Comparison test block, evaluation method and system for defect detection capability of CT (Computed Tomography) system

Abstract

The invention provides a CT (Computed Tomography) system defect detection capability reference block, evaluation method and system. The CT system defect detection capability reference block comprises a defect simulation sample A and a matrix disc B, the method comprises the following steps: preprocessing a CT detection image, extracting a defect region of interest, and obtaining a defect detection capability evaluation two-dimensional plane; correcting the voxel size of the CT image; carrying out sub-pixel edge detection; fitting a result, and establishing an image pixel position coordinate; according to the sub-pixel edge point set, matrix gray scale mu l (h) and defect gray scale mu b (h) at different positions are obtained; establishing a relational expression of the defect size and the gray level, drawing a relation curve I of the defect size D and # imgabs0 # and a human eye 50% probability recognition probability curve II, and obtaining a minimum recognizable defect; and establishing a relation curve of the defect size and the measurement error thereof. According to the method, the defect size measurement precision is high, the influence of image noise is small, and the minimum defect detection capability and defect measurement error can be quantitatively evaluated.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial nondestructive testing, and specifically relates to a CT system defect detection capability comparison test block, an evaluation method and a system, and in particular to a CT system defect detection capability comparison test block and an evaluation method based on a triangle model. Background Art

[0002] With the rapid progress of material preparation and processing technology, the quantitative evaluation of defects inside materials has become particularly important, which has put forward higher requirements for the detection and characterization of defects. CT is a high-resolution computer tomography scanning technology, and its highest resolution can reach the micron or even nanometer level, which is particularly suitable for the detection and characterization of small and complex components. However, the quantitative evaluation of the defect detection capability of CT systems in practical applications has always been a puzzling problem.

[0003] The limited number and width of the beam will cause feature blurring, which will change the shape of the feature and the distinguishability of multiple features. This blurring will also affect the perceived contrast, especially the contrast of small features. In addition, CT images are inevitably affected by factors such as noise, artifacts and scattering, which will also seriously affect the identification and measurement of defects. The identification of small defects is mainly determined by human observation and is seriously affected by the image window position. There are few studies on the quantitative evaluation of defect detection.

[0004] The main defect measurement methods include half-height width method, template matching algorithm and sub-pixel edge detection method. Among them, the half-height width method is simple to operate but has large random errors, and the measurement accuracy is related to the diameter of the defect size; the template matching algorithm can match according to the image gray value, image edge and image shape, but the measurement result is greatly affected by the image quality; the classic segmentation algorithm often cannot meet the detection requirements of high-precision parts. Many scholars have conducted research on sub-pixel edge segmentation methods, but severe artifacts in CT images can cause image edges to be blurred, making it difficult to accurately segment.

[0005] At present, the defect detection capability of CT systems is mainly evaluated by image quality, namely spatial resolution, density resolution and noise. The measured values ​​are generally obtained under the optimal parameter process and can only measure a single plane. However, in the CT inspection of actual products, due to the limitations of conditions such as workpiece size, the process parameters are often not optimal, and the defect detection capability can only be estimated by image quality.

[0006] The patent document "A method for evaluating the defect detection performance limit of a CT detection system" (CN110060293A) discloses a method for estimating defect detection capabilities based on a point spread function, but does not take into account the imaging differences caused by different ray attenuation coefficients inside actual defects. The resulting limit defect detection performance is not equivalent to the actual minimum detectable defect. Summary of the invention

[0007] In view of the defects in the prior art, the object of the present invention is to provide a CT system defect detection capability comparison test block, evaluation method and system.

[0008] A CT system defect detection capability comparison test block provided by the present invention includes: a defect simulation sample A and a base disc B.

[0009] The defect simulation sample A is a cube, and both the bottom and the side surfaces have an isosceles triangle groove and two circular grooves.

[0010] The base disc B is composed of a plurality of nested circular rings and has a square opening on the top.

[0011] The defect simulation sample A is embedded in the base disk B.

[0012] The CT system defect detection capability comparison test block is a variable diameter model.

[0013] Preferably, the outer diameter of the base disc B is 4 mm;

[0014] The defect simulation sample A is embedded in the center of the upper surface of the base disk B;

[0015] The defect simulation sample A is a cube with a side length of 1 mm.

[0016] The base of the isosceles triangle groove is 0.4 mm and the height is 0.8 mm.

[0017] The diameter of the circular groove is 0.4 mm, the distance between the centers of the two circles is 0.6 mm, and the depth of the groove is 0.05 mm.

[0018] The outer diameter of the CT system defect detection capability comparison test block is 10 mm and the thickness is 2 mm.

