A method for nondestructive testing of the quality of a continuously cast strand for steel billets

Abstract

The present application relates to the technical field of image analysis, and particularly relates to a continuous casting billet quality nondestructive testing method for billet casting, which comprises the following steps: obtaining an interference image corresponding to a continuous casting billet to be detected, and performing phase unwrapping on the interference image to obtain a target image; screening out a suspected defect region according to a phase value, a height value and a phase gradient value of a pixel point in the target image; determining a phase feature index, a height feature index, a direction consistency index and a defect contribution index corresponding to each suspected defect region, so as to determine a real crack defect index corresponding to each suspected defect region; and performing quality nondestructive testing on the continuous casting billet to be detected based on the real crack defect index. The present application quantifies the real crack defect index representing the crack defect condition by analyzing the interference image and the target image, so as to realize the quality nondestructive testing of the continuous casting billet and improve the accuracy of the quality testing of the continuous casting billet.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image analysis, and in particular to a continuous casting billet quality nondestructive testing method for billet casting. BACKGROUND

[0002] With the development of science and technology, image analysis is applied more and more widely, for example, it can be applied to continuous casting billet quality nondestructive testing of billet casting. At present, when the quality of an object is detected based on image analysis, the method adopted is: the defect condition of the object is analyzed based on the gray image of the object, so as to judge the quality of the object. Specifically, the defect area is often identified by different gray values, and the quality of the object is judged by the identified defect area.

[0003] However, when the defect area on the surface of the continuous casting billet is identified by different gray values, the following technical problems often exist:

[0004] In actual situations, the common defect on the surface of the continuous casting billet is crack defect, and the gray difference between the crack defect area and the normal area on the surface of the continuous casting billet is often not obvious. Therefore, when the crack defect is directly identified by different gray values, the crack defect pixel points may be misjudged, thereby causing poor accuracy of the continuous casting billet quality detection. SUMMARY

[0005] In order to solve the technical problem of poor accuracy of continuous casting billet quality detection, the present application provides a continuous casting billet quality nondestructive testing method for billet casting.

[0006] In the first aspect, the present application provides a continuous casting billet quality nondestructive testing method for billet casting, which comprises:

[0007] Obtaining an interference image corresponding to a continuous casting billet to be detected, and performing phase unwrapping on the interference image to obtain a target image;

[0008] According to the phase value, height value and phase gradient value of the pixel points in the target image, a suspected defect area is screened out from the target image;

[0009] According to the distribution of the phase gradient value and the phase gradient direction of the pixel points in each suspected defect area, a phase feature index corresponding to each suspected defect area is determined;

[0010] According to the distribution of the height value, height gradient value and height gradient direction of the pixel points in each suspected defect area, a height feature index corresponding to each suspected defect area is determined;

[0011] According to the proportion of the LBP value of the pixel points in each suspected defect area under different preset scales, a direction consistency index corresponding to each suspected defect area is determined;

[0012] According to the information entropy change of the LBP value of each pixel point in each suspected defect area under different preset scales, a defect contribution index corresponding to each suspected defect area is determined;

[0013] According to the phase feature index, the height feature index, the direction consistency index and the defect contribution index corresponding to each suspected defect area, a real crack defect index corresponding to each suspected defect area is determined.

[0014] Based on the real crack defect index, the quality of the to-be-detected continuous casting billet is nondestructively detected.

[0015] In combination with the above first aspect, in a possible implementation manner, the method further includes:

[0016] Based on the real crack defect index corresponding to all suspected defect areas in the target image, a real defect area is screened out from all suspected defect areas in the target image, and the real defect area in the target image is determined as a target defect area;

[0017] The interference image corresponding to each historical continuous casting billet obtained in advance is phase unwrapped to obtain a historical reference image, and a real defect area is identified from each historical reference image as a reference defect area;

[0018] From all the reference defect areas, a reference defect area matched with each target defect area is screened out to form a matching area set corresponding to each target defect area;

[0019] If the number of reference defect areas in the matching area set corresponding to the target defect area is not 0, the target defect area is determined as a current repeated risk area;

[0020] According to the area and value of all target defect areas, the real crack defect index corresponding to all target defect areas, the number of current repeated risk areas, and the number of reference defect areas in the matching area set corresponding to all current repeated risk areas, an abnormal early warning priority corresponding to the to-be-detected continuous casting billet is determined.

[0021] In combination with the above first aspect, in a possible implementation manner, the screening of the suspected defect area from the target image according to the phase value, the height value and the phase gradient value corresponding to the pixel point in the target image includes:

[0022] The difference between the phase value corresponding to each pixel point in the target image and the phase value before phase unwrapping is determined as a target phase difference corresponding to each pixel point in the target image.

[0023] determine a mean value of the height values corresponding to all pixel points in the target image as a height representative factor;

[0024] determine an absolute value of a difference between the height value corresponding to each pixel point in the target image and the height representative factor as a height difference factor corresponding to each pixel point in the target image;

[0025] determine a defect suspicion factor corresponding to each pixel point in the target image according to the target phase difference corresponding to each pixel point in the target image, the height difference factor and the phase gradient value;

[0026] if the defect suspicion factor corresponding to a pixel point is greater than a preset suspicion threshold, determine the pixel point as a defect suspicion pixel point;

[0027] filter out a suspected defect region from the target image based on all defect suspicion pixel points in the target image.

[0028] In a possible implementation manner of the first aspect, the determining of the phase feature index corresponding to each suspected defect region according to the distribution of the phase gradient value and the phase gradient direction of the pixel points in each suspected defect region comprises:

[0029] determining the phase feature index corresponding to each suspected defect region according to a range of the phase gradient values corresponding to all pixel points in each suspected defect region and information entropy of the phase gradient directions corresponding to all pixel points in each suspected defect region.

[0030] In a possible implementation manner of the first aspect, the determining of the height feature index corresponding to each suspected defect region according to the distribution of the height value, the height gradient value and the height gradient direction of the pixel points in each suspected defect region comprises:

[0031] determining any suspected defect region in the target image as a marked suspected defect region;

[0032] determining a main extension direction corresponding to the marked suspected defect region according to the distribution of the height gradient directions of the pixel points in the marked suspected defect region;

[0033] making a main extension pixel sequence corresponding to the marked suspected defect region according to the main extension direction;

[0034] determining a temporary difference value between the height gradient values corresponding to each adjacent pixel point in the main extension pixel sequence as a temporary difference value sequence corresponding to the marked suspected defect region;

[0035] constructing a temporary difference value segment from the temporary difference values with the same sign function value in the temporary difference value sequence.

