CN116645367B - Steel plate cutting quality detection method for high-end manufacturing - Google Patents
Steel plate cutting quality detection method for high-end manufacturing Download PDFInfo
- Publication number
- CN116645367B CN116645367B CN202310926389.1A CN202310926389A CN116645367B CN 116645367 B CN116645367 B CN 116645367B CN 202310926389 A CN202310926389 A CN 202310926389A CN 116645367 B CN116645367 B CN 116645367B
- Authority
- CN
- China
- Prior art keywords
- image
- value
- connected domain
- gray
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 129
- 239000010959 steel Substances 0.000 title claims abstract description 129
- 238000005520 cutting process Methods 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 title claims abstract description 22
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 19
- 238000009826 distribution Methods 0.000 claims abstract description 85
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000003064 k means clustering Methods 0.000 claims abstract description 34
- 238000012216 screening Methods 0.000 claims abstract description 10
- 238000004891 communication Methods 0.000 claims abstract description 8
- 230000007547 defect Effects 0.000 claims description 35
- 238000003708 edge detection Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 4
- 230000008859 change Effects 0.000 description 23
- 238000004364 calculation method Methods 0.000 description 5
- 238000003698 laser cutting Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000002950 deficient Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/457—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by analysing connectivity, e.g. edge linking, connected component analysis or slices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/54—Extraction of image or video features relating to texture
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image processing, in particular to a steel plate cutting quality detection method for high-end manufacturing. The method comprises the following steps: acquiring a gray image of a section of the steel plate; obtaining corresponding stability according to the position distribution of each pixel point in each communication domain in the gray level image and the gray level difference between each pixel point and the adjacent pixel points; determining a distribution interval of texture characteristic values of each connected domain based on the stability degree and the gray value of the pixel point; screening a target historical image according to the distribution interval, the gray values of the pixel points in the gray images and the gray values of the pixel points in each historical image in the image dataset; and clustering pixel points in the gray level image based on a target K value corresponding to the target historical image in K-means clustering and a distribution interval of texture characteristic values of each connected domain in the gray level image to obtain a clustering result, and further judging whether the cutting quality of the steel plate is qualified. The invention improves the accuracy of the steel plate cutting quality detection result.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a steel plate cutting quality detection method for high-end manufacturing.
Background
The laser cutting technology can cut steel plates accurately at high speed and high quality, and is important in the whole cutting field. Under the industrial environment, the laser cutting technology cannot always keep proper cutting speed and power, the condition of poor cutting quality inevitably exists, the surface of a section can have cracks with different depths, and the cracks have great influence on the quality evaluation of the steel plate for high-end manufacturing, so that the cut steel plate for high-end manufacturing after cutting needs to be subjected to cutting quality detection so as to be convenient for timely adjusting the cutting method of the steel plate.
The existing cutting quality detection for the surface of the steel plate generally adopts a K-means clustering algorithm, but the algorithm depends on K selection and is over sensitive to noise, and different initial clustering centers can lead to different finally obtained clustering results, so that the defect quantity can only be roughly reflected, the defect degree can not be quantified, and further the cutting quality of the steel plate can not be accurately evaluated.
Disclosure of Invention
In order to solve the problem of lower accuracy of a detection result when detecting the cutting quality of a steel plate in the existing method, the invention aims to provide a steel plate cutting quality detection method for high-end manufacturing, and the adopted technical scheme is as follows:
The invention provides a steel plate cutting quality detection method for high-end manufacturing, which comprises the following steps:
acquiring a gray image of a section of a steel plate to be detected;
sliding on the gray level image by utilizing a sliding window to obtain each window area, and carrying out edge detection on each window area to obtain each connected domain in each window area; obtaining the corresponding stability of each connected domain according to the position distribution of each pixel point in each connected domain and the gray difference between each pixel point and the adjacent pixel point; determining a distribution interval of texture characteristic values of each connected domain based on the stability and the gray value of the pixel point;
screening a target historical image from an image dataset according to the distribution interval of the texture characteristic values of each connected domain in the gray image, the gray value of the pixel point in the gray image and the gray value of the pixel point in each historical image in the image dataset;
based on a target K value corresponding to a target historical image in K-means clustering and a distribution interval of texture characteristic values of each connected domain in the gray image, clustering pixel points in the gray image by adopting a K-means clustering algorithm to obtain a clustering result; and judging whether the cutting quality of the steel plate to be detected is qualified or not by combining the clustering result and the distribution interval of the texture characteristic values of each connected domain in the target historical image.
Preferably, the obtaining of the size of the sliding window includes:
threshold segmentation is carried out on the gray level image to obtain each target subarea, and the minimum circumscribed rectangle of each target subarea is respectively obtained; calculating the average width of all the minimum circumscribed rectangles;
and setting the length and the width of the sliding window to be preset multiples of the average width.
Preferably, the obtaining the stability degree corresponding to each connected domain according to the position distribution of each pixel point in each connected domain and the gray scale difference between each pixel point and the adjacent pixel point comprises:
for any communication domain:
for any row in the connected domain: determining the gray level difference of two adjacent pixel points of each pixel point of the row as a transverse difference value of the corresponding pixel point; respectively calculating square values of all the transverse differential values of the row, and marking the sum of the square values of all the transverse differential values of the row as a first index corresponding to the row; determining the ratio of the first index to the average width of all the minimum circumscribed rectangles as a line gray scale variation parameter of the line;
for any column in the connected domain: the gray level difference of two adjacent pixels of each pixel of the row is determined as a longitudinal difference value of the corresponding pixel; respectively calculating the square value of each longitudinal differential value of the row, and marking the sum of the square values of all the longitudinal differential values of the row as a second index corresponding to the column; determining the ratio of the second index to the average width of all the minimum circumscribed rectangles as a column gray scale variation parameter of the column;
Performing linear detection on the connected domain to obtain each linear segment in the connected domain;
and obtaining the corresponding stability degree of the connected domain according to the direction of each straight line segment in the connected domain, the line gray scale change parameters of each line and the column gray scale change parameters of each column.
Preferably, the obtaining the corresponding stability of the connected domain according to the direction of each straight line segment in the connected domain, the line gray scale variation parameter of each line and the column gray scale variation parameter of each column includes:
calculating the variance of the line gray scale variation parameters of all the lines according to the line gray scale variation parameters of each line in the connected domain; determining the variance of the line gray scale variation parameter as the transverse stability of the connected domain;
calculating the variance of the column gray scale variation parameters of all columns according to the column gray scale variation parameters of each column in the connected domain; determining the variance of the column gray scale variation parameters as the longitudinal stability of the connected domain;
respectively counting the number of straight line segments in the transverse direction and the number of straight lines in the longitudinal direction in the connected domain, and determining the direction with the maximum number of straight line segments in the two directions as a target direction;
and obtaining the corresponding stability degree of the connected domain based on the transverse stability, the longitudinal stability, the target direction and the preset weight.