[0019] According to a method for evaluating defect detection capability of a CT system provided by the present invention, the defect detection capability comparison block of the CT system is used for detection, comprising:

[0020] Step S1: preprocessing the CT detection image, extracting the defect region of interest, and obtaining a two-dimensional plane for defect detection capability evaluation;

[0021] Step S2: Correcting the CT image voxel size;

[0022] Step S3: performing sub-pixel edge detection;

[0023] Step S4: fitting the sub-pixel edge detection results to establish the pixel position coordinates of the entire slice image;

[0024] Step S5: Obtain μ at different positions based on the results of sub-pixel edge detection l (h) and μ b (h);

[0025] Step S6: Establish the relationship between defect size and grayscale, and plot the defect size D and The relationship curve 1 and the 50% probability recognition probability curve 2 of the human eye are used to obtain the minimum recognizable defect;

[0026] Step S7: Establish a relationship curve between defect size and its measurement error.

[0027] Preferably, the CT system defect detection capability comparison test block is placed on the product surface during the test, and the outer diameter is consistent with the maximum diameter of the product actually detected by the CT system, and the material is consistent with the material of the actual detected product. If the material is inconsistent, it is multiplied by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness;

[0028] Each artificial defect is measured three times, and the average value of the three measurement results is used as the standard value of the actual size.

[0029] In the step S1, the median filter is used to preprocess the CT detection image, the defect segmentation algorithm is used to extract the simulated defect interest region A through image cropping, the canny operator is used to obtain the initial edge contour of the image, the position of the image edge point is confirmed, the least squares method is used to fit the edge area, and the area surrounded by the edge contour is extracted as the interest region B, which is recorded as the defect interest region.

[0030] Slice images are sorted and segmented by three-dimensional threshold to roughly extract isosceles triangle grooves and circular groove models. The vertical direction of the scanning direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0.

[0031] Select the slice images with the median number and two intervals apart as slice Z for defect detection capability evaluation;

[0032] Define the plane perpendicular to the Z axis as the xy plane, and select the slice XY for defect detection capability evaluation in the same way.

[0033] In step S2, level set segmentation is used to obtain the sub-pixel point cloud of the central area plane, the positions of the two circle centers are obtained based on least squares circle fitting, the distance between the circle centers is measured, and the voxel size error of the CT system is corrected according to the actual measured value of the distance between the circle centers.

[0034] Preferably, in step S3, a sobel operator is used to obtain a gradient image Rix,y), and sub-pixel edge detection is performed based on a Lagrange interpolation method in the gradient direction.

[0035] Take three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l 0 (x)+R i l 1 (x)+R i+1 l 2 (x);

[0036] l j (x) represents the Lagrange polynomial,

[0037] Among them, w represents the pixel direction spacing;

[0038] x i ,y j All represent interpolation nodes;

[0039] i, j represent the sequence number, i, j∈[0,N];

[0040] N represents the total number of gradient image points obtained by the Sobel operator.

[0041] Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as

[0042] Take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, respectively (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is

[0043] In step S4, a triangle fitting is performed on the point set of sub-pixel edge points through Hough transform, and the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H are measured. The pixel position (s, h) of the entire slice image is established with the short side of the triangle as the S axis, the line perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

[0044] Preferably, in step S5, the grayscale distribution μp(s, h) of the pixels inside the sub-pixel edge is calculated, and the average grayscale value of the defects under the same h value is calculated.

[0045] The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μq(s, h) of the matrix material is obtained by the erosion and expansion algorithm.

[0046] The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ l (s, h)≠0;

[0047] Among them, n represents the nth pixel at the same height;

[0048] H represents the height of the triangular groove.

[0049] In step S6, the simulated defect width is the distance between the hypotenuses of the triangle. Based on the triangle morphology characteristics, the defect size relationship D = L*(Hh) / h is established, and the defect grayscale μ b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l (h) is converted into the grayscale relationship μ about the defect size b (D);

[0050] According to the measured grayscale distribution, the relationship curve 1 and the recognition probability curve 2 are drawn, and the horizontal coordinate of the intersection of the curve 1 and the curve 2 is the minimum identifiable defect size;

[0051] The curve 1 is a graph showing the defect size D and The relationship between the recognition probability curve 2 is

[0052] Where c is a constant between 2 and 5, indicating the probability of a defect being identified;

[0053] Δp represents the voxel size;

[0054] σ represents the noise value.

[0055] In step S7, based on the triangle fitting data, the distance between the two hypotenuses at different height y values ​​and different h values ​​inside the triangle fitting is measured as the defect size, the difference is counted, and a relationship curve between the defect size and its measurement error is established.