[0036] determining the temporary difference segment with a non-constant 0 as a height change segment;

[0037] selecting a height change segment with the most temporary differences from all height change segments as a target height change segment;

[0038] selecting a pixel point with a continuous same height value from the main extension pixel sequence to form a contour pixel segment;

[0039] selecting a contour pixel segment with the most pixel points from all contour pixel segments as a target contour pixel segment;

[0040] determining a height feature index corresponding to the marked suspected defect region according to the number of temporary differences in the target height change segment, the number of pixel points in the target contour pixel segment, and the absolute value of the mean of all temporary differences in the target height change segment.

[0041] In a possible implementation manner of the first aspect, the determining the main extension direction corresponding to the marked suspected defect region according to the distribution of the height gradient directions corresponding to the pixel points in the marked suspected defect region comprises:

[0042] selecting a height gradient direction with the most number from the height gradient directions corresponding to all pixel points in the marked suspected defect region as a main gradient representative direction corresponding to the marked suspected defect region;

[0043] drawing a straight line perpendicular to the main gradient representative direction in the marked suspected defect region, denoted as a target perpendicular line corresponding to the marked suspected defect region;

[0044] determining an intersection point between the target perpendicular line and a horizontal line as a target intersection point, and segmenting the target perpendicular line by taking the target intersection point as a segmentation point to obtain two target rays;

[0045] determining a target included angle between the extension direction of each target ray and a horizontal right direction as a target included angle corresponding to each target ray;

[0046] selecting a target ray corresponding to a minimum target included angle from the two target rays as a reference ray;

[0047] determining the extension direction of the reference ray as the main extension direction corresponding to the marked suspected defect region.

[0048] In a possible implementation manner of the first aspect, the drawing the main extension pixel sequence corresponding to the marked suspected defect region according to the main extension direction comprises:

[0049] The target vertical line is translated within the marked suspected defect area, and the intersection segment of the target vertical line and the marked suspected defect area after each translation is determined as a candidate intersection segment;

[0050] The number of pixels in each candidate intersection segment is determined as the pixel count factor for each candidate intersection segment;

[0051] Select the candidate intersection segment with the largest corresponding pixel number factor from all candidate intersection segments and use it as the target intersection segment;

[0052] Along the main extension direction, the pixels in the target intersection segment are sorted to obtain the main extension pixel sequence corresponding to the marked suspected defect area.

[0053] In conjunction with the first aspect above, in one possible implementation, determining the orientation consistency index corresponding to each suspected defect region based on the proportion of LBP values ​​of pixels within each suspected defect region at different preset scales includes:

[0054] The percentage of pixels corresponding to the most frequent LBP value in each suspected defect area at each preset scale is determined as the target percentage of each suspected defect area at each preset scale.

[0055] The average percentage of the target area for each suspected defective region across all preset scales is used as the directional consistency index for each suspected defective region.

[0056] In conjunction with the first aspect above, in one possible implementation, determining the defect contribution index corresponding to each suspected defect region based on the information entropy change of the LBP values ​​of pixels within each suspected defect region at different preset scales includes:

[0057] Any suspected defect region in the target image is identified as a marked suspected defect region;

[0058] The information entropy of the LBP values ​​of all pixels in the suspected defect area under the same preset scale is determined as the LBP entropy of the label under that preset scale.

[0059] Sort all preset scales in ascending order to obtain a preset scale sequence;

[0060] The labeled LBP entropy at all preset scales in the preset scale sequence is used to form a labeled LBP entropy sequence;

[0061] The difference between each adjacent LBP entropy in the labeled LBP entropy sequence is determined as the target entropy difference.

[0062] If the target entropy difference value is less than a constant 0, the target entropy difference value is determined as a reference entropy negative value;

[0063] According to the number of reference entropy negative values and the absolute value of the mean of all reference entropy negative values, a defect contribution index corresponding to the marked suspected defect region is determined.

[0064] In combination with the first aspect, in a possible implementation manner, the filtering of the reference defect regions matched with each target defect region from all reference defect regions to form a matched region set corresponding to each target defect region comprises:

[0065] Any one target defect region is determined as a marked defect region, and a region in each historical reference image with the same position as the marked defect region is determined as a defect representative region;

[0066] The intersection of the defect representative region in each historical reference image and each reference defect region is determined as a marked overlapping region corresponding to each reference defect region in each historical reference image;

[0067] The area ratio of the marked overlapping region corresponding to each reference defect region in each historical reference image in the defect representative region is determined as a marked overlapping rate corresponding to each reference defect region in each historical reference image;

[0068] If the marked overlapping rate corresponding to the reference defect region is greater than a preset overlapping threshold, the reference defect region is determined as a matched region corresponding to the marked defect region;

[0069] All matched regions corresponding to the marked defect region form a matched region set corresponding to the marked defect region.

[0070] In a second aspect, the present application provides a continuous casting billet quality nondestructive testing system for billet casting, which comprises:

[0071] An image acquisition and phase unwrapping module is configured to acquire an interference image corresponding to the continuous casting billet to be detected, and perform phase unwrapping on the interference image to obtain a target image;

[0072] A suspected defect region screening module is configured to screen a suspected defect region from the target image according to the phase value, height value and phase gradient value of a pixel point in the target image;

[0073] A phase feature index determination module is configured to determine a phase feature index corresponding to each suspected defect region according to the distribution of the phase gradient value and phase gradient direction of a pixel point in each suspected defect region;

[0074] a height feature index determination module configured to determine a height feature index corresponding to each suspected defect region according to a distribution of height values, height gradient values and height gradient directions of the pixel points in each suspected defect region;

[0075] a direction consistency index determination module configured to determine a direction consistency index corresponding to each suspected defect region according to a proportion of LBP values of the pixel points in each suspected defect region under different preset scales;

[0076] a defect contribution index determination module configured to determine a defect contribution index corresponding to each suspected defect region according to a change in information entropy of LBP values of the pixel points in each suspected defect region under different preset scales;

[0077] a real crack defect index determination module configured to determine a real crack defect index corresponding to each suspected defect region according to the phase feature index, the height feature index, the direction consistency index and the defect contribution index corresponding to each suspected defect region;

[0078] a quality nondestructive detection module configured to perform quality nondestructive detection on the continuous casting billet to be detected based on the real crack defect index.

