifies the degree of difference in the direction of the normal vector at each point within the spherical neighborhood; and the multi-scale roughness feature is a comprehensive parameter that characterizes the degree of microscopic undulation of the shell surface at multiple spatial scales.

[0036] In specific implementation, firstly, setting multiple spherical neighborhoods with different radii within the suspected defect area can be achieved as follows: For each point within the suspected defect area, three spherical neighborhoods with different radii are set around that point, with the three radii being 1 mm, 2 mm, and 3 mm respectively. Each spherical neighborhood spatially constitutes a sphere containing all point cloud data points within that sphere. These three spherical neighborhoods with different radii are used as the basic spatial units for subsequent calculations. Then, calculating the change in the normal vector angle of the point cloud within each spherical neighborhood based on the local geometric deviation can be achieved as follows: For each spherical neighborhood, the normal vectors of all points within that neighborhood are extracted, and the angle between the normal vector of the center point and the normal vectors of every other point within the neighborhood is calculated, resulting in a set of angle values. The arithmetic mean of this set of angle values ​​is calculated, and this average value is used as the change in the normal vector angle of one spherical neighborhood. The above calculations are performed on three spherical neighborhoods with different radii to obtain three normal vector angle changes, corresponding to spatial scales of radii of 1 mm, 2 mm, and 3 mm, respectively. Finally, the material uniformity index is used as a weighting coefficient to weight and fuse the various normal vector angle changes to obtain the multi-scale roughness features of the shell in the suspected defect area. This can be achieved in the following way: the three normal vector angle changes are denoted as angle change 1, angle change 2, and angle change 3 in ascending order of radius. The material uniformity index of each spherical neighborhood is calculated using the above method. The material uniformity index is used as the corresponding weighting coefficient and multiplied by angle change 1, angle change 2, and angle change 3 to obtain three weighted results. The sum of the three weighted results is used as the multi-scale roughness feature of that point. The above process is repeated for each point in the suspected defect area to obtain the multi-scale roughness features of all points.

[0037] In some embodiments, constructing a defect confidence map of the shell based on the multi-scale roughness features and locating multiple defect positions of the shell can be achieved by the following steps: The multi-scale roughness features are spatially rasterized in the shell point cloud coordinate system, and the average roughness value within each grid is calculated. Rasteres with average roughness values ​​greater than the confidence threshold are marked as high-confidence defect rasters, and adjacent high-confidence defect rasters are merged by region growing to obtain multiple defect locations. Map the confidence values ​​of all defect locations to color labels to generate a defect confidence map.

[0038] It should be noted that, in this application, the average roughness value is a statistical measure representing the central tendency of the multi-scale roughness features of all points within a single grid; the confidence threshold is the lower limit of the average roughness value for screening high-confidence defect grids; the high-confidence defect grid is a spatial unit that marks the probability of a defect; the defect location is the point that identifies the spatial coordinates of the defect on the shell surface; the color label is a visual coding symbol that indicates the level of defect confidence; and the defect confidence map is a color distribution map that visualizes the probability of defects in different areas of the shell surface.

[0039] In specific implementation, firstly, the multi-scale roughness features are spatially rasterized in the shell point cloud coordinate system. The average roughness value within each grid can be calculated as follows: using the shell point cloud coordinate system as a spatial reference, the entire point cloud space is divided into a series of closely arranged cubic grids with a grid side length of 1 mm. For each grid, the multi-scale roughness feature values ​​of all points contained within the grid are extracted, and the arithmetic mean of all feature values ​​is calculated. This arithmetic mean is used as the average roughness value. Then, a roughness value greater than a confidence threshold is considered. The grid is marked as a high-confidence defect grid, and adjacent high-confidence defect grids are merged by region growing to obtain multiple defect locations. This can be achieved as follows: Calculate the average roughness value of all grids, and then calculate the 90th percentile of all average roughness values. Use this percentile as the confidence threshold. Traverse each grid, marking grids with average roughness values ​​greater than the confidence threshold as high-confidence defect grids. Starting from any unvisited high-confidence defect grid, search for adjacent grids in six directions (up, down, left, right, front, and back) adjacent to that grid surface. Grids that are also marked as high-confidence defective grids in adjacent grids are grouped into the same connected region. Starting from the newly added grid, the search continues outwards until no more adjacent high-confidence defective grids are added. This connected region is then considered a defective region. The average value of the geometric centers of all grids within this defective region is calculated, and this average value is taken as the defect location. This process is repeated until all high-confidence defective grids have been visited, thus obtaining multiple defect locations. Finally, the confidence values ​​of all defect locations are mapped to color labels to generate a defect confidence map. The following method can be used to achieve this. The implementation is as follows: For each defect location, extract the average roughness value of all grids within the defect area corresponding to that defect location, calculate the arithmetic mean of all average roughness values, and use this arithmetic mean as the confidence value of that defect location. Map the confidence values ​​of all defect locations from low to high as a continuous color gradient from blue to red, where blue represents low confidence and red represents high confidence. In the shell point cloud coordinate system, assign a corresponding color label to each defect location to generate a color distribution map, and use this color distribution map as the defect confidence map.