[0056] According to a CT system defect detection capability evaluation system provided by the present invention, the CT system defect detection capability comparison test block according to claim 1 or 2 is used for detection, comprising:

[0057] Module M1: Preprocess the CT detection image, extract the defect area of ​​interest, and obtain the two-dimensional plane for defect detection capability evaluation;

[0058] Module M2: Correction of CT image voxel size;

[0059] Module M3: perform sub-pixel edge detection;

[0060] Module M4: Fit the sub-pixel edge detection results to establish the pixel position coordinates of the entire slice image;

[0061] Module M5: Obtain μ at different positions based on the results of sub-pixel edge detection l (h) and μ b (h);

[0062] Module M6: Establish the relationship between defect size and grayscale, and plot the defect size D and The relationship curve 1 and the 50% probability recognition probability curve 2 of the human eye are used to obtain the minimum recognizable defect;

[0063] Module M7: Establish the relationship curve between defect size and its measurement error.

[0064] Preferably, the CT system defect detection capability comparison test block is placed on the product surface during the test, and the outer diameter is consistent with the maximum diameter of the product actually detected by the CT system, and the material is consistent with the material of the actual detected product. If the material is inconsistent, it is multiplied by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness;

[0065] Each artificial defect is measured three times, and the average value of the three measurement results is used as the standard value of the actual size.

[0066] In the module M1, median filtering is used to preprocess the CT detection image, a defect segmentation algorithm is used to extract the simulated defect interest region A through image cropping, the canny operator is used to obtain the initial edge contour of the image, the position of the image edge point is confirmed, the least squares method is used to fit the edge region, and the area surrounded by the edge contour is extracted as the interest region B, which is recorded as the defect interest region.

[0067] Slice images are sorted and segmented by three-dimensional threshold to roughly extract isosceles triangle grooves and circular groove models. The vertical direction of the scanning direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0.

[0068] Select the slice images with the median number and two intervals apart as slice Z for defect detection capability evaluation;

[0069] Define the plane perpendicular to the Z axis as the xy plane, and select the slice XY for defect detection capability evaluation in the same way.

[0070] The module M2 adopts level set segmentation to obtain the sub-pixel point cloud of the central area plane, obtains the positions of the two circle centers based on least squares circle fitting, measures the distance between the circle centers, and corrects the voxel size error of the CT system according to the actual measured value of the distance between the circle centers.

[0071] Preferably, the module M3 uses a sobel operator to obtain a gradient image R(x, y), and performs sub-pixel edge detection based on a Lagrange interpolation method in the gradient direction.

[0072] Take three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l 0 (x)+R i l 1 (x)+R i+1 l 2 (x).

[0073] l j (x) represents the Lagrange polynomial,

[0074] Where w represents the pixel spacing;

[0075] x i represents the interpolation node;

[0076] i, j represent the sequence number, i, j∈[0,N];

[0077] N represents the total number of gradient image points obtained by the Sobel operator.

[0078] Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as

[0079] Take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, respectively (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is

[0080] In the module M4, a triangle fitting is performed on the point set of sub-pixel edge points through Hough transform, and the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H are measured. The pixel position (s, h) of the entire slice image is established with the short side of the triangle as the S axis, the perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

[0081] Preferably, the grayscale distribution μp(s, h) of the pixels inside the sub-pixel edge in the module M5 is calculated by counting the average grayscale value of the defects under the same h value. μ l (s, h)≠0;

[0082] The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μ of the matrix material is obtained by the erosion and expansion algorithm. q (s, h);

[0083] The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ l (s, h)≠0;

[0084] Among them, n represents the nth pixel at the same height;

[0085] H represents the height of the triangular groove.

[0086] The simulated defect width in the module M6 is the distance between the hypotenuses of the triangle. Based on the triangle morphology characteristics, the defect size relationship D = L*(Hh) / h is established, and the defect grayscale μ b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l(h) is converted into the grayscale relationship μ about the defect size b (D).

[0087] According to the measured grayscale distribution, a relationship curve 1 and an identification probability curve 2 are drawn, and the horizontal coordinate of the intersection of curve 1 and curve 2 is the minimum identifiable defect size.

[0088] The curve 1 is a graph showing the defect size D and The relationship between the recognition probability curve 2 is

[0089] Where c is a constant between 2 and 5, indicating the probability of a defect being identified;

[0090] Δp represents the voxel size;

[0091] σ represents the noise value.

[0092] In the module M7, based on the triangle fitting data, the distance between the two hypotenuses at different height y values ​​and different h values ​​inside the triangle fitting is measured as the defect size, the difference is counted, and a relationship curve between the defect size and its measurement error is established.

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] 1. The evaluation method provided by the present invention has high defect size measurement accuracy, is less affected by image noise, and has small random fluctuations in defect contrast measurement.

[0095] 2. The defect detection capability comparison test block provided by the present invention can meet the real-time monitoring during the inspection of products with different diameters and materials, obtain the actual size information of simulated defects, and can simultaneously evaluate the longitudinal and circumferential defect detection capabilities of the CT system.