[0079] In a third aspect, a server is provided, including a memory and a processor. The memory is configured to store executable program code, and the processor is configured to call and run the executable program code from the memory, so that the device executes the method in the first aspect or any possible implementation manner of the first aspect.

[0080] In a fourth aspect, a computer program product is provided, which includes computer program code. When the computer program code is run on a computer, the computer program code causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0081] In a fifth aspect, a computer readable storage medium is provided, which stores computer program code. When the computer program code is run on a computer, the computer program code causes the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0082] The present application has the following beneficial effects:

[0083] This invention provides a non-destructive testing method for the quality of continuously cast billets used in steel billet casting. By analyzing interference images and target images, it quantifies the true crack defect indicators characterizing crack defects, thereby achieving non-destructive testing of continuously cast billet quality. This solves the technical problem of poor accuracy in continuously cast billet quality testing and improves the accuracy of continuous cast billet quality testing. Specifically, this invention analyzes the interference image corresponding to the continuously cast billet to be tested and its target image after phase unwrapping, quantifying multiple indicators related to crack defects, such as phase characteristic indicators, height characteristic indicators, direction consistency indicators, and defect contribution indicators. This quantifies the true crack defect indicators characterizing crack defects. Finally, based on the true crack defect indicators, non-destructive testing of continuously cast billet quality is achieved, and the accuracy of continuously cast billet quality testing is improved. Attached Figure Description

[0084] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0085] Figure 1 This is a flowchart of a non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to the present invention;

[0086] Figure 2 This is a schematic diagram of the composition and structure of a non-destructive testing system for the quality of continuously cast billets used in steel billet casting according to the present invention.

[0087] Figure 3 This is a schematic diagram of the structure of a computer device according to the present invention. Detailed Implementation

[0088] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0089] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0090] refer to Figure 1, shows the flow of some embodiments of the present application for a continuous casting billet quality nondestructive testing method for billet casting. The continuous casting billet quality nondestructive testing method for billet casting comprises the following steps:

[0091] Step S1, obtaining the interference image corresponding to the continuous casting billet to be detected, and phase unwrapping the interference image to obtain the target image.

[0092] Wherein, the continuous casting billet to be detected can be the billet casting continuous casting billet to be detected for quality detection.

[0093] As an example, this step can include the following steps:

[0094] First, obtain the interference image corresponding to the continuous casting billet to be detected.

[0095] For example, a monitoring point can be set 5-8 meters from the cutting machine outlet after cutting of the continuous casting machine, a CCD (Charge-Coupled Device, Charge-Coupled Device) camera is installed at the monitoring point, a green laser line (wavelength 532nm) emitted by a high-brightness laser line light source is projected onto the surface of the continuous casting billet to be detected by using CCD camera technology, the green light reflected by the surface of the continuous casting billet is collected by a line array CCD camera and converted into a digital image; the phase of the laser can be changed to form interference fringes on the surface of the billet, and the converted digital image is used to obtain the interference fringe image, i.e. the interference image; wherein the roll encoder triggers the collection once every 5mm movement, and four-step phase shift (pzt phase switching) is performed.

[0096] It should be noted that the interference image can reduce temperature interference to some extent.

[0097] Second, phase unwrapping the interference image to obtain the target image.

[0098] For example, Goldstein phase unwrapping algorithm can be used to phase unwrap the interference image, the surface height information can be restored from the interference image, and the phase unwrapped interference image is recorded as the target image. According to the height mapping, three-dimensional reconstruction can be carried out, and the height obtained is recorded as the height value of the pixel point in the target image.

[0099] It should be noted that the interference image reflects the height change of the surface of the steel continuous casting billet, and under normal circumstances, when the surface of the billet has a defect (such as a crack), since the crack is recessed compared to the normal surface, the height and phase of the crack area often have a mutation compared to the normal area, and the phase of the original interference image is wrapped, often ignoring the true period of the phase, so that the true height difference of the local surface may be covered up, therefore, combined with the above analysis, the original interference image is unwrapped and three-dimensional reconstruction is performed, and the area where the defect may exist is screened based on the phase and height change before and after reconstruction.

[0100] Step S2, screening a suspected defect area from the target image according to the phase value, height value and phase gradient value corresponding to the pixel point in the target image.

[0101] Among them, the phase value corresponding to the pixel point in the target image can be the phase value after phase unwrapping. The height value corresponding to the pixel point in the target image can be the surface height restored from the interference image. The phase gradient value corresponding to the pixel point in the target image can be the phase gradient value after phase unwrapping.

[0102] As an example, the present step can include the following steps:

[0103] Firstly, the difference between the phase value corresponding to each pixel point in the above target image and the phase value before phase unwrapping is determined as the target phase difference corresponding to each pixel point in the above target image.

[0104] Secondly, the mean value of the height values corresponding to all pixel points in the above target image is determined as the height representative factor.

[0105] Thirdly, the absolute value of the difference between the height value corresponding to each pixel point in the above target image and the above height representative factor is determined as the height difference factor corresponding to each pixel point in the above target image.

[0106] Fourthly, the defect suspected factor corresponding to each pixel point in the above target image is determined according to the target phase difference, height difference factor and phase gradient value corresponding to each pixel point in the above target image.

[0107] For example, the formula for determining the defect suspected factor corresponding to the pixel point in the target image can be:

[0108] ;

[0109] Among them, is the defect suspected factor corresponding to the i-th pixel point in the target image. i is the serial number of the pixel point in the target image. is a normalization function. is an absolute value function. is the height value corresponding to the i-th pixel point in the target image. h is a height representative factor, that is, the average value of the height values corresponding to all pixel points in the target image. is the height difference factor corresponding to the i-th pixel point in the target image. is the phase value corresponding to the i-th pixel point in the target image. The phase value corresponding to the pixel point in the target image is the phase value obtained after phase unwrapping of the pixel point in the interference image. is the phase value before phase unwrapping of the i-th pixel point in the target image. is the target phase difference corresponding to the i-th pixel point in the target image. is the phase gradient value corresponding to the i-th pixel point in the target image. is the average value of the phase gradient values corresponding to all pixel points in the target image.