Preferably, determining the distribution interval of the texture feature value of each connected domain based on the stability and the gray value of the pixel point includes:
for any communication domain:
respectively acquiring an average gray value, a maximum gray value and a minimum gray value of all the pixel points in each target direction in the connected domain according to the gray value of each pixel point in each target direction in the connected domain;
obtaining a lower limit value and an upper limit value of a distribution interval of the texture characteristic value of the connected domain according to the average gray value, the maximum gray value, the minimum gray value and the corresponding stability degree of the connected domain; and obtaining a distribution interval of the connected domain texture characteristic value based on the lower limit value and the upper limit value.
Preferably, the lower limit value and the upper limit value of the distribution interval of the connected domain texture feature value are calculated by adopting the following formula:
,
wherein,a lower limit value of a distribution interval of the connected domain texture characteristic value, < ->K is the number of target directions in the connected domain, which is the upper limit value of the distribution interval of the texture characteristic value of the connected domain>For the average gray value of all pixels in the kth target direction in the connected domain, +.>For the minimum gray value of all pixels in the kth target direction in the connected domain, +. >And C is the corresponding stability of the connected domain, and is the maximum gray value of all pixel points in the kth target direction in the connected domain.
Preferably, the screening the target historical image from the image dataset according to the distribution interval of the texture feature value of each connected domain in the grayscale image, the gray value of the pixel point in the grayscale image, and the gray value of the pixel point in each historical image in the image dataset includes:
for any window region in the grayscale image: determining the minimum value in the lower limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic minimum value of the window area; determining the maximum value in the upper limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic maximum value of the window area; obtaining a texture feature interval of the window area based on the texture feature minimum value and the texture feature maximum value;
the number of the lower limit values of the texture feature intervals of all window areas in the gray level image is recorded as a first mode corresponding to the gray level image; recording the numbers of the upper limit values of the texture feature intervals of all window areas in the gray level image as second numbers corresponding to the gray level image; recording the number of the lower limit values of the texture feature intervals of all window areas in each historical image in the image data set as a third mode corresponding to the corresponding historical image; recording the number of the upper limit values of the texture feature intervals of all window areas in each historical image in the image data set as a fourth mode corresponding to the corresponding historical image;
Marking the median value of the texture feature interval of each window area in the gray level image as the first median value of the corresponding window area in the gray level image, and respectively calculating the standard deviation and the entropy value of the first median value of all window areas in the gray level image; marking the median value of the texture feature interval of each window area in each historical image in the image data set as the second median value of the corresponding window area in the corresponding historical image, and respectively calculating the standard deviation and the entropy value of the second median value of all window areas in each historical image in the image data set;
and screening a target historical image from an image dataset according to the first mode, the second mode, the third mode, the fourth mode, the standard deviation and entropy of the first median and the standard deviation and entropy of the second median.
Preferably, filtering the target historical image from the image dataset according to the first mode, the second mode, the third mode, the fourth mode, the standard deviation and entropy of the first median, and the standard deviation and entropy of the second median, includes:
determining the difference between the first mode and the third mode corresponding to each history image as the first difference between the gray image and each history image; determining the difference between the second mode and a fourth mode corresponding to each history image as a second difference between the gray scale image and each history image;
Obtaining matching parameters of the gray scale image and each historical image according to the first difference, the second difference, the difference between the standard deviation of the first median value of all window areas in the gray scale image and the standard deviation of the second median value of all window areas in each historical image, and the difference between the entropy value of the first median value of all window areas in the gray scale image and the entropy value of the second median value of all window areas in each historical image;
and taking the history image corresponding to the smallest matching parameter as a target history image.
Preferably, based on a target K value corresponding to a target history image in K-means clustering and a distribution interval of texture feature values of each connected domain in the gray image, clustering pixel points in the gray image by using a K-means clustering algorithm to obtain a clustering result, including:
taking a target K value corresponding to a target historical image in K-means clustering as a K value of the gray level image in K-means clustering;
taking the median value of the distribution interval of the texture characteristic values of each connected domain in the gray image as the characteristic value of each connected domain, obtaining the category corresponding to each connected domain based on the characteristic value of each connected domain, and selecting one pixel point in the connected domain with the largest area in each category as an initial clustering center;
And based on the K value of the gray image when K-means clustering is carried out and the initial clustering center, obtaining a clustering result when K-means clustering is carried out on the gray image.
Preferably, the step of determining whether the cutting quality of the steel plate to be detected is qualified by combining the clustering result and the distribution interval of the texture feature value of each connected domain in the target historical image includes:
determining a texture feature interval corresponding to each cluster in the target historical image based on the distribution interval of the texture feature values of each connected domain in the target historical image; determining a texture feature interval corresponding to each cluster in the gray image based on the distribution interval of the texture feature values of each connected domain in the gray image;
marking a cluster where a defect area in the target historical image is located as a defect cluster in the target historical image;
for any cluster in the grayscale image: respectively judging whether the texture feature interval corresponding to the cluster is larger than a preset duty ratio threshold of the texture feature interval corresponding to any defect cluster in the target historical image, and if so, determining the cluster as the defect cluster in the gray level image;
judging whether the total area of all the defect cluster clusters in the gray level image is larger than an area threshold value, if so, judging that the cutting quality of the steel plate to be detected is unqualified; if the cutting quality of the steel plate to be detected is smaller than or equal to the cutting quality of the steel plate to be detected, the steel plate to be detected is judged to be qualified.
The invention has at least the following beneficial effects:
according to the invention, firstly, the position distribution of the pixel points in each connected domain in the gray level image of the section of the steel plate to be detected and the gray level value of the pixel points and the adjacent pixel points are combined to determine the distribution interval of the texture characteristic value of each connected domain; and then, according to the distribution interval of the texture characteristic value of each connected domain in the gray level image of the to-be-detected steel plate tangent plane, the gray level value of the pixel point in the gray level image of the to-be-detected steel plate tangent plane and the gray level value of the pixel point in each historical image in the image data set, the target historical image is screened out from the image data set, the similarity degree of the target historical image and the gray level image of the to-be-detected steel plate tangent plane is highest, so that the target K value corresponding to the target historical image in K-means clustering is more suitable for the K value of the to-be-detected steel plate tangent plane in K-means clustering.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for detecting cutting quality of a steel plate for high-end manufacturing according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purpose, the following detailed description is given to a steel plate cutting quality detection method for high-end manufacturing according to the present invention with reference to the accompanying drawings and the preferred embodiments.
Unless defined otherwise, 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 belongs.
The following specifically describes a specific scheme of the steel plate cutting quality detection method for high-end manufacturing provided by the invention with reference to the accompanying drawings.