[0096] 3. The present invention is not limited by workpiece size and process parameters, and can combine multiple measurement and analysis methods to quantitatively evaluate the minimum defect detection capability and dimensional measurement error, and quantitatively evaluate the defect detection capability of the CT system. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0098] Figure 1 It is a flow chart of the method for evaluating the defect detection capability of the CT system;

[0099] Figure 2 It is a schematic diagram of radiographic equivalent coefficient values ​​based on aluminum;

[0100] Figure 3a Schematic diagram of defect simulation specimen A for comparison of defect detection capabilities of CT system;

[0101] Figure 3b Schematic diagram of the base disc B of the CT system defect detection capability comparison test block;

[0102] Figure 4 The plane selection for defect detection capability evaluation and the schematic diagram of matrix grayscale distribution and defect grayscale distribution;

[0103] Figure 5 The defect size and Schematic diagram of the relationship curve;

[0104] Figure 6 Schematic diagram of the defect size measurement error relationship curve. DETAILED DESCRIPTION

[0105] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0106] A CT system defect detection capability comparison test block provided by the present invention, taking FIG. 3 as an example, includes: a defect simulation sample A and a base disc B.

[0107] The defect simulation specimen A is embedded in the base disc B. The overall appearance of the defect simulation specimen A is a cube, with an isosceles triangle groove and two circular grooves on its bottom and side surfaces; the base disc B is composed of multiple nested circular rings and has a square opening on the top. A traceable measurement device, namely a three-coordinate measuring machine, is used to obtain the actual size information of the defect simulation specimen, including measuring the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H, the roundness of the circular groove, the distance between the centers of the two circles, and the verticality and roughness of the groove edge. Each artificial defect is measured three times, and the average value of the three measurement results is used as the standard value of the actual size. It can simultaneously evaluate the longitudinal and circumferential defect detection capabilities of the CT system, and quantitatively evaluate the minimum defect detection capability and the measurement error of different defect sizes.

[0108] In more preferred embodiments, the outer diameter of the base disk B is 4 mm, and the defect simulation sample A is embedded in the center of its upper surface;

[0109] Defect simulation specimen A is a cube with a side length of 1 mm, and has an isosceles triangle groove with a base of 0.4 mm and a height of 0.8 mm on the bottom and side, and two circular grooves with a diameter of 0.4 mm.

[0110] The distance between the centers of the two circular grooves is 0.6 mm, and the groove depth is 0.05 mm.

[0111] The actual product detected by the CT system has a diameter of 10 mm, the material is aluminum, and the voxel size is 10 μm.

[0112] The outer diameter of the comparison test block is 10mm and the thickness is 2mm. The test block is placed on the product surface for synchronous testing.

[0113] Compared with conventional comparison test blocks, the specimen processing is simple, the defect size can be measured traceably, and the defect size measurement accuracy is higher.

[0114] A defect detection capability evaluation method based on a triangle model is proposed, which includes a sub-pixel defect size measurement method, a defect contrast measurement method based on an erosion and expansion algorithm, a quantitative evaluation of the minimum detectable defect, and an error analysis method for defect size measurement. Figure 1 For example, including:

[0115] The CT system defect detection capability comparison test block is a variable diameter model. The outer diameter of the model disk is consistent with the maximum diameter of the actual product detected by the CT system. The model material is aluminum alloy. The material of the actual product to be detected is consistent with the model material. If the material is inconsistent, it is necessary to multiply it by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness. The equivalent coefficient value is Figure 2 For example.

[0116] From the perspective of actual defect detection, the influence of point spread function, scattering noise, beam hardening, edge effect and other multiple parameters are taken into consideration, and defect samples based on triangle model are used for detection.

[0117] Step S1: Preprocess the CT detection image using median filtering, extract the defect simulation sample area as the region of interest A by image cropping, use the least squares method to fit the edge area, extract the edge contour enclosed area as the region of interest B, and record it as the defect region of interest.

[0118] Specifically, the defect segmentation algorithm is used to extract the region of interest of the simulated defect. The canny operator is used to obtain the initial edge contour of the image, and the position of the edge point of the image is confirmed. The region of interest B is the CT detection data of the simulated defect. This can be used for the extraction of conventional defects and is universal.

[0119] Extract CT inspection data of defect simulation specimens to reduce the data processing workload of subsequent algorithms.

[0120] Specifically, the defect detection capability evaluation two-dimensional plane is obtained by sorting slice images, and the triangular groove and circular groove models are roughly extracted through three-dimensional threshold segmentation. The vertical direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0. Figure 4 For example, the three slice images with the median number and two intervals are selected as the slice Z for defect detection capability evaluation. The plane perpendicular to the Z axis is defined as the xy plane, and the slice XY for defect detection capability evaluation is selected in the same way.

[0121] Step S2: Use level set segmentation to measure the distance between the centers of the two circles and calibrate the voxel size detected by the CT system.