[0110] It should be noted that if the phase difference of a point before and after three-dimensional reconstruction is large, or the phase difference before and after unwrapping the phase is large, and the height of the point after unwrapping the phase is greatly different from the overall image, and the phase gradient amplitude is also large, then the possibility of local mutation at the point is often large. may represent the possibility of local mutation at the i-th pixel point in the target image. The larger the value, the more likely it is that the i-th pixel point has a local mutation, and the more likely it is that the i-th pixel point is a defect pixel point.

[0111] In the fifth step, if the defect suspected factor corresponding to the pixel point is greater than a preset suspected threshold, the pixel point is determined as a defect suspected pixel point.

[0112] The preset suspected threshold can be a pre-set threshold, which can be 0.5.

[0113] In the sixth step, based on all defect suspected pixel points in the target image, a suspected defect region is screened from the target image.

[0114] For example, the region where the defect suspected pixel point is located can be extracted from the target image, and a morphological closing operation is performed to mark the obtained region as a suspected defect region.

[0115] It should be noted that since the surface of the billet is oxidized or interfered by water, it may also show a mutation in height, phase and other characteristics in the interference image due to local optical path mutation and other factors. Therefore, the suspected defect region can be a water stain region, an oxidation region or a crack defect region. Therefore, the real crack defect region needs to be further screened from the suspected defect region based on the rules of the crack defect.

[0116] Step S3, determining a phase feature index corresponding to each suspected defect region according to the distribution of the phase gradient value and the phase gradient direction corresponding to the pixel points in each suspected defect region.

[0117] The phase gradient value corresponding to the pixel points in the suspected defect region can be a phase gradient value after phase unwrapping. The phase gradient direction corresponding to the pixel points in the suspected defect region can be a phase gradient direction after phase unwrapping.

[0118] As an example, the phase feature index corresponding to each suspected defect region can be determined according to the range of the phase gradient values corresponding to all pixel points in each suspected defect region and the information entropy of the phase gradient directions corresponding to all pixel points in each suspected defect region.

[0119] For example, the formula corresponding to the phase feature index corresponding to the suspected defect region can be:

[0120] ;

[0121] wherein, is the phase feature index corresponding to the jth suspected defect region in the target image. j is the serial number of the suspected defect region in the target image. is the information entropy of the phase gradient directions corresponding to all pixel points in the jth suspected defect region in the target image. is an exponential function with a natural constant as the base. is the range of the phase gradient values corresponding to all pixel points in the jth suspected defect region in the target image.

[0122] It should be noted that if the difference between the phase gradient amplitudes in the region is relatively small and the gradient direction is relatively disordered, the region is often a crack defect region. When is larger, it often means that the gradient direction in the jth suspected defect region is more likely to present a disordered phenomenon. When is smaller, it often means that the difference between the phase gradient amplitudes in the jth suspected defect region is relatively smaller. Therefore, when is larger, it often means that the difference between the phase gradient amplitudes in the jth suspected defect region is relatively small and the gradient direction is relatively disordered, which often means that the jth suspected defect region is more likely to be a crack defect region.

[0123] Step S4, determining a height feature index corresponding to each suspected defect region according to the distribution of the height value, the height gradient value and the height gradient direction corresponding to the pixel points in each suspected defect region.

[0124] The height value corresponding to the pixel point in the suspected defect area can be the height after phase unwrapping, that is, the surface height restored from the interference image. The height gradient value corresponding to the pixel point in the suspected defect area can be the height gradient value after phase unwrapping. The height gradient direction corresponding to the pixel point in the suspected defect area can be the height gradient direction after phase unwrapping.

[0125] As an example, the present step can include the following steps:

[0126] Firstly, any one of the suspected defect areas in the target image is determined as a marked suspected defect area.

[0127] Secondly, according to the distribution of the height gradient directions corresponding to the pixel points in the marked suspected defect area, the main extension direction corresponding to the marked suspected defect area is determined.

[0128] If the marked suspected defect area is a crack defect area, the main extension direction corresponding to the marked suspected defect area is the main extension direction of the crack defect.

[0129] For example, the determination of the main extension direction corresponding to the marked suspected defect area can include the following sub-steps:

[0130] Firstly, the most common height gradient direction is selected from the height gradient directions corresponding to all the pixel points in the marked suspected defect area as the main gradient representative direction corresponding to the marked suspected defect area.

[0131] For example, if there are 560 pixel points in the marked suspected defect area, and the number of pixel points with the same height gradient direction as the first height gradient direction in the marked suspected defect area is 260; the number of pixel points with the same height gradient direction as the second height gradient direction in the marked suspected defect area is 160; and the number of pixel points with the same height gradient direction as the third height gradient direction in the marked suspected defect area is 140; then the main gradient representative direction corresponding to the marked suspected defect area is the first height gradient direction.

[0132] Secondly, a straight line perpendicular to the main gradient representative direction in the marked suspected defect area is drawn, which is recorded as the target perpendicular line corresponding to the marked suspected defect area.

[0133] Thirdly, the intersection point between the target perpendicular line and the horizontal line is determined as the target intersection point, and the target perpendicular line is divided into two target rays by the target intersection point.

[0134] Fourthly, the included angle between the extension direction of each target ray and the horizontal right direction is determined as the target included angle corresponding to each target ray.

[0135] A fifth sub-step, selecting a target ray with the minimum corresponding target angle from the two target rays as a reference ray.

[0136] A sixth sub-step, determining the extension direction of the reference ray as the main extension direction corresponding to the marked suspected defect region.

[0137] A third step, making a main extension pixel sequence corresponding to the marked suspected defect region according to the main extension direction.

[0138] If the marked suspected defect region is a crack defect region, the main extension pixel sequence corresponding to the marked suspected defect region can be composed of pixel points in the main extension direction of the crack defect.

[0139] For example, making the main extension pixel sequence corresponding to the marked suspected defect region can include the following sub-steps:

[0140] A first sub-step, translating the target perpendicular line in the marked suspected defect region, and determining the intersection segment of the target perpendicular line after each translation and the marked suspected defect region as a candidate intersection segment.

[0141] A second sub-step, determining the number of pixel points in each candidate intersection segment as a pixel number factor corresponding to each candidate intersection segment.

[0142] A third sub-step, selecting a candidate intersection segment with the maximum corresponding pixel number factor from all candidate intersection segments as a target intersection segment.

[0143] A fourth sub-step, sorting the pixel points in the target intersection segment along the main extension direction to obtain the main extension pixel sequence corresponding to the marked suspected defect region.