An embodiment of a steel plate cutting quality detection method for high-end manufacturing comprises the following steps:
the specific scene aimed at by this embodiment is: when the laser cutting is carried out on the steel plate for high-end manufacturing, cracks and scratches with different depths can be generated on the section, so that the section is rough and uneven, and at the moment, image defect detection is needed on the section of the steel plate. The defect degree of the section is judged by the existing method mainly by adopting a mean value clustering algorithm, but the mean value clustering algorithm depends on the selection of the quantity K, is too sensitive to noise, cannot quantify the defect degree, and further causes lower accuracy of a steel plate quality detection result. According to the embodiment, a K-means clustering algorithm for improving initial clustering center point selection is provided on the basis of a K-means clustering algorithm, texture features of defects are determined through windows with proper sizes, defect degrees in different windows are quantized, proper K values are selected, and an initial clustering center is determined; and finally, evaluating the quality of the steel plate to be detected based on the clustering result.
The embodiment provides a method for detecting cutting quality of a steel plate for high-end manufacturing, as shown in fig. 1, the method for detecting cutting quality of the steel plate for high-end manufacturing of the embodiment comprises the following steps:
Step S1, acquiring a gray image of a section of the steel plate to be detected.
According to the embodiment, vision detection is mainly performed on cracks of the section during laser cutting, so that a high-resolution camera is arranged on one side of the section, a lens of the camera is enabled to be right against the section of the steel plate to be detected, the camera is used for collecting RGB images of the section of the steel plate to be detected, gray level images are obtained by carrying out gray level processing on the collected RGB images of the section of the steel plate to be detected, gaussian filtering processing is carried out on the obtained gray level images, noise in the images is filtered, a section gray level image with good effect is obtained, and the filtered images are recorded as gray level images of the section of the steel plate to be detected. Wherein the RGB image is a color image.
Thus, a gray image of the section of the steel plate to be detected is obtained and is used for evaluating the cutting quality of the steel plate to be detected subsequently.
S2, sliding a sliding window on the gray level image to obtain window areas, and performing edge detection on the window areas to obtain communication areas in the window areas; obtaining the corresponding stability of each connected domain according to the gray level difference of each pixel point in the connected domain and the adjacent pixel points; and determining the distribution interval of the texture characteristic values of each connected domain based on the stability degree.
After the gray level image of the section of the steel plate to be detected is obtained, a sliding window with proper size is required to be selected, so that the number of included connected domains is moderate, the excessive number of various small through domains in the window is avoided, and finally, the representativeness of texture characteristics is lost.
Specifically, firstly, a gray level image of a section of a steel plate to be detected is processed by adopting an Ojin threshold method, cracks and grooves in the image can become a plurality of connected domains with gray level values of 0, morphological operation is carried out on the obtained image, small noise points and broken line segments are removed, a smooth connected domain is obtained, the connected domain obtained at the moment is marked as a target subarea, namely, the gray level image of the section of the steel plate to be detected is subjected to threshold segmentation, and a plurality of target subareas are obtained. The oxford thresholding method is a self-adaptive binarization method, which is a prior art and will not be described in detail here. Respectively obtaining the minimum circumscribed rectangle of each target subarea; calculating the average width of all the minimum circumscribed rectangles; and setting the length and the width of the sliding window to be preset multiples of the average width. The preset multiple in this embodiment is 1.8, and in a specific application, the practitioner can set according to the specific situation.
After the size of the sliding window is obtained, the sliding window is utilized to slide from left to right and from top to bottom on the gray level image of the section of the steel plate to be detected, each sliding corresponds to one window area, the sliding step length of the sliding window is equal to the width of the sliding window and is 1.8 times of the average width of all the minimum circumscribed rectangles, and the method can be used for obtaining a plurality of window areas in the gray level image of the section of the steel plate to be detected.
And (3) carrying out edge detection on each window area by adopting a Canny operator, and extracting a connected domain in each window area. Next, in this embodiment, the analysis is performed on the single connected domain first, and if the gray scales of the pixels of the single connected domain are more consistent, that is, do not change drastically, it is indicated that the texture stability in the connected domain is higher. Therefore, in this embodiment, firstly, gray level difference values are obtained according to gray level differences between each pixel point in each connected domain and adjacent pixel points, wherein the gray level difference values include a horizontal difference value and a vertical difference value, and the gray level difference values are obtained for evaluating the horizontal gray level change condition and the vertical gray level change condition of the pixel points; after the gray level change condition is obtained, the horizontal gray level change and the longitudinal gray level change of the pixel point can be synthesized, and the texture change degree of the connected domain can be analyzed; the texture change degree is weighted to characterize the texture characteristics of the connected domain, the embodiment combines the transverse differential value to determine the row gray change parameter and the longitudinal differential value to determine the column gray change parameter, the row gray change parameter reflects the intensity of the gray change of the pixel points in the transverse direction in the connected domain, and the larger the row gray change parameter is, the larger the gray change amplitude of each row of pixel points in the connected domain is, which means that deeper and shallower places possibly exist in the connected domain at the same time, namely, the more inconsistent the texture change in the connected domain is; the column gray scale change parameter reflects the intensity of the gray scale change of the pixel points in the longitudinal direction in the connected domain, and the larger the column gray scale change parameter is, the larger the gray scale change amplitude of each column of pixel points in the connected domain is, so that deeper and shallower places possibly exist in the connected domain at the same time, namely, the texture change in the connected domain is more inconsistent. Therefore, the embodiment determines the stability degree corresponding to each connected domain according to the row gray scale variation parameter and the column gray scale variation parameter.
Specifically, for any connected domain:
for any row in the connected domain: determining the gray level difference of two adjacent pixel points of each pixel point of the row as a transverse difference value of the corresponding pixel point; for example: for the 5 th pixel point in the row, determining the absolute value of the difference value of the gray values of the 4 th pixel point and the 6 th pixel point in the row as the transverse difference value of the 5 th pixel point in the row. It should be noted that: since there is no pixel point on the left side of the 1 st pixel point of the row and there is no pixel point on the right side of the last 1 pixel point of the row, the lateral difference value between the 1 st pixel point and the last 1 pixel point of the row is not calculated. Respectively calculating square values of all the transverse differential values of the row, and marking the sum of the square values of all the transverse differential values of the row as a first index corresponding to the row; and determining the ratio of the first index to the average width of all the minimum circumscribed rectangles as a line gray scale variation parameter of the line. Calculating the variance of the line gray scale variation parameters of all the lines according to the line gray scale variation parameters of each line in the connected domain; and determining the variance of the line gray scale variation parameters as the lateral stability of the connected domain.
For any column in the connected domain: the gray level difference of two adjacent pixels of each pixel of the row is determined as a longitudinal difference value of the corresponding pixel; for example: for the 3 rd pixel point in the column, determining the absolute value of the difference value of the gray values of the 2 nd pixel point and the 4 th pixel point in the column as the longitudinal difference value of the 3 rd pixel point in the column. It should be noted that: since there is no pixel above the 1 st pixel in the column and there is no pixel below the last 1 pixel in the row, the vertical difference value between the 1 st pixel and the last 1 pixel in the column is not calculated. Respectively calculating the square value of each longitudinal differential value of the row, and marking the sum of the square values of all the longitudinal differential values of the row as a second index corresponding to the column; and determining the ratio of the second index to the average width of all the minimum circumscribed rectangles as a column gray scale variation parameter of the column. Calculating the variance of the column gray scale variation parameters of all columns according to the column gray scale variation parameters of each column in the connected domain; and determining the variance of the column gray scale variation parameters as the longitudinal stability of the connected domain.