[0122] Specifically, the sub-pixel point cloud of the central area plane is obtained by level set segmentation, and then the positions of the two circle centers are obtained by the double circle model based on the least squares circle fitting method, and the distance between the centers is measured. The voxel size error of the CT system is corrected according to the actual measured value of the distance between the centers.

[0123] Step S3: Use the Sobel operator to obtain the gradient image R(x,y), perform sub-pixel edge detection based on the Lagrange interpolation method in the gradient direction, calculate the coordinates of the sub-pixel edge points, and thus realize the point cloud extraction of the sub-pixel edge.

[0124] Specifically, three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l 0 (x)+R i l 1 (x)+R i+1 l 2 (x).

[0125] l j (x) represents the Lagrange polynomial,

[0126] Among them, w represents the pixel direction spacing;

[0127] x i represents the interpolation node;

[0128] i, j represent the sequence number, i, j∈[0,N];

[0129] N represents the total number of gradient image points obtained by the Sobel operator.

[0130] Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as

[0131] Similarly, take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, and (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is The coordinates of the sub-pixel edge points are calculated to achieve point cloud extraction of the sub-pixel edge. This step can be used for the extraction of conventional defects and is universal.

[0132] Step S4: Perform triangle fitting on the point set of the sub-pixel edge through Hough transform, measure the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H, and establish the pixel position (s, h) of the entire slice image with the short side of the triangle as the S axis, the perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

[0133] Step S5: The dilation algorithm is used to dilate the sub-pixel edge point set by 7 pixels and 2 pixels respectively. The two dilation results are subtracted and multiplied with the original image pixel by pixel. μ is obtained by erosion and dilation algorithm. n (s, h).

[0134] Specifically, the grayscale distribution of pixels inside the sub-pixel edge is μ p (s, h), statistics of the average grayscale value of defects under the same h value μ l (s, h)≠0;

[0135] The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μ of the matrix material is obtained by the erosion and expansion algorithm. q (s, h);

[0136] The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ 1 (s, h)≠0;

[0137] Among them, n represents the nth pixel at the same height;

[0138] H represents the height of the triangular groove.

[0139] Furthermore, the grayscale measurement method of the defect and the matrix material can be further used for grayscale value analysis of the defect and the matrix material.

[0140] Step S6: Simulate the defect width as the distance between the hypotenuses of the triangle, and establish the defect size relationship D = L*(Hh) / h based on the triangle morphology characteristics. b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l (h) is converted into the grayscale relationship μ about the defect size b (D).

[0141] According to the grayscale distribution measured above, Figure 5 For example, plot the defect size D versus The relationship curve 1 is plotted on the same graph, and the probability curve 2 of 50% recognition probability of human eyes is plotted on the same graph.

[0142] Where c is a constant between 2 and 5, indicating that the probability of a defect being identified is greater than 50%;

[0143] Δp represents the voxel size;

[0144] σ represents the noise value.

[0145] by Figure 6 For example, the horizontal coordinate of the intersection of curve 1 and curve 2 is the minimum identifiable defect size.

[0146] Step S7: Draw and establish a relationship curve between defect size and its measurement error. Based on the triangle fitting data, measure the distance between the two hypotenuses at different heights (h values) inside the triangle fitting as the defect size, and count the defect width at the same height inside, and compare it with the measurement results of the three-dimensional coordinate measuring machine to count the difference.

[0147] The present invention also provides a CT system defect detection capability evaluation system, which can be implemented by executing the process steps of the CT system defect detection capability evaluation method, that is, technical personnel in this field can understand the CT system defect detection capability evaluation method as a preferred implementation of the CT system defect detection capability evaluation system.

[0148] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0149] According to a CT system defect detection capability evaluation system provided by the present invention, the CT system defect detection capability comparison test block according to claim 1 or 2 is used for detection, comprising:

[0150] Module M1: Preprocess the CT detection image, extract the defect area of ​​interest, and obtain the two-dimensional plane for defect detection capability evaluation;

[0151] Module M2: Correction of CT image voxel size;

[0152] Module M3: perform sub-pixel edge detection;

[0153] Module M4: Fit the sub-pixel edge detection results to establish the pixel position coordinates of the entire slice image;

[0154] Module M5: Obtain μ at different positions based on the results of sub-pixel edge detection l (h) and μ b (h);

[0155] Module M6: Establish the relationship between defect size and grayscale, and plot the defect size D and The relationship curve 1 and the 50% probability recognition probability curve 2 of the human eye are used to obtain the minimum recognizable defect;

[0156] Module M7: Establish the relationship curve between defect size and its measurement error.

[0157] In more preferred examples, the CT system defect detection capability comparison test block is placed on the product surface during detection, and the outer diameter is consistent with the maximum diameter of the product actually detected by the CT system, and the material is consistent with the material of the actual detected product. If the material is inconsistent, it is multiplied by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness;

[0158] Each artificial defect is measured three times, and the average value of the three measurement results is used as the standard value of the actual size.