[0144] The serial numbers of the pixel points in the main extension pixel sequence gradually increase along the main extension direction.

[0145] A fourth step, determining the difference between the height gradient values corresponding to each adjacent pixel point in the main extension pixel sequence as a temporary difference value to obtain a temporary difference value sequence corresponding to the marked suspected defect region.

[0146] A fifth step, constructing a temporary difference value segment from temporary difference values with the same continuous sign function value in the temporary difference value sequence.

[0147] When the sign function value of a temporary difference value is 1, the temporary difference value is positive. When the sign function value of a temporary difference value is -1, the temporary difference value is negative. When the sign function value of a temporary difference value is 0, the temporary difference value is 0.

[0148] For example, if the temporary difference sequence is {-5, -1, -6, 1, 3, 0, 0, 0, 0, 0, 2, 2, 1, 2}, four temporary difference segments can be obtained, which are {-5, -1, -6}, {1, 3}, {0, 0, 0, 0, 0}, and {2, 2, 1, 2} respectively.

[0149] In the sixth step, the temporary difference segment with a corresponding sign function value that is not constant 0 is determined as a height change segment.

[0150] In the seventh step, the height change segment with the most temporary differences in all height change segments is selected as a target height change segment.

[0151] In the eighth step, the pixel points with consecutive same height values corresponding to the above main extension pixel sequence are selected to form a contour pixel segment.

[0152] In the ninth step, the contour pixel segment with the most pixel points in all contour pixel segments is selected as a target contour pixel segment.

[0153] In the tenth step, the height feature index corresponding to the marked suspected defect region is determined according to the number of temporary differences in the target height change segment, the number of pixel points in the target contour pixel segment, and the absolute value of the mean value of all temporary differences in the target height change segment.

[0154] For example, the formula for determining the height feature index corresponding to the marked suspected defect region can be:

[0155] ;

[0156] wherein, is the height feature index corresponding to the marked suspected defect region. w is the number of temporary differences in the target height change segment. is the absolute value function. d is the mean value of all temporary differences in the target height change segment. is the number of pixel points in the target contour pixel segment. is the number of pixel points in the main extension pixel sequence corresponding to the marked suspected defect region.

[0157] It should be noted that if a region has a height gradient amplitude that gradually changes along the suspected crack extension direction and has a large change range, and there are multiple consecutive height invariant points along the suspected crack extension direction, the region is more likely to be a real crack defect region. Therefore, when the greater the value of the height feature index corresponding to the marked suspected defect region, the more likely the marked suspected defect region is a real crack defect region.

[0158] Step S5, according to the proportion of the LBP value of each pixel point in each suspected defect area under different preset scales, determine the direction consistent index corresponding to each suspected defect area.

[0159] Wherein, the preset scale can be a pre-set scale, representing the radius of the neighborhood (Radius). It determines how far the pixel points participating in the calculation are from the center point. The larger the radius, the more macro the captured texture pattern is, and the larger the scale is. The adjacent preset scales can differ by 1. The value range of the preset scale can be [1, R]. R can be equal to the maximum value among the distances between different edge pixel points in the suspected defect area. The method for obtaining the LBP value of the pixel point under the preset scale can be: under the preset scale, the LBP (Local Binary Pattern, Local Binary Pattern) operator is used to obtain the LBP value of the pixel point under the preset scale.

[0160] As an example, the present step can include the following steps:

[0161] Firstly, the proportion of the pixel points corresponding to the LBP value that appears most frequently in each suspected defect area under each preset scale is determined as the target proportion of each suspected defect area under each preset scale.

[0162] For example, the formula for determining the target proportion of the suspected defect area under different preset scales can be:

[0163] ;

[0164] Wherein, is the target proportion of the jth suspected defect area in the target image under the ath preset scale. j is the serial number of the suspected defect area in the target image. a is the serial number of the preset scale. is the number of pixel points corresponding to the LBP value that appears most frequently in the jth suspected defect area in the target image under the ath preset scale. is the number of pixel points in the jth suspected defect area in the target image.

[0165] Secondly, the average value of the target proportion of each suspected defect area under all preset scales is determined as the direction consistent index corresponding to each suspected defect area.

[0166] It should be noted that in actual situations, the oxidation, water stains and the like present overall changes (oxidation or water stains cause uniform changes in the overall reflectivity or refractivity of the surface) in the optical path, and therefore the LBP values of the oxidation, water stains and the like at the same scale are relatively random. The real crack is a linear recess, and the crack edge has a refractive mutation in the light path, so that the gradient direction of the phase of the real crack is relatively concentrated, and therefore the LBP values of the crack defect area at the same scale are relatively uniform. Secondly, the LBP value to some extent expresses the directionality of the light and shade distribution at each point, and if there are more pixel points corresponding to the LBP value equal to the main mode in the region, the LBP direction in the region is more consistent. Therefore, when the direction consistency index corresponding to the suspected defect area is larger, it is often indicated that the suspected defect area is more likely to be a crack defect area.

[0167] Step S6, according to the information entropy change of the LBP value of each pixel point in the suspected defect area at different preset scales, determine the defect contribution index corresponding to each suspected defect area.

[0168] As an example, the present step can include the following steps:

[0169] Firstly, any one of the above target image suspected defect areas is determined as a marked suspected defect area.

[0170] Secondly, the information entropy of the LBP value of all pixel points in the above marked suspected defect area at the same kind of preset scale is determined as the marked LBP entropy at the kind of preset scale.

[0171] Thirdly, all preset scales are sorted in ascending order to obtain a preset scale sequence.

[0172] Fourthly, the marked LBP entropy at all preset scales in the above preset scale sequence is used to form a marked LBP entropy sequence.

[0173] Fifthly, the difference between each adjacent marked LBP entropy in the above marked LBP entropy sequence is determined as a target entropy difference.

[0174] Sixthly, if the target entropy difference is less than a constant 0, the target entropy difference is determined as a reference entropy negative value.

[0175] Seventhly, according to the number of reference entropy negative values and the absolute value of the mean value of all reference entropy negative values, the defect contribution index corresponding to the above marked suspected defect area is determined.

[0176] For example, the formula for determining the defect contribution index corresponding to the marked suspected defect area can be:

[0177] ;

[0178] Wherein, is a defect contribution indicator corresponding to the suspected defect region. is an absolute value function. M is the mean of all reference entropy negative values. is the number of reference entropy negative values. is the number of target entropy difference values.