The specific calculation formulas of the row gray scale variation parameter of the ith row and the column gray scale variation parameter of the jth column in the connected domain are respectively as follows:
Wherein,for the line gray scale variation parameter of the i-th line in the connected domain,>for the column gray scale variation parameter of the j-th column in the connected domain, d is the average width of all the minimum bounding rectangles, M is the number of the lateral differential values of the pixel points in the i-th row in the connected domain, N is the number of the longitudinal differential values of the pixel points in the j-th column in the connected domain, and->For the image in the ith row in the connected domainMth lateral difference value of the pixel, < ->The n-th vertical differential value of the pixel point in the j-th column in the connected domain.
A first index corresponding to the ith row; />And the second index corresponding to the j-th column. If the crack defect in the connected domain is obvious, the gray level difference value of the pixel points in the connected domain is larger, namely the transverse difference value and the longitudinal difference value are larger; defects are represented on an image, and besides the magnitude of gray scale difference values and the ratio of the gray scale difference values to the ratio of the sum of squares of difference values of pixel points in a connected domain in the horizontal axis direction or in the longitudinal direction to the average width of all minimum circumscribed rectangles are calculated, and the larger the ratio is, the more obvious the defects in the connected domain are indicated.
After obtaining the row gray scale variation parameters of each row and the column gray scale variation parameters of each column in the connected domain, the embodiment determines the variance of the row gray scale variation parameters of all rows in the connected domain as the lateral stability of the connected domain, and determines the variance of the row gray scale variation parameters of all columns in the connected domain as the longitudinal stability of the connected domain. By adopting the method, the lateral stability and the longitudinal stability of the connected domain are obtained, and then the embodiment combines the lateral stability and the longitudinal stability of the connected domain to determine the overall stability of the connected domain, namely, calculate the corresponding stability of the connected domain, and because the direction of the crack or the cutting mark generated when the laser cuts the steel plate is fixed, namely, most of the scratch or the cutting mark extends towards the same direction, the direction should be given a larger weight when calculating the corresponding stability of the connected domain.
Specifically, performing linear detection on the connected domain to obtain each linear segment in the connected domain; in this embodiment, a hough line detection method is used to perform line detection on the connected domain, where the method is in the prior art and will not be described in detail here. Respectively counting the number of straight line segments in the transverse direction and the number of straight lines in the longitudinal direction in the communication domain, determining the direction with the maximum number of straight line segments in the two directions as a target direction, and taking the transverse direction as the target direction if the number of straight line segments in the transverse direction in the communication domain is larger than the number of straight lines in the longitudinal direction; if the number of straight line segments in the lateral direction in the connected domain is less than or equal to the number of straight lines in the longitudinal direction, the weight of the stability in the target direction should be large with the longitudinal direction as the target direction. In this embodiment, the stability degree corresponding to the connected domain is determined by combining the lateral stability of the connected domain, the longitudinal stability of the connected domain, the target direction and the preset weight, and the specific calculation formula of the stability degree corresponding to the connected domain is as follows:
wherein Q is the corresponding stability of the connected domain,for the lateral stability of the connected domain +. >For the longitudinal stability of the connected domain +.>For the first weight, ++>Is a second weight.
Since the weight of the stability in the target direction should be large, the present embodiment sets when the target direction is lateral0.7%>The total number of the components is 0.3,that is, the weight corresponding to the lateral stability is greater than the weight corresponding to the longitudinal stability; when the target direction is longitudinal, the present embodiment sets +.>0.3%>The weight corresponding to the longitudinal stability is 0.7, namely, the weight corresponding to the transverse stability is larger than the weight corresponding to the transverse stability. In a specific application, the practitioner may set up according to the specific circumstances.
The distribution interval of the texture characteristic values needs to represent the texture distribution condition of the whole connected domain, the stability degree cannot be weighted directly by using the upper and lower gray limits, and the upper and lower limits of the distribution interval of the texture characteristic values of the connected domain can be obtained by carrying out convergence and re-weighting on the stability degree with the average value, so that the distribution interval of the texture characteristic values of the connected domain is determined. After the stability corresponding to the connected domain is obtained, the embodiment combines the stability corresponding to the connected domain and the gray distribution of the pixel points in the connected domain to determine the distribution interval of the texture characteristic value of the connected domain.
Specifically, respectively acquiring an average gray value, a maximum gray value and a minimum gray value of all pixel points in each target direction in the connected domain; obtaining a lower limit value and an upper limit value of a distribution interval of the texture characteristic value of the connected domain according to the average gray value, the maximum gray value, the minimum gray value and the corresponding stability degree of the connected domain; and obtaining a distribution interval of the connected domain texture characteristic value based on the lower limit value and the upper limit value. It should be noted that: if the target direction is transverse, respectively acquiring an average gray value, a maximum gray value and a minimum gray value of all pixel points of each row in the connected domain, wherein each row in the connected domain corresponds to one average gray value, one maximum gray value and one minimum gray value; if the target direction is longitudinal, respectively acquiring an average gray value, a maximum gray value and a minimum gray value of all pixel points in each column in the connected domain, wherein each column in the connected domain corresponds to one average gray value, one maximum gray value and one minimum gray value. The specific expressions of the lower limit value and the upper limit value of the distribution interval of the connected domain texture characteristic value are respectively as follows:
wherein,a lower limit value of a distribution interval of the connected domain texture characteristic value, < - >K is the number of target directions in the connected domain, which is the upper limit value of the distribution interval of the texture characteristic value of the connected domain>For the average gray value of all pixels in the kth target direction in the connected domain, +.>For the minimum gray value of all pixels in the kth target direction in the connected domain, +.>And C is the corresponding stability of the connected domain, and is the maximum gray value of all pixel points in the kth target direction in the connected domain.
Is the average of both the average gray value and the minimum gray value of all pixel points in the kth target direction,dividing 255 by normalization means that the smaller the normalized value, the closer the distribution interval of texture feature values is to the minimum gray value, and the utilization +.>Correcting the corresponding stability of the connected domain to change the stability with smaller value into a value with moderate size and reflecting gray distribution; the lower limit value of the window texture characteristic value distribution interval where the connected domain with low gray level is located is smaller, the window is closer to the abnormal region of the steel plate, and on the contrary, the upper limit value of the window texture characteristic value distribution interval where the connected domain with high gray level is located is larger, and the window is closer to the normal region of the steel plate.
At the lower limit value of the distribution interval for obtaining the texture characteristic value of the connected domainAnd upper limit value->Afterwards, will [As a distribution interval of the connected domain texture feature value.
By adopting the method, the distribution interval of the texture characteristic value of each connected domain in the gray level image of the section of the steel plate to be detected can be obtained.