[0040] In step 104, the defect depth level of each defect location is determined based on the attenuation degree of the echo intensity sequence at each defect location, and then the defect score of the shell is output, and a three-dimensional defect distribution visualization report of the finished shell is generated.

[0041] In some embodiments, determining the defect depth level at each defect location based on the attenuation degree of the echo intensity sequence at each defect location can be achieved using the following steps: The attenuation coefficient at each defect location is determined based on the minimum intensity value of the echo intensity sequence at each defect location and the background intensity value. Each attenuation coefficient is compared with multiple preset depth level threshold ranges to obtain the initial depth level of each defect location; The initial depth levels are corrected based on the material uniformity index of the neighborhood of each defect location, and the defect depth level of each defect location is output.

[0042] It should be noted that, in this application, the minimum intensity value is a numerical value characterizing the weakest echo signal at the defect location; the background intensity value is a numerical value used to represent the baseline level of echo intensity in the normal area surrounding the defect; the attenuation coefficient is a dimensionless parameter quantifying the proportion of attenuation of the echo signal at the defect location relative to the normal area; the depth level threshold range is the range of attenuation coefficient values ​​used to classify the severity levels of defect depth; the initial depth level is a level used to initially identify the longitudinal severity of the defect; the material uniformity index is a parameter used to reflect the degree of consistency of the material around the defect; and the defect depth level is the final level used to comprehensively evaluate the longitudinal severity of the defect.

[0043] In specific implementation, firstly, based on the minimum intensity value and background intensity value of the echo intensity sequence at each defect location, the attenuation coefficient at each defect location can be determined in the following way: For each defect location, extract the echo intensity values ​​of all points within the defect area corresponding to that defect location, find the minimum value, and use this minimum value as the minimum intensity value. Extend outwards by 2 mm from the defect location as a buffer zone, extract the echo intensity values ​​of all points within the buffer zone that do not belong to any defect area, calculate the arithmetic mean of these echo intensity values, and use this arithmetic mean as the background intensity value. The result of dividing the minimum intensity value by the background intensity value is used as the attenuation coefficient. Then, compare each attenuation coefficient with multiple preset depth level threshold intervals to obtain the initial depth level of each defect location. This can be achieved in the following way: Divide the defect depth into 3 levels, with the corresponding preset depth level threshold intervals as follows: Attenuation coefficient greater than 0.8 indicates a mild level interval; attenuation coefficient greater than 0.5 and less than or equal to 0.8 indicates a moderate level interval; and attenuation coefficient less than or equal to 0.5 indicates a severe level interval. The attenuation coefficient calculated at each defect location is compared with three intervals. The level corresponding to the interval in which the attenuation coefficient falls is taken as the initial depth level of the defect location. Finally, the initial depth levels are corrected according to the material uniformity index of the neighborhood of each defect location. The defect depth level of each defect location can be output in the following way: For each defect location, the material uniformity index of all points within a radius of 3 mm is extracted with the defect location as the center. The arithmetic mean of the material uniformity index is calculated as the neighborhood material mean. When the neighborhood material mean is greater than 0.7, the initial depth level is adjusted one level towards the more severe direction, but the adjustment does not exceed the severe level. When the neighborhood material mean is less than 0.3, the initial depth level is adjusted one level towards the less severe direction, but the adjustment does not fall below the mild level. When the neighborhood material mean is between 0.3 and 0.7, the initial depth level remains unchanged, and the corrected level is taken as the defect depth level of the defect location. The above correction is performed on all defect locations in sequence to output the defect depth level corresponding to each defect location.

[0044] It should be noted that in this application, the defect score is a quantitative score that evaluates the overall severity of defects in the shell; the three-dimensional defect distribution visualization report is a graphic output document that intuitively displays the location, type, and severity of all defects in the shell.