[0179] It should be noted that the real crack is a linear depression, and there is a light path refraction mutation at the crack edge, so that the gradient direction of the real crack phase is relatively concentrated, and the entropy value of the LBP value of the real crack region usually decreases with the increase of the scale, and the greater the decrease, the more it indicates that with the increase of the LBP scale, the defect characteristics of the current region can be more clearly reflected. Therefore, when is greater, it often means that the suspected defect region is more likely to be a crack defect region.

[0180] Step S7, determining a real crack defect indicator corresponding to each suspected defect region according to the phase feature indicator, the height feature indicator, the direction consistency indicator and the defect contribution indicator corresponding to each suspected defect region.

[0181] As an example, the present step can include the following steps:

[0182] Firstly, the LBP main mode value of each suspected defect region at each preset scale is determined as the LBP value that appears most frequently at each preset scale.

[0183] Secondly, the real crack defect indicator corresponding to each suspected defect region is determined according to the phase feature indicator, the height feature indicator, the direction consistency indicator and the defect contribution indicator corresponding to each suspected defect region, and the LBP main mode value of each suspected defect region at different preset scales.

[0184] For example, the formula for determining the real crack defect indicator corresponding to the suspected defect region can be:

[0185] ;

[0186] wherein, is a real crack defect indicator corresponding to the jth suspected defect region in the target image. j is the serial number of the suspected defect region in the target image. is a normalization function. is a phase feature indicator corresponding to the jth suspected defect region in the target image. is a height feature indicator corresponding to the jth suspected defect region in the target image. is a direction consistency indicator corresponding to the jth suspected defect region in the target image. is a defect contribution indicator corresponding to the jth suspected defect region in the target image. is an exponential function with a natural constant as a base. is a variance of the LBP primary pattern value of the jth suspected defect region in the target image at all preset scales.

[0187] It should be noted that when is smaller, the influence of the LBP scale on the LBP value of the jth suspected defect region is greater. When is greater, it often indicates that the jth suspected defect region is more likely to be a crack defect region.

[0188] Step S8, based on the real crack defect index, performing nondestructive testing on the continuous casting billet to be detected.

[0189] As an example, the present step can include the following steps:

[0190] First, if the real crack defect index corresponding to the suspected defect region is greater than the preset defect threshold, the suspected defect region is determined as a real defect region.

[0191] Wherein, the preset defect threshold can be a preset threshold, which can be 0.5.

[0192] Second, if there is a real defect region in the target image, it is determined that the quality of the continuous casting billet to be detected is unqualified.

[0193] Optionally, all the continuous casting billets in the same batch are determined to have real defect regions according to the above method, wherein when multiple continuous casting billets in the same batch have real defects at similar positions, it is more likely that there is a systematic process problem in the current continuous casting process, such as mold scratching, secondary cooling zone nozzle blockage, etc., rather than random individual defects, so it is more necessary to give an early warning. For a single continuous casting billet, if there are more repeated risk regions in the current continuous casting billet, and the area of the repeated risk region is larger, the distribution position is more concentrated, it is more likely to cause stress concentration, and the possibility of each region being a real defect region is larger, so the damage of the region is more serious, and it is more necessary to give an early warning. Therefore, the embodiments of the present application can further include the following steps:

[0194] First, based on the real crack defect index corresponding to all suspected defect regions in the above target image, the real defect regions are screened from all suspected defect regions in the above target image, and the real defect regions in the above target image are determined as target defect regions.

[0195] Second, the phase unwrapping is performed on each historical continuous casting billet corresponding to the interference image obtained in advance to obtain a historical reference image, and the real defect region is identified from each historical reference image as a reference defect region.

[0196] The historical continuous casting billet can be a continuous casting billet of the same batch as the continuous casting billet to be detected, that is, the historical continuous casting billet and the continuous casting billet to be detected can be of the same specification and model.

[0197] It should be noted that the identification method of the real defect area in the historical reference image can be the same as the identification method of the real defect area in the target image, which will not be described here.

[0198] The third step is to screen the reference defect area matched with each target defect area from all reference defect areas to form a matching area set corresponding to each target defect area, which can include the following sub-steps:

[0199] The first sub-step is to determine any one target defect area as a marker defect area, and determine the area in each historical reference image which is the same as the position of the marker defect area as a defect representative area.

[0200] The second sub-step is to determine the intersection of the defect representative area in each historical reference image and each reference defect area as the marker overlap area corresponding to each reference defect area in each historical reference image.

[0201] The third sub-step is to determine the area ratio of the marker overlap area corresponding to each reference defect area in each historical reference image within the defect representative area as the marker overlap rate corresponding to each reference defect area in each historical reference image.

[0202] The fourth sub-step is to determine the reference defect area as the matching area corresponding to the marker defect area if the marker overlap rate corresponding to the reference defect area is greater than a preset overlap threshold.

[0203] The preset overlap threshold can be a preset threshold, which can be 0.7.

[0204] The fifth sub-step is to form a matching area set corresponding to the marker defect area by using all the matching areas corresponding to the marker defect area.

[0205] It should be noted that the defect represented by the marker defect area and the area in the matching area set corresponding to the marker defect area can be the same, and the more areas in the matching area set corresponding to the marker defect area, the greater the probability of repeated occurrence of the defect represented by the marker defect area.

[0206] The fourth step is to determine the target defect area as the current repeated risk area if the number of reference defect areas in the matching area set corresponding to the target defect area is not 0.

[0207] It should be noted that the current repeated risk area often represents a repeated defect.

[0208] In the fifth step, according to the area and value of all target defect regions, the real crack defect index corresponding to all target defect regions, the number of current repetitive risk regions, and the number of reference defect regions in the matching region set corresponding to all current repetitive risk regions, the abnormal early warning priority corresponding to the to-be-detected continuous casting billet is determined.

[0209] It should be noted that the greater the abnormal early warning priority corresponding to the to-be-detected continuous casting billet, the more likely it is to be warned.

[0210] For example, determining the abnormal early warning priority corresponding to the to-be-detected continuous casting billet can include the following sub-steps:

[0211] In the first sub-step, the number of reference defect regions in the matching region set corresponding to each current repetitive risk region is determined as the repetitive defect factor corresponding to each current repetitive risk region.

[0212] In the second sub-step, the number of historical continuous casting billets is determined as the reference representative number.