And step S3, screening the target historical images from the image data set according to the distribution interval of the texture characteristic values of each connected domain in the gray image, the gray values of the pixel points in the gray image and the gray values of the pixel points in each historical image in the image data set.
Next, in this embodiment, an image with the highest similarity to the gray image of the section of the steel plate to be detected is obtained from the image data set, and the K value of the section of the steel plate to be detected when the gray images of the section of the steel plate to be detected are clustered is determined based on the K value corresponding to the image in the image data set when the K-means is clustered. The image data set is composed of images of a plurality of steel plate sections, the images are all images collected in the history detection process, the images comprise various defect images and normal images without defects, the images are gray images, each image in the image data set is recorded as a history image, and the size of each history image is the same as that of the gray image of the section of the steel plate to be detected.
For any window region in the grayscale image: determining the minimum value in the lower limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic minimum value of the window area; determining the maximum value in the upper limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic maximum value of the window area; obtaining a texture feature interval of the window area based on the texture feature minimum value and the texture feature maximum value; that is, the lower limit value of the texture feature interval of the window area is the minimum texture feature value of the window area, and the upper limit value of the texture feature interval of the window area is the maximum texture feature value of the window area. The number of the lower limit values of the texture feature intervals of all window areas in the gray level image is recorded as a first mode corresponding to the gray level image; recording the numbers of the upper limit values of the texture feature intervals of all window areas in the gray level image as second numbers corresponding to the gray level image; recording the number of the lower limit values of the texture feature intervals of all window areas in each historical image in the image data set as a third mode corresponding to the corresponding historical image; recording the number of the upper limit values of the texture feature intervals of all window areas in each historical image in the image data set as a fourth mode corresponding to the corresponding historical image; marking the median value of the texture feature interval of each window area in the gray level image as the first median value of the corresponding window area in the gray level image, and respectively calculating the standard deviation and the entropy value of the first median value of all window areas in the gray level image; the calculation methods of the standard deviation and the entropy value are all the prior art, and are not repeated here; and marking the median value of the texture feature interval of each window area in each historical image in the image data set as the second median value of the corresponding window area in the corresponding historical image, and respectively calculating the standard deviation and the entropy value of the second median value of all window areas in each historical image in the image data set. Determining the difference between the first mode and the third mode corresponding to each history image as the first difference between the gray image and each history image; determining the difference between the second mode and a fourth mode corresponding to each history image as a second difference between the gray scale image and each history image; and obtaining matching parameters of the gray image and each historical image according to the first difference, the second difference, the difference between the standard deviation of the first median value of all window areas in the gray image and the standard deviation of the second median value of all window areas in each historical image, and the difference between the entropy value of the first median value of all window areas in the gray image and the entropy value of the second median value of all window areas in each historical image.
For any historical image in the image data set, a specific calculation formula of matching parameters of the gray level image of the section of the steel plate to be detected and the historical image is as follows:
wherein Y is a matching parameter of a gray level image of a section of the steel plate to be detected and the historical image;the difference between the mode of the lower limit value of the texture feature intervals of all window areas in the gray level image of the section of the steel plate to be detected and the mode of the lower limit value of the texture feature intervals of all window areas in the history image, namely the difference between the first mode corresponding to the gray level image of the section of the steel plate to be detected and the third mode corresponding to the history image; />The difference between the mode of the upper limit value of the texture feature intervals of all window areas in the gray level image of the section of the steel plate to be detected and the mode of the upper limit value of the texture feature intervals of all window areas in the history image, namely the difference between the second mode corresponding to the gray level image of the section of the steel plate to be detected and the fourth mode corresponding to the history image; />In the gray image of the section of the steel plate to be detectedStandard deviation of the first median value of all window areas, < >>For the standard deviation of the second median value of all window areas in the history image +. >Entropy value of the first median value of all window areas in gray level images of the section of the steel plate to be detected, < ->And (3) taking the entropy value of the second median value of all window areas in the historical image, wherein e is a natural constant, and I is an absolute value sign.
The method for calculating the difference between the first mode corresponding to the gray level image of the section of the steel plate to be detected and the third mode corresponding to the historical image comprises the following steps: and calculating the absolute value of the difference between the first mode corresponding to the gray level image of the section of the steel plate to be detected and the third mode corresponding to the historical image, and taking the absolute value as the difference between the first mode corresponding to the gray level image of the section of the steel plate to be detected and the third mode corresponding to the historical image. The calculation method of the difference between the second mode corresponding to the gray level image of the section of the steel plate to be detected and the fourth mode corresponding to the historical image comprises the following steps: and calculating the absolute value of the difference between the second mode corresponding to the gray level image of the section of the steel plate to be detected and the fourth mode corresponding to the historical image, and taking the absolute value as the difference between the second mode corresponding to the gray level image of the section of the steel plate to be detected and the fourth mode corresponding to the historical image.
The difference condition between standard deviations can be reflected, the standard deviations of the two are reflected to be the difference degree of textures in corresponding images, and the smaller the difference between the standard deviations is, the more similar the gray level image of the section of the steel plate to be detected is to the texture complexity degree of the historical image is. / >Can reflect entropy valueThe smaller the difference between the entropy values, the more similar the texture complexity of the gray level image of the section of the steel plate to be detected and the history image is. />The larger, i.e. the closer to 1, & gt>The smaller the image is, the higher the similarity between the gray image of the section of the steel plate to be detected and the historical image is.
By adopting the method, the matching parameters of the gray level image of the section of the steel plate to be detected and each historical image in the image data set can be obtained, and the smaller the matching parameters are, the higher the similarity degree between the corresponding historical image and the gray level image of the section of the steel plate to be detected is, so that the embodiment takes the historical image corresponding to the smallest matching parameter as the target historical image.
So far, the target historical image which is most similar to the gray level image of the section of the steel plate to be detected is screened from the image data set.
Step S4, clustering pixel points in the gray level image by adopting a K-means clustering algorithm based on a target K value corresponding to a target historical image in K-means clustering and a distribution interval of texture characteristic values of each connected domain in the gray level image to obtain a clustering result; and judging whether the cutting quality of the steel plate to be detected is qualified or not by combining the clustering result and the distribution interval of the texture characteristic values of each connected domain in the target historical image.
In the embodiment, the target historical image is screened out in the steps, the similarity between the target historical image and the gray image of the section of the steel plate to be detected is highest, and then the target K value corresponding to the target historical image in K-means clustering is used as the K value of the gray image in K-means clustering.
The method for acquiring the corresponding target K value of the target historical image in K-means clustering comprises the following steps: and clustering the target historical images for a plurality of times by adopting a method of manually setting the K value, manually evaluating the clustering effect of each clustering, taking the K value corresponding to the best clustering result as the target K value corresponding to the target historical images in K-means clustering, and sequentially carrying out K-means clustering on pixel points in the target historical images for a plurality of times by increasing 1 each time when the initial value of K is 2 in the process of manually setting the K value.