[0045] In practice, firstly, the total number of defect locations on the shell is counted. For each defect location, a score is assigned based on its defect depth level: 1 point for mild, 3 points for moderate, and 5 points for severe. The scores of all defect locations are summed to obtain the total defect score. The total defect score is then divided by the total number of defect locations, and the result is used as the defect score of the shell. Next, a standard 3D model of the shell is loaded as the base map. At the corresponding coordinates of the standard 3D model, a 3D marker symbol is added to each defect location. The color of the marker symbol is determined according to the defect depth level: green for mild, yellow for moderate, and red for severe. The size of the marker symbol is scaled according to the area of ​​the defect area corresponding to the defect location; the larger the area, the larger the marker symbol. A defect statistics table is added to the report page, listing the coordinates, defect depth level, and defect score of each defect location. The defect score is displayed in a prominent font at the top of the report. The above 3D model view, defect statistics table, and defect score are integrated into a multi-page document as a 3D defect distribution visualization report.

[0046] Furthermore, in another aspect of this application, in some embodiments, this application provides an intelligent detection system for defects in finished silica sol shell products, referencing... Figure 3 The figure is a schematic diagram of the structure of an intelligent detection system for defects in silica sol shell-made products according to some embodiments of this application. The intelligent detection system for defects in silica sol shell-made products includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to perform three-dimensional laser scanning on the surface of the finished product made of silica sol shell to obtain the point cloud data of the shell, and then extract the curvature distribution characteristics and local geometric deviations of the shell from the point cloud data. Processing module 202, in this application, is used to identify suspected defect areas of the shell based on the curvature distribution characteristics, collect the echo intensity sequence of the shell in the suspected defect areas, and calculate the material uniformity index of the shell based on the echo intensity sequence. It should be noted that the processing module 202 is also used to determine the multi-scale roughness characteristics of the shell in the suspected defect area through the material uniformity index and the local geometric deviation, construct the defect confidence map of the shell based on the multi-scale roughness characteristics, and locate multiple defect locations of the shell. The execution module 203 in this application is mainly used to determine the defect depth level of each defect location based on the attenuation degree of the echo intensity sequence at each defect location, and then output the defect score of the shell and generate a three-dimensional defect distribution visualization report of the finished shell.

[0047] The foregoing has detailed examples of the intelligent detection method and system for defects in silica sol-coated finished products provided in the embodiments of this application. It is understood that the corresponding apparatus, in order to achieve the above functions, includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0048] In some embodiments, this application also provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device performs the above-described intelligent detection method for defects in silica sol shell-making products.

[0049] In some embodiments, reference Figure 4 The dashed lines in the figure indicate that the unit or module is optional. This figure is a schematic diagram of the structure of a computer device for implementing an intelligent detection method for defects in silica sol-coated finished products according to an embodiment of this application. The intelligent detection method for defects in silica sol-coated finished products described in the above embodiments can... Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a memory 302 and at least one communication unit 305. The computer device may be a terminal device, a server or a chip.

[0050] Processor 301 can be a general-purpose processor or a special-purpose processor. For example, processor 301 can be a central processing unit (CPU), which can be used to control computer devices, execute software programs, and process data from software programs. The computer device may also include a communication unit 305 for inputting (receiving) and outputting (transmitting) signals.

[0051] For example, the computer device may be a chip, and the communication unit 305 may be the input and / or output circuit of the chip, or the communication unit 305 may be the communication interface of the chip, which may be a component of a terminal device, network device or other device.

[0052] For example, the computer device may be a terminal device or a server, and the communication unit 305 may be a transceiver of the terminal device or the server, or the communication unit 305 may be a transceiver circuit of the terminal device or the server.

[0053] The computer device may include one or more memories 302 storing a program 304. The program 304 can be executed by a processor 301 to generate instructions 303, causing the processor 301 to execute the method described in the above method embodiments according to the instructions 303. Optionally, the memory 302 may also store data (such as a target audit model). Optionally, the processor 301 may also read data stored in the memory 302, which may be stored at the same storage address as the program 304, or it may be stored at a different storage address than the program 304.

[0054] The processor 301 and memory 302 can be configured separately or integrated together, for example, integrated on the system on chip (SOC) of the terminal device.

[0055] It should be understood that each step of the above method embodiment can be completed by hardware logic circuits or software instructions in the processor 301. The processor 301 can be a CPU, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, such as discrete gate, transistor logic devices, or discrete hardware components.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] For example, in some embodiments, this application also provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to implement the above-described intelligent detection method for defects in silica sol shell-making products.