[0159] In the module M1, median filtering is used to preprocess the CT detection image, a defect segmentation algorithm is used to extract the simulated defect interest region A through image cropping, the canny operator is used to obtain the initial edge contour of the image, the position of the image edge point is confirmed, the least squares method is used to fit the edge region, and the area surrounded by the edge contour is extracted as the interest region B, which is recorded as the defect interest region.

[0160] Slice images are sorted and segmented by three-dimensional threshold to roughly extract isosceles triangle grooves and circular groove models. The vertical direction of the scanning direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0.

[0161] Select the slice images with the median number and two intervals apart as slice Z for defect detection capability evaluation;

[0162] Define the plane perpendicular to the Z axis as the xy plane, and select the slice XY for defect detection capability evaluation in the same way.

[0163] The module M2 adopts level set segmentation to obtain the sub-pixel point cloud of the central area plane, obtains the positions of the two circle centers based on least squares circle fitting, measures the distance between the circle centers, and corrects the voxel size error of the CT system according to the actual measured value of the distance between the circle centers.

[0164] In more preferred examples, the module M3 uses a sobel operator to obtain a gradient image R(x, y), and performs sub-pixel edge detection based on a Lagrange interpolation method in the gradient direction.

[0165] Take three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l 0 (x)+R i l 1 (x)+R i+1 l 2 (x).

[0166] l j (x) represents the Lagrange polynomial,

[0167] Among them, w represents the pixel direction spacing;

[0168] x i represents the interpolation node;

[0169] i, j represent the sequence number, i, j∈[0,N];

[0170] N represents the total number of gradient image points obtained by the Sobel operator.

[0171] Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as

[0172] Take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, respectively (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is

[0173] In the module M4, a triangle fitting is performed on the point set of sub-pixel edge points through Hough transform, and the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H are measured. The pixel position (s, h) of the entire slice image is established with the short side of the triangle as the S axis, the perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

[0174] In more preferred embodiments, the grayscale distribution of the pixels inside the sub-pixel edge in the module M5 is μp(s, h), and the average grayscale value of the defects under the same h value is calculated. μ l (s, h)≠0;

[0175] The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μ of the matrix material is obtained by the erosion and expansion algorithm. q (s, h);

[0176] The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ l (s, h)≠0;

[0177] Among them, n represents the nth pixel at the same height;

[0178] H represents the height of the triangular groove.

[0179] The simulated defect width in the module M6 is the distance between the hypotenuses of the triangle. Based on the triangle morphology characteristics, the defect size relationship D = L*(Hh) / h is established, and the defect grayscale μ b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l (h) is converted into the grayscale relationship μ about the defect size b (D).

[0180] According to the measured grayscale distribution, a relationship curve 1 and an identification probability curve 2 are drawn, and the horizontal coordinate of the intersection of curve 1 and curve 2 is the minimum identifiable defect size.

[0181] The curve 1 is a graph showing the defect size D and The relationship between the recognition probability curve 2 is

[0182] Where c is a constant between 2 and 5, indicating the probability of a defect being identified;

[0183] Δp represents the voxel size;

[0184] σ represents the noise value.

[0185] In the module M7, based on the triangle fitting data, the distance between the two hypotenuses at different height y values ​​and different h values ​​inside the triangle fitting is measured as the defect size, the difference is counted, and a relationship curve between the defect size and its measurement error is established.

[0186] In the description of the present application, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0187] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A CT system defect detection capability comparison test block, characterized in that: include: Defect simulation specimen A and substrate disc B; The defect simulation specimen A is a cube, and both the bottom and the side surfaces have an isosceles triangle notch and two circular notches; The base disc B is composed of a plurality of nested circular rings and has a square opening on the top; The defect simulation specimen A is embedded in the base disc B; The CT system defect detection capability comparison test block is a variable diameter model.

2. The CT system defect detection capability comparison test block according to claim 1, characterized in that: The outer diameter of the base disc B is 4 mm; The defect simulation sample A is embedded in the center of the upper surface of the base disk B; The defect simulation sample A is a cube with a side length of 1 mm; The isosceles triangle groove has a base of 0.4 mm and a height of 0.8 mm; The diameter of the circular groove is 0.4 mm, the distance between the two circles is 0.6 mm, and the groove depth is 0.05 mm; The outer diameter of the CT system defect detection capability comparison test block is 10 mm and the thickness is 2 mm.

3. A method for evaluating defect detection capability of a CT system, characterized in that: The CT system defect detection capability comparison test block according to claim 1 or 2 is used for testing, comprising: Step S1: preprocessing the CT detection image, extracting the defect region of interest, and obtaining a two-dimensional plane for defect detection capability evaluation; Step S2: Correcting the CT image voxel size; Step S3: performing sub-pixel edge detection; Step S4: fitting the sub-pixel edge detection results to establish the pixel position coordinates of the entire slice image; Step S5: Obtain μ at different positions based on the results of sub-pixel edge detection l (h) and μ b (h); Step S6: Establish the relationship between defect size and grayscale, and plot the defect size D and The relationship curve 1 and the 50% probability recognition probability curve 2 of the human eye are used to obtain the minimum recognizable defect; Step S7: Establish a relationship curve between defect size and its measurement error.