After determining the K value of the gray level image K-means of the section of the steel plate to be detected during clustering, in order to further improve the accuracy of the clustering result, in this embodiment, the initial clustering center of the section of the steel plate to be detected during the clustering of the gray level image K-means is determined, specifically, the median value of the distribution interval of the texture characteristic value of each connected domain in the gray level image of the section of the steel plate to be detected is used as the characteristic value of each connected domain, and each connected domain has a characteristic value. By adopting the method provided by the embodiment, the distribution interval of the texture characteristic value of each connected domain in the target historical image can be obtained, and because a plurality of complete connected domains possibly exist in the clustering clusters obtained after the target historical image is clustered by using the corresponding target K value, the embodiment determines the distribution interval of the texture characteristic value of each clustering cluster based on the distribution intervals of the texture characteristic values of all complete connected domains in each clustering cluster in the target historical image. Specifically, for any cluster, determining the minimum value of the lower limit value of the distribution interval of all the complete connected domain texture characteristic values in the cluster as the lower limit value of the texture characteristic interval corresponding to the cluster, determining the maximum value of the upper limit value of the distribution interval of all the complete connected domain texture characteristic values in the cluster as the upper limit value of the texture characteristic interval corresponding to the cluster, and obtaining the texture characteristic interval corresponding to the cluster based on the lower limit value of the texture characteristic interval corresponding to the cluster and the upper limit value of the texture characteristic interval corresponding to the cluster. By adopting the method, the texture feature interval corresponding to each cluster in the target historical image can be obtained.
Because each connected domain in the gray level image of the section of the steel plate to be detected has a corresponding distribution interval of the texture feature value in the embodiment, it is then determined which distribution interval each connected domain belongs to in the gray level image of the section of the steel plate to be detected, and the category corresponding to each connected domain is determined. Specifically, for any connected domain, it is determined in which cluster in the target history image the feature value of the connected domain is located in the texture feature interval corresponding to, and the serial number of the texture feature interval in which the feature value of the connected domain is located is used as the category corresponding to the connected domain, for example: and if the characteristic value of the connected domain is positioned in the texture characteristic interval corresponding to the 3 rd cluster in the target historical image, the class corresponding to the connected domain is the 3 rd class. By adopting the method, the class corresponding to each connected domain in the gray level image of the section of the steel plate to be detected is judged, the class corresponding to each connected domain in the gray level image of the section of the steel plate to be detected is determined, and in the embodiment, one pixel point in the connected domain with the largest area in each class is selected as an initial clustering center, namely K initial clustering centers are screened in the gray level image of the section of the steel plate to be detected, and then K-means clustering is adopted to cluster the pixels in the gray level image of the section of the steel plate to be detected, so that a clustering result is obtained.
After the clustering result of the gray level image of the section of the steel plate to be detected is obtained, the embodiment judges the cutting quality of the steel plate to be detected based on the clustering result. The method for acquiring the texture feature interval corresponding to each cluster in the target historical image is analogous to the method for acquiring the texture feature interval corresponding to each cluster in the gray level image of the section of the steel plate to be detected, and the embodiment has already described the method for acquiring the texture feature interval corresponding to each cluster in the target historical image in detail, so that the method for acquiring the texture feature interval corresponding to each cluster in the gray level image of the section of the steel plate to be detected is not described. In this embodiment, the cluster where the defect area in the target history image is located is recorded as the defect cluster in the target history image, and it should be noted that the defect area in the target history image is determined by a person in advance, and the number of defect clusters in the target history image is not necessarily 1. For any cluster in gray level images of a section of the steel plate to be detected: respectively judging whether the texture feature interval corresponding to the cluster is greater than 65% of the texture feature interval corresponding to any defect cluster in the target historical image, if so, determining the cluster in the gray level image of the section of the steel plate to be detected as the defect cluster in the gray level image of the section of the steel plate to be detected; if less than or equal to the predetermined value, then to be inspected And determining the cluster in the gray level image of the section of the steel plate to be detected as a normal cluster in the gray level image of the section of the steel plate to be detected, namely, no defect exists in the cluster. 65% is a preset duty cycle threshold, which may be set by the practitioner according to the particular situation in a particular application. For example: the texture feature interval corresponding to the first defect cluster in the target historical image is [10, 50]The texture feature interval corresponding to the first defect cluster in the target historical image is [40, 80 ]]The texture feature interval corresponding to one cluster in the gray level image of the section of the steel plate to be detected is [15, 55 ]]For the first defective cluster in the target history image,,87.5%>65, the cluster in the gray-scale image of the section of the steel sheet to be detected belongs to the defective area. By adopting the method, all defect clusters in the gray level image of the section of the steel plate to be detected can be screened out.
Counting the total area of all defect clusters in the gray level image of the section of the steel plate to be detected, wherein the larger the total area of all defect clusters in the gray level image of the section of the steel plate to be detected is, the worse the cutting quality of the steel plate to be detected is, so that the embodiment judges whether the total area of all defect clusters in the gray level image of the section of the steel plate to be detected is larger than an area threshold value, and if so, judges that the cutting quality of the steel plate to be detected is unqualified; if the cutting quality of the steel plate to be detected is smaller than or equal to the cutting quality of the steel plate to be detected, the steel plate to be detected is judged to be qualified. In a specific application, the area threshold implementer sets based on the size of the image acquired.
The method provided by the embodiment completes the judgment of the cutting quality of the steel plate to be detected.
According to the embodiment, firstly, the position distribution of the pixel points in each connected domain in the gray level image of the section of the steel plate to be detected and the gray level value of the pixel points and the adjacent pixel points are combined to determine the distribution interval of the texture characteristic value of each connected domain, the texture characteristic in each connected domain is analyzed, and the accuracy of the obtained distribution interval of the texture characteristic value is higher; and then screening out a target historical image from the image data set according to the distribution interval of the texture characteristic value of each connected domain in the gray image of the to-be-detected steel plate section, the gray value of the pixel point in the gray image of the to-be-detected steel plate section and the gray value of the pixel point in each historical image in the image data set, wherein the similarity degree of the target historical image and the gray image of the to-be-detected steel plate section is highest, the corresponding target K value of the target historical image in K-means clustering is more suitable for being used as the K value of the gray image of the to-be-detected steel plate section in K-means clustering, so that the embodiment clusters the pixel point in the gray image of the to-be-detected steel plate section by adopting a K-means clustering algorithm based on the corresponding target K value of the target historical image in K-means clustering and the distribution interval of the texture characteristic value of each connected domain in the gray image of the to-be-detected steel plate section, thereby obtaining a clustering result.
It should be noted that: the foregoing description of the preferred embodiments of the present invention is not intended to be limiting, but rather, any modifications, equivalents, improvements, etc. that fall within the principles of the present invention are intended to be included within the scope of the present invention.