[0058] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0059] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for intelligent detection of defects in finished products made from silica sol shells, characterized in that, Includes the following steps: A three-dimensional laser scan is performed on the surface of the finished shell made of silica sol to obtain point cloud data of the shell, and then the curvature distribution characteristics and local geometric deviations of the shell are extracted from the point cloud data. Based on the curvature distribution characteristics, suspected defect areas of the shell are identified, and echo intensity sequences of the shell are collected within the suspected defect areas. The material uniformity index of the shell is calculated based on the echo intensity sequences. The multi-scale roughness characteristics of the shell in the suspected defect area are determined by the material uniformity index and the local geometric deviation. Based on the multi-scale roughness characteristics, a defect confidence map of the shell is constructed, and the locations of multiple defects in the shell are located. The defect depth level at each defect location is determined based on the attenuation degree of the echo intensity sequence at each defect location, and then the defect score of the shell is output, and a three-dimensional defect distribution visualization report of the finished shell is generated.

2. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, Extracting the curvature distribution features and local geometric deviations of the shell from the point cloud data specifically includes: The point cloud data is subjected to voxel filtering for noise reduction, and then the normal vector and curvature value of each point in the point cloud data are calculated to obtain the curvature distribution characteristics of all points. The standard 3D model of the shell and the point cloud data are registered, and the normal distance from each point cloud point to the surface of the standard model is calculated to obtain the local geometric deviation.

3. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, Identifying suspected defect areas of the shell based on the aforementioned curvature distribution characteristics specifically includes: Points with curvature values ​​exceeding a preset threshold are marked as high curvature points; Euclidean distance clustering is performed on the high curvature points to obtain multiple candidate regions; Candidate regions with an area smaller than the minimum defect size are removed, and the remaining regions are retained as suspected defect regions.

4. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, The calculation of the material uniformity index of the shell based on the echo intensity sequence specifically includes: Calculate the mean and standard deviation of the echo intensity sequence within the suspected defect area; Using a defect-free area on the shell surface as a reference area, the baseline value of the echo intensity of the reference area is calculated. The material uniformity index of the shell is obtained based on the echo intensity benchmark value, mean, and standard deviation.

5. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, Determining the multi-scale roughness characteristics of the shell in the suspected defect region using the material uniformity index and the local geometric deviation specifically includes: Multiple spherical neighborhoods with different radii are set within the suspected defect area; The change in the normal vector angle of the point cloud within each spherical neighborhood is calculated based on the local geometric deviation. Using the material uniformity index as a weighting coefficient, the changes in the angles of each normal vector are weighted and fused to obtain the multi-scale roughness characteristics of the shell in the suspected defect area.

6. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, Constructing a defect confidence map of the shell based on the multi-scale roughness features and locating multiple defect locations of the shell specifically includes: The multi-scale roughness features are spatially rasterized in the shell point cloud coordinate system, and the average roughness value within each grid is calculated. Rasteres with average roughness values ​​greater than the confidence threshold are marked as high-confidence defect rasters, and adjacent high-confidence defect rasters are merged by region growing to obtain multiple defect locations. Map the confidence values ​​of all defect locations to color labels to generate a defect confidence map.

7. The intelligent detection method for defects in silica sol shell-made products as described in claim 1, characterized in that, Determining the defect depth level at each defect location based on the attenuation degree of the echo intensity sequence at each defect location specifically includes: The attenuation coefficient at each defect location is determined based on the minimum intensity value of the echo intensity sequence at each defect location and the background intensity value. Each attenuation coefficient is compared with multiple preset depth level threshold ranges to obtain the initial depth level of each defect location; The initial depth levels are corrected based on the material uniformity index of the neighborhood of each defect location, and the defect depth level of each defect location is output.

8. A smart detection system for defects in finished products made from silica sol shells, characterized in that, include: The acquisition module is used to perform three-dimensional laser scanning on the surface of the finished product made of silica sol to obtain point cloud data of the shell, and then extract the curvature distribution characteristics and local geometric deviations of the shell from the point cloud data; The processing module is used to identify suspected defect areas of the shell based on the curvature distribution characteristics, collect the echo intensity sequence of the shell in the suspected defect areas, and calculate the material uniformity index of the shell based on the echo intensity sequence. The processing module is also used to determine the multi-scale roughness characteristics of the shell in the suspected defect area through the material uniformity index and the local geometric deviation, construct the defect confidence map of the shell based on the multi-scale roughness characteristics, and locate multiple defect locations of the shell. The execution module is used to determine the defect depth level of each defect location based on the attenuation degree of the echo intensity sequence at each defect location, and then output the defect score of the shell and generate a three-dimensional defect distribution visualization report of the finished shell.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to call and run the computer programs from the memory, so that the computer device performs the intelligent detection method for defects in silica sol shell-making products as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions or code that, when executed on a computer, cause the computer to implement the intelligent detection method for defects in silica sol shell products as described in any one of claims 1 to 7.

Images