4. The method for evaluating defect detection capability of a CT system according to claim 3, characterized in that: The CT system defect detection capability comparison test block is placed on the product surface during the test, and the outer diameter is consistent with the maximum diameter of the product actually detected by the CT system, and the material is consistent with the material of the actual detected product. If the material is inconsistent, it is multiplied by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness; Each artificial defect is measured three times, and the average of the three measurement results is used as the standard value of the actual size; In the step S1, the CT detection image is preprocessed by using a median filter, a defect segmentation algorithm is used to extract a simulated defect region of interest A by image cropping, a canny operator is used to obtain the initial edge contour of the image, the position of the edge point of the image is confirmed, and the edge region is fitted by the least squares method, and the region surrounded by the edge contour is extracted as the region of interest B, which is recorded as the defect region of interest; Slice images are sorted and segmented by three-dimensional threshold to roughly extract isosceles triangle grooves and circular groove models. The vertical direction of the scanning direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0. Select the slice images with the median number and two intervals apart as slice Z for defect detection capability evaluation; Define the plane perpendicular to the Z axis as the xy plane, and select the slice XY for defect detection capability evaluation in the same way; In step S2, level set segmentation is used to obtain the sub-pixel point cloud of the central area plane, the positions of the two circle centers are obtained based on least squares circle fitting, the distance between the circle centers is measured, and the voxel size error of the CT system is corrected according to the actual measured value of the distance between the circle centers.

5. The method for evaluating defect detection capability of a CT system according to claim 3, characterized in that: In step S3, a sobel operator is used to obtain a gradient image R(x, y), and sub-pixel edge detection is performed based on the Lagrange interpolation method in the gradient direction; Take three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l0(x)+R i l1(x)+R i+1 l2(x); l j (x) represents the Lagrange polynomial, Among them, w represents the pixel direction spacing; x i represents the interpolation node; i, j represent the sequence number, i, j∈[0,N]; N represents the total number of gradient image points obtained by the Sobel operator; Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as Take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, respectively (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is In step S4, a triangle fitting is performed on the point set of sub-pixel edge points through Hough transform, and the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H are measured. The pixel position (s, h) of the entire slice image is established with the short side of the triangle as the S axis, the line perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

6. The method for evaluating defect detection capability of a CT system according to claim 3, characterized in that: In step S5, the grayscale distribution μ of the pixels inside the sub-pixel edge is p (s, h), statistics of the average grayscale value of defects under the same h value The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μ of the matrix material is obtained by the corrosion and expansion algorithm. q (s,h); The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ l (s,h)≠0; Among them, n represents the nth pixel at the same height; H represents the height of the triangular groove; In step S6, the simulated defect width is the distance between the hypotenuses of the triangle. Based on the triangle morphology characteristics, the defect size relationship D = L*(Hh) / h is established, and the defect grayscale μ b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l (h) is converted into the grayscale relationship μ about the defect size b (D); According to the measured grayscale distribution, the relationship curve 1 and the recognition probability curve 2 are drawn, and the horizontal coordinate of the intersection of the curve 1 and the curve 2 is the minimum identifiable defect size; The curve 1 is a graph showing the defect size D and The relationship between the recognition probability curve 2 is Where c is a constant between 2 and 5, indicating the probability of a defect being identified; Δp represents the voxel size; σ represents the noise value; In step S7, based on the triangle fitting data, the distance between the two hypotenuses at different height y values ​​and different h values ​​inside the triangle fitting is measured as the defect size, the difference is counted, and a relationship curve between the defect size and its measurement error is established.

7. A CT system defect detection capability evaluation system, characterized in that: The CT system defect detection capability comparison test block according to claim 1 or 2 is used for testing, comprising: Module M1: Preprocess the CT detection image, extract the defect area of ​​interest, and obtain the two-dimensional plane for defect detection capability evaluation; Module M2: Correction of CT image voxel size; Module M3: perform sub-pixel edge detection; Module M4: Fit the sub-pixel edge detection results to establish the pixel position coordinates of the entire slice image; Module M5: Obtain μ at different positions based on the results of sub-pixel edge detection l (h) and μ b (h); Module M6: Establish the relationship between defect size and grayscale, and plot the defect size D and The relationship curve 1 and the 50% probability recognition probability curve 2 of the human eye are used to obtain the minimum recognizable defect; Module M7: Establish the relationship curve between defect size and its measurement error.