[0213] In the third sub-step, the ratio of the repetitive defect factor corresponding to each current repetitive risk region to the reference representative number is determined as the repetitive defect proportion corresponding to each current repetitive risk region.

[0214] In the fourth sub-step, the formula of the abnormal early warning priority corresponding to the to-be-detected continuous casting billet can be:

[0215] ;

[0216] Wherein, p is the abnormal early warning priority corresponding to the to-be-detected continuous casting billet. is a normalization function. is the mean value of the real crack defect indexes corresponding to all target defect regions. m is the area and value of all target defect regions. S is the number of current repetitive risk regions. N is the mean value of the repetitive defect proportions corresponding to all current repetitive risk regions. is an exponential function with a natural constant as the base. T is the mean value of the distances between the center points of different target defect regions.

[0217] Reference Figure 2 Based on the same inventive concept as the above method embodiments, the present application provides a continuous casting billet quality nondestructive testing system for billet casting, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the above computer program is executed by the processor, the steps of a continuous casting billet quality nondestructive testing method for billet casting are implemented, which can specifically include:

[0218] The image acquisition and phase unwrapping module 201 is configured to acquire an interference image corresponding to the continuous casting billet to be detected, and perform phase unwrapping on the interference image to obtain a target image.

[0219] The suspected defect area screening module 202 is configured to screen a suspected defect area from the target image according to a phase value, a height value and a phase gradient value corresponding to a pixel point in the target image.

[0220] The phase feature index determination module 203 is configured to determine a phase feature index corresponding to each suspected defect area according to a distribution of the phase gradient value and a phase gradient direction corresponding to a pixel point in each suspected defect area.

[0221] The height feature index determination module 204 is configured to determine a height feature index corresponding to each suspected defect area according to a distribution of the height value, a height gradient value and a height gradient direction corresponding to a pixel point in each suspected defect area.

[0222] The direction consistency index determination module 205 is configured to determine a direction consistency index corresponding to each suspected defect area according to a proportion of an LBP value of a pixel point in each suspected defect area under different preset scales.

[0223] The defect contribution index determination module 206 is configured to determine a defect contribution index corresponding to each suspected defect area according to a change of an information entropy of the LBP value of a pixel point in each suspected defect area under different preset scales.

[0224] The real crack defect index determination module 207 is configured to determine a real crack defect index corresponding to each suspected defect area according to the phase feature index, the height feature index, the direction consistency index and the defect contribution index corresponding to each suspected defect area.

[0225] The quality nondestructive detection module 208 is configured to perform quality nondestructive detection on the continuous casting billet to be detected based on the real crack defect index.

[0226] Figure 3 is a structural schematic diagram of a computer device provided by an embodiment of the present application. As shown in the example, Figure 3 the computer device 300 includes a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302, wherein the processor 302 executes the computer program 303, so that the computer device can execute any one of the aforementioned continuous casting billet quality nondestructive detection methods for steel billet casting.

[0227] Based on the same inventive concept as the above method embodiments, the present application provides a server comprising a memory and a processor. The memory is configured to store executable program code, and the processor is configured to call and run the executable program code from the memory, so that the device executes any one of the above-mentioned non-destructive testing methods for the quality of the continuous casting billet for steel billet casting.

[0228] Based on the same inventive concept as the above method embodiments, the present application provides a computer program product comprising: computer program code, which, when executed on a computer, causes the computer to execute any one of the above-mentioned non-destructive testing methods for the quality of the continuous casting billet for steel billet casting.

[0229] Based on the same inventive concept as the above method embodiments, the present application provides a computer readable storage medium storing computer program code, which, when executed on a computer, causes the computer to execute any one of the above-mentioned non-destructive testing methods for the quality of the continuous casting billet for steel billet casting.

[0230] In summary, by analyzing the interference image corresponding to the continuous casting billet to be detected and the target image after phase unwrapping, the present application quantifies a plurality of indicators related to the crack defect condition, such as the phase feature indicator, the height feature indicator, the direction consistency indicator and the defect contribution indicator, thereby quantifying the real crack defect indicator representing the crack defect condition. Finally, based on the real crack defect indicator, the present application realizes non-destructive testing of the quality of the continuous casting billet and improves the accuracy of the quality detection of the continuous casting billet.

[0231] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A non-destructive testing method for the quality of continuously cast billets used in steel billet casting, characterized in that, Includes the following steps: The interference image corresponding to the continuous casting billet to be detected is obtained, and the phase unwrapping of the interference image is performed to obtain the target image; Based on the phase value, height value, and phase gradient value of the pixels in the target image, suspected defect areas are selected from the target image. Based on the distribution of phase gradient values ​​and phase gradient directions of pixels within each suspected defect region, the phase feature index corresponding to each suspected defect region is determined. Based on the distribution of height value, height gradient value, and height gradient direction of pixels within each suspected defect area, the height feature index corresponding to each suspected defect area is determined. Based on the proportion of LBP values ​​of pixels in each suspected defect area at different preset scales, determine the orientation consistency index corresponding to each suspected defect area. Based on the information entropy change of LBP values ​​of pixels in each suspected defect area at different preset scales, the defect contribution index corresponding to each suspected defect area is determined. Based on the phase characteristic index, height characteristic index, direction consistency index and defect contribution index corresponding to each suspected defect area, the actual crack defect index corresponding to each suspected defect area is determined. Based on the actual crack defect index, non-destructive testing of the quality of the continuous casting billet to be tested is carried out. The step of filtering out suspected defect areas from the target image based on the phase value, height value, and phase gradient value corresponding to pixels in the target image includes: The difference between the phase value corresponding to each pixel in the target image and its phase value before phase unwrapping is determined as the target phase difference corresponding to each pixel in the target image; The average height value of all pixels in the target image is determined as the height representative factor. The absolute value of the difference between the height value corresponding to each pixel in the target image and the height representative factor is determined as the height difference factor corresponding to each pixel in the target image; Based on the target phase difference, height difference factor and phase gradient value corresponding to each pixel in the target image, the defect suspicion factor corresponding to each pixel in the target image is determined. The larger the target phase difference, height difference factor and phase gradient value, the larger the defect suspicion factor. If the defect suspicion factor corresponding to a pixel is greater than the preset suspicion threshold, then the pixel is identified as a defect suspicion pixel. Based on all suspected defect pixels in the target image, suspected defect areas are selected from the target image.

2. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 1, characterized in that, The method further includes: Based on the actual crack defect index corresponding to all suspected defect areas in the target image, the actual defect areas are screened out from all suspected defect areas in the target image, and the actual defect areas in the target image are determined as the target defect areas. Phase unwrapping is performed on the interference image corresponding to each historical continuous casting billet obtained in advance to obtain a historical reference image, and the real defect area is identified from each historical reference image as a reference defect area; From all reference defect regions, select the reference defect regions that match each target defect region to form a set of matching regions corresponding to each target defect region; If the number of reference defect areas in the matching area set corresponding to the target defect area is not 0, then the target defect area is determined as the current repeated risk area. The abnormal warning priority corresponding to the continuously cast billet to be inspected is determined based on the area and value of all target defect areas, the actual crack defect index corresponding to all target defect areas, the number of current repeated risk areas, and the number of reference defect areas in the matching area set corresponding to all current repeated risk areas.

3. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 1, characterized in that, The step of determining the phase feature index corresponding to each suspected defect region based on the distribution of phase gradient values ​​and phase gradient directions of pixels within each suspected defect region includes: Based on the range of the phase gradient values ​​of all pixels in each suspected defect region and the information entropy of the phase gradient direction of all pixels in each suspected defect region, the phase feature index corresponding to each suspected defect region is determined.

4. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 1, characterized in that, The height feature index corresponding to each suspected defect region is determined based on the distribution of the height value, height gradient value, and height gradient direction of the pixels within each suspected defect region, including: Any suspected defect region in the target image is identified as a marked suspected defect region; Based on the distribution of the height gradient direction corresponding to the pixels within the marked suspected defect area, the main extension direction corresponding to the marked suspected defect area is determined; Based on the main extension direction, construct the main extension pixel sequence corresponding to the marked suspected defect area; The difference between the height gradient values ​​corresponding to each adjacent pixel in the main extended pixel sequence is determined as a temporary difference, and a temporary difference sequence corresponding to the marked suspected defect area is obtained. Temporary difference segments are formed by taking temporary differences with consecutive identical sign function values ​​in the temporary difference sequence. The temporary difference segment whose corresponding symbolic function value is not a constant 0 is identified as the height variation segment; Select the height change segment with the largest temporary difference from all height change segments and use it as the target height change segment. Pixels with consecutive height values ​​are selected from the main extended pixel sequence to form equal-height pixel segments; Select the segment with the most pixels from all segments of equal height as the target segment of equal height. The height feature index corresponding to the marked suspected defect area is determined based on the number of temporary differences in the target height variation segment, the number of pixels in the target equal height pixel segment, and the absolute value of the mean of all temporary differences in the target height variation segment.

5. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 4, characterized in that, The step of determining the main extension direction corresponding to the marked suspected defect region based on the distribution of the height gradient direction corresponding to the pixels within the marked suspected defect region includes: The height gradient direction with the largest number of pixels in the marked suspected defect area is selected as the main gradient representative direction corresponding to the marked suspected defect area. Within the marked suspected defect area, draw a straight line perpendicular to the direction represented by the principal gradient, and denote it as the target perpendicular line corresponding to the marked suspected defect area; The intersection point between the target vertical line and the horizontal line is determined as the target intersection point, and the target vertical line is divided into two target rays using the target intersection point as the dividing point. The angle between the extension direction of each target ray and the horizontal rightward direction is determined as the target angle corresponding to each target ray; The target ray with the smallest included angle between the two target rays is selected as the reference ray. The extension direction of the reference ray is determined as the main extension direction corresponding to the marked suspected defect area.

6. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 5, characterized in that, The step of creating the main extension pixel sequence corresponding to the marked suspected defect region according to the main extension direction includes: The target vertical line is translated within the marked suspected defect area, and the intersection segment of the target vertical line and the marked suspected defect area after each translation is determined as a candidate intersection segment; The number of pixels in each candidate intersection segment is determined as the pixel count factor for each candidate intersection segment; Select the candidate intersection segment with the largest corresponding pixel number factor from all candidate intersection segments and use it as the target intersection segment; Along the main extension direction, the pixels in the target intersection segment are sorted to obtain the main extension pixel sequence corresponding to the marked suspected defect area.

7. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 1, characterized in that, The step of determining the orientation consistency index for each suspected defect region based on the proportion of LBP values ​​of pixels within each suspected defect region at different preset scales includes: The percentage of pixels corresponding to the most frequent LBP value in each suspected defect area at each preset scale is determined as the target percentage of each suspected defect area at each preset scale. The average percentage of the target area for each suspected defective region across all preset scales is used as the directional consistency index for each suspected defective region.

8. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 1, characterized in that, The step of determining the defect contribution index corresponding to each suspected defect region based on the information entropy change of the LBP value of pixels within each suspected defect region at different preset scales includes: Any suspected defect region in the target image is identified as a marked suspected defect region; The information entropy of the LBP values ​​of all pixels in the suspected defect area under the same preset scale is determined as the LBP entropy of the label under that preset scale. Sort all preset scales in ascending order to obtain a preset scale sequence; The labeled LBP entropy at all preset scales in the preset scale sequence is used to form a labeled LBP entropy sequence; The difference between each adjacent LBP entropy in the labeled LBP entropy sequence is determined as the target entropy difference. If the target entropy difference is less than the constant 0, then the target entropy difference is determined as the negative value of the reference entropy; Based on the number of negative reference entropy values ​​and the absolute value of the mean of all negative reference entropy values, the defect contribution index corresponding to the marked suspected defect area is determined.

9. The non-destructive testing method for the quality of continuously cast billets used in steel billet casting according to claim 2, characterized in that, The step of selecting reference defect regions that match each target defect region from all reference defect regions to form a matching region set corresponding to each target defect region includes: Any target defect area is identified as a marked defect area, and the area in each historical reference image that is at the same position as the marked defect area is identified as a defect representative area. The intersection of the defect-representing region in each historical reference image and each reference defect region is used to determine the marked overlapping region corresponding to each reference defect region in each historical reference image. The area ratio of the overlapping region of the marker corresponding to each reference defect region in each historical reference image within the defect representative region is determined as the marker overlap rate corresponding to each reference defect region in each historical reference image. If the overlap rate of the markers corresponding to the reference defect region is greater than the preset overlap threshold, then the reference defect region is determined as the matching region corresponding to the marked defect region. All matching regions corresponding to the marked defect region are combined to form the matching region set corresponding to the marked defect region.

Images