Claims (4)
1. The method for detecting the cutting quality of the steel plate for high-end manufacturing is characterized by comprising the following steps of:
acquiring a gray image of a section of a steel plate to be detected;
sliding on the gray level image by utilizing a sliding window to obtain each window area, and carrying out edge detection on each window area to obtain each connected domain in each window area; obtaining the corresponding stability of each connected domain according to the position distribution of each pixel point in each connected domain and the gray difference between each pixel point and the adjacent pixel point; determining a distribution interval of texture characteristic values of each connected domain based on the stability and the gray value of the pixel point;
screening a target historical image from an image dataset according to the distribution interval of the texture characteristic values of each connected domain in the gray image, the gray value of the pixel point in the gray image and the gray value of the pixel point in each historical image in the image dataset;
based on a target K value corresponding to a target historical image in K-means clustering and a distribution interval of texture characteristic values of each connected domain in the gray image, clustering pixel points in the gray image by adopting a K-means clustering algorithm to obtain a clustering result; judging whether the cutting quality of the steel plate to be detected is qualified or not by combining the clustering result and the distribution interval of the texture characteristic values of each connected domain in the target historical image;
The acquisition of the size of the sliding window comprises the following steps:
threshold segmentation is carried out on the gray level image to obtain each target subarea, and the minimum circumscribed rectangle of each target subarea is respectively obtained; calculating the average width of all the minimum circumscribed rectangles;
setting the length and the width of the sliding window to be preset multiples of the average width;
the obtaining the corresponding stability of each connected domain according to the position distribution of each pixel point in each connected domain and the gray level difference between each pixel point and the adjacent pixel point comprises the following steps:
for any communication domain:
for any row in the connected domain: determining the gray level difference of two adjacent pixel points of each pixel point of the row as a transverse difference value of the corresponding pixel point; respectively calculating square values of all the transverse differential values of the row, and marking the sum of the square values of all the transverse differential values of the row as a first index corresponding to the row; determining the ratio of the first index to the average width of all the minimum circumscribed rectangles as a line gray scale variation parameter of the line;
for any column in the connected domain: the gray level difference of two adjacent pixels of each pixel of the row is determined as a longitudinal difference value of the corresponding pixel; respectively calculating the square value of each longitudinal differential value of the row, and marking the sum of the square values of all the longitudinal differential values of the row as a second index corresponding to the column; determining the ratio of the second index to the average width of all the minimum circumscribed rectangles as a column gray scale variation parameter of the column;
Performing linear detection on the connected domain to obtain each linear segment in the connected domain;
obtaining the corresponding stability of the connected domain according to the direction of each straight line segment in the connected domain, the line gray scale variation parameters of each line and the column gray scale variation parameters of each column;
obtaining the corresponding stability of the connected domain according to the direction of each straight line segment in the connected domain, the line gray scale variation parameters of each line and the column gray scale variation parameters of each column, wherein the method comprises the following steps:
calculating the variance of the line gray scale variation parameters of all the lines according to the line gray scale variation parameters of each line in the connected domain; determining the variance of the line gray scale variation parameter as the transverse stability of the connected domain;
calculating the variance of the column gray scale variation parameters of all columns according to the column gray scale variation parameters of each column in the connected domain; determining the variance of the column gray scale variation parameters as the longitudinal stability of the connected domain;
respectively counting the number of straight line segments in the transverse direction and the number of straight lines in the longitudinal direction in the connected domain, and determining the direction with the maximum number of straight line segments in the two directions as a target direction;
based on the transverse stability, the longitudinal stability, the target direction and a preset weight, obtaining the corresponding stability of the connected domain;
Determining a distribution interval of texture characteristic values of each connected domain based on the stability degree and the gray value of the pixel point, wherein the distribution interval comprises the following steps:
for any communication domain:
respectively acquiring an average gray value, a maximum gray value and a minimum gray value of all the pixel points in each target direction in the connected domain according to the gray value of each pixel point in each target direction in the connected domain;
obtaining a lower limit value and an upper limit value of a distribution interval of the texture characteristic value of the connected domain according to the average gray value, the maximum gray value, the minimum gray value and the corresponding stability degree of the connected domain; obtaining a distribution interval of the texture characteristic value of the connected domain based on the lower limit value and the upper limit value;
screening a target historical image from an image dataset according to a distribution interval of texture characteristic values of each connected domain in the gray image, gray values of pixel points in the gray image and gray values of pixel points in each historical image in the image dataset, wherein the method comprises the following steps:
for any window region in the grayscale image: determining the minimum value in the lower limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic minimum value of the window area; determining the maximum value in the upper limit value of the distribution interval of all the connected domain texture characteristic values in the window area as the texture characteristic maximum value of the window area; obtaining a texture feature interval of the window area based on the texture feature minimum value and the texture feature maximum value;
The number of the lower limit values of the texture feature intervals of all window areas in the gray level image is recorded as a first mode corresponding to the gray level image; recording the numbers of the upper limit values of the texture feature intervals of all window areas in the gray level image as second numbers corresponding to the gray level image; recording the number of the lower limit values of the texture feature intervals of all window areas in each historical image in the image data set as a third mode corresponding to the corresponding historical image; recording the number of the upper limit values of the texture feature intervals of all window areas in each historical image in the image data set as a fourth mode corresponding to the corresponding historical image;
marking the median value of the texture feature interval of each window area in the gray level image as the first median value of the corresponding window area in the gray level image, and respectively calculating the standard deviation and the entropy value of the first median value of all window areas in the gray level image; marking the median value of the texture feature interval of each window area in each historical image in the image data set as the second median value of the corresponding window area in the corresponding historical image, and respectively calculating the standard deviation and the entropy value of the second median value of all window areas in each historical image in the image data set;
Screening a target historical image from an image dataset according to the first mode, the second mode, the third mode, the fourth mode, the standard deviation and entropy of the first median value and the standard deviation and entropy of the second median value;
the step of judging whether the cutting quality of the steel plate to be detected is qualified or not by combining the clustering result and the distribution interval of the texture characteristic values of each connected domain in the target historical image comprises the following steps:
determining a texture feature interval corresponding to each cluster in the target historical image based on the distribution interval of the texture feature values of each connected domain in the target historical image; determining a texture feature interval corresponding to each cluster in the gray image based on the distribution interval of the texture feature values of each connected domain in the gray image;
marking a cluster where a defect area in the target historical image is located as a defect cluster in the target historical image;
for any cluster in the grayscale image: respectively judging whether the texture feature interval corresponding to the cluster is larger than a preset duty ratio threshold of the texture feature interval corresponding to any defect cluster in the target historical image, and if so, determining the cluster as the defect cluster in the gray level image;
Judging whether the total area of all the defect cluster clusters in the gray level image is larger than an area threshold value, if so, judging that the cutting quality of the steel plate to be detected is unqualified; if the cutting quality of the steel plate to be detected is smaller than or equal to the cutting quality of the steel plate to be detected, the steel plate to be detected is judged to be qualified.