8. The CT system defect detection capability evaluation system according to claim 7, characterized in that: The CT system defect detection capability comparison test block is placed on the product surface during the test, and the outer diameter is consistent with the maximum diameter of the product actually detected by the CT system, and the material is consistent with the material of the actual detected product. If the material is inconsistent, it is multiplied by the equivalent coefficient under the voltage used to convert the equivalent aluminum thickness; Each artificial defect is measured three times, and the average of the three measurement results is used as the standard value of the actual size; In the module M1, median filtering is used to preprocess the CT detection image, a defect segmentation algorithm is used to extract the simulated defect interest region A through image cropping, the canny operator is used to obtain the initial edge contour of the image, the position of the edge point of the image is confirmed, the least squares method is used to fit the edge region, and the area surrounded by the edge contour is extracted as the interest region B, which is recorded as the defect interest region; Slice images are sorted and segmented by three-dimensional threshold to roughly extract isosceles triangle grooves and circular groove models. The vertical direction of the scanning direction is the Z axis, and the direction away from the turntable is the positive direction. The first image with a pixel value of 1 is numbered as 1, and the numbering stops when the pixel value is 0. Select the slice images with the median number and two intervals apart as slice Z for defect detection capability evaluation; Define the plane perpendicular to the Z axis as the xy plane, and select the slice XY for defect detection capability evaluation in the same way; The module M2 adopts level set segmentation to obtain the sub-pixel point cloud of the central area plane, obtains the positions of the two circle centers based on least squares circle fitting, measures the distance between the circle centers, and corrects the voxel size error of the CT system according to the actual measured value of the distance between the circle centers.

9. The CT system defect detection capability evaluation system according to claim 7, characterized in that: In the module M3, the sobel operator is used to obtain the gradient image R(x, y), and sub-pixel edge detection is performed based on the Lagrange interpolation method in the gradient direction; Take three points R(x (i-1) ,y j )、R(x i ,y j )、R(x (i+1) ,y j ), take the x-direction gradient values ​​of these three points as the function values, respectively (x i -w)、x i 、(x i -w) is the interpolation point, and substituting it into the above formula, we get L(x) = R i-1 l0(x)+R i l1(x)+R i+1 l2(x); l j (x) represents the Lagrange polynomial, Among them, w represents the pixel direction spacing; x i represents the interpolation node; i, j represent the sequence number, i, j∈[0,N]; N represents the total number of gradient image points obtained by the Sobel operator; Differentiate L(x) and let The x coordinate of the sub-pixel edge is obtained as Take three points R(x i ,y (j-1) )、R(x i ,y j )、R(x i ,y (j+1) ), take the y-direction gradient values ​​of these three points as the function value, respectively (y i -w), y i ,(y i -w) is the interpolation point, and the y coordinate of the sub-pixel edge is In the module M4, a triangle fitting is performed on the point set of sub-pixel edge points through Hough transform, and the length L of the short side of the triangle, the angle θ between the short side and the hypotenuse, and the height H are measured. The pixel position (s, h) of the entire slice image is established with the short side of the triangle as the S axis, the perpendicular to the midline of the short side as the H axis, and the midpoint of the short side as the origin.

10. The CT system defect detection capability evaluation system according to claim 7, characterized in that: The grayscale distribution μ of the sub-pixel edge internal pixel in the module M5 p (s, h), statistics of the average grayscale value of defects under the same h value m l (s,h)≠0; The point set of the sub-pixel edge is expanded by 7 pixels and 2 pixels respectively using the expansion algorithm. The two expansion results are subtracted and intersected with the original image pixel by pixel. The grayscale distribution μ of the matrix material is obtained by the erosion and expansion algorithm. q (s,h); The average gray value of the matrix material under the same h value in the statistical area h∈[0,H+7] and μ l (s,h)≠0; Among them, n represents the nth pixel at the same height; H represents the height of the triangular groove; The simulated defect width in the module M6 is the distance between the hypotenuses of the triangle. Based on the triangle morphology characteristics, the defect size relationship D = L*(Hh) / h is established, and the defect grayscale μ b (h) is converted into the grayscale relationship μ about the defect size b (D), grayscale μ of matrix material l (h) is converted into the grayscale relationship μ about the defect size b (D); According to the measured grayscale distribution, the relationship curve 1 and the recognition probability curve 2 are drawn, and the horizontal coordinate of the intersection of the curve 1 and the curve 2 is the minimum identifiable defect size; The curve 1 is a graph showing the defect size D and The relationship between the recognition probability curve 2 is Where c is a constant between 2 and 5, indicating the probability of a defect being identified; Δp represents the voxel size; σ represents the noise value; In the module M7, based on the triangle fitting data, the distance between the two hypotenuses at different height y values ​​and different h values ​​inside the triangle fitting is measured as the defect size, the difference is counted, and a relationship curve between the defect size and its measurement error is established.

Images