2. The method for detecting the cutting quality of a steel plate for high-end manufacturing according to claim 1, wherein the lower limit value and the upper limit value of the distribution interval of the connected domain texture feature value are calculated by using the following formulas:
,
wherein,a lower limit value of a distribution interval of the connected domain texture characteristic value, < ->K is the number of target directions in the connected domain, which is the upper limit value of the distribution interval of the texture characteristic value of the connected domain>For the average gray value of all pixels in the kth target direction in the connected domain, +.>For the minimum gray value of all pixels in the kth target direction in the connected domain, +.>And C is the corresponding stability of the connected domain, and is the maximum gray value of all pixel points in the kth target direction in the connected domain.
3. The method for detecting the cutting quality of a steel sheet for high-end production according to claim 1, wherein screening a target history image from an image dataset based on the first mode, the second mode, the third mode, the fourth mode, a standard deviation and an entropy value of the first median, and a standard deviation and an entropy value of the second median, comprises:
Determining the difference between the first mode and the third mode corresponding to each history image as the first difference between the gray image and each history image; determining the difference between the second mode and a fourth mode corresponding to each history image as a second difference between the gray scale image and each history image;
obtaining matching parameters of the gray scale image and each historical image according to the first difference, the second difference, the difference between the standard deviation of the first median value of all window areas in the gray scale image and the standard deviation of the second median value of all window areas in each historical image, and the difference between the entropy value of the first median value of all window areas in the gray scale image and the entropy value of the second median value of all window areas in each historical image;
and taking the history image corresponding to the smallest matching parameter as a target history image.
4. The method for detecting the cutting quality of the steel plate for high-end manufacturing according to claim 1, wherein clustering the pixel points in the gray level image by using a K-means clustering algorithm based on a target K value corresponding to a target history image in K-means clustering and a distribution interval of texture feature values of each connected domain in the gray level image to obtain a clustering result comprises:
Taking a target K value corresponding to a target historical image in K-means clustering as a K value of the gray level image in K-means clustering;
taking the median value of the distribution interval of the texture characteristic values of each connected domain in the gray image as the characteristic value of each connected domain, obtaining the category corresponding to each connected domain based on the characteristic value of each connected domain, and selecting one pixel point in the connected domain with the largest area in each category as an initial clustering center;
and based on the K value of the gray image when K-means clustering is carried out and the initial clustering center, obtaining a clustering result when K-means clustering is carried out on the gray image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310926389.1A CN116645367B (en) | 2023-07-27 | 2023-07-27 | Steel plate cutting quality detection method for high-end manufacturing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310926389.1A CN116645367B (en) | 2023-07-27 | 2023-07-27 | Steel plate cutting quality detection method for high-end manufacturing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116645367A CN116645367A (en) | 2023-08-25 |
CN116645367B true CN116645367B (en) | 2023-12-01 |
Family
ID=87643744
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310926389.1A Active CN116645367B (en) | 2023-07-27 | 2023-07-27 | Steel plate cutting quality detection method for high-end manufacturing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116645367B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117113156B (en) * | 2023-10-20 | 2024-01-09 | 浙江鸿昌铝业有限公司 | Saw cutting section quality analysis method for aluminum profile |
CN117171693B (en) * | 2023-10-30 | 2024-01-26 | 山东交通学院 | Cutting abnormality detection method in woodworking polishing process |
CN117237385B (en) * | 2023-11-16 | 2024-01-26 | 江苏龙达纺织科技有限公司 | Textile transfer printing pattern extraction method and system based on pattern cutting |
CN117237245B (en) * | 2023-11-16 | 2024-01-26 | 湖南云箭智能科技有限公司 | Industrial material quality monitoring method based on artificial intelligence and Internet of things |
CN117333489B (en) * | 2023-12-01 | 2024-02-02 | 苏州普洛泰科精密工业有限公司 | Film damage detection device and detection system |
CN117522875B (en) * | 2024-01-08 | 2024-04-05 | 深圳市新创源精密智造有限公司 | Visual detection method for production quality of semiconductor carrier tape based on image filtering |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007062563A1 (en) * | 2005-12-01 | 2007-06-07 | Bohai Shipbuilding Industry Co., Ltd. | On-line automatic inspection method for detecting surface flaws of steel during the pretreatment of the ship steel |
CN112636642A (en) * | 2020-12-17 | 2021-04-09 | 广东工业大学 | Method and device for evaluating performance state of numerical control cutting tool bit made of flexible material |
CN115018828A (en) * | 2022-08-03 | 2022-09-06 | 深圳市尹泰明电子有限公司 | Defect detection method for electronic component |
CN115311277A (en) * | 2022-10-11 | 2022-11-08 | 南通美乐柯材料科技有限公司 | Pit defect identification method for stainless steel product |
CN115311292A (en) * | 2022-10-12 | 2022-11-08 | 南通创铭伊诺机械有限公司 | Strip steel surface defect detection method and system based on image processing |
CN115861290A (en) * | 2022-12-30 | 2023-03-28 | 南京林业大学 | Method for detecting surface defects of skin-touch wooden door |
CN116228798A (en) * | 2023-05-10 | 2023-06-06 | 青岛星跃铁塔有限公司 | Intelligent iron tower cutting detection method based on machine vision |
-
2023
- 2023-07-27 CN CN202310926389.1A patent/CN116645367B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007062563A1 (en) * | 2005-12-01 | 2007-06-07 | Bohai Shipbuilding Industry Co., Ltd. | On-line automatic inspection method for detecting surface flaws of steel during the pretreatment of the ship steel |
CN112636642A (en) * | 2020-12-17 | 2021-04-09 | 广东工业大学 | Method and device for evaluating performance state of numerical control cutting tool bit made of flexible material |
CN115018828A (en) * | 2022-08-03 | 2022-09-06 | 深圳市尹泰明电子有限公司 | Defect detection method for electronic component |
CN115311277A (en) * | 2022-10-11 | 2022-11-08 | 南通美乐柯材料科技有限公司 | Pit defect identification method for stainless steel product |
CN115311292A (en) * | 2022-10-12 | 2022-11-08 | 南通创铭伊诺机械有限公司 | Strip steel surface defect detection method and system based on image processing |
CN115861290A (en) * | 2022-12-30 | 2023-03-28 | 南京林业大学 | Method for detecting surface defects of skin-touch wooden door |
CN116228798A (en) * | 2023-05-10 | 2023-06-06 | 青岛星跃铁塔有限公司 | Intelligent iron tower cutting detection method based on machine vision |
Non-Patent Citations (2)
Title |
---|
Real-time Detection of Steel Strip Surface Defects Based on Improved YOLO Detection Network;Jiangyun Li et al;《IFAC》;76-81 * |
板带钢表面缺陷目标检测与分类算法研究_田思洋2019年第07期;田思洋;《中国博士学位论文全文数据库(电子期刊)》;第2019卷(第07期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN116645367A (en) | 2023-08-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116645367B (en) | Steel plate cutting quality detection method for high-end manufacturing | |
CN115294113B (en) | Quality detection method for wood veneer | |
CN115311292B (en) | Strip steel surface defect detection method and system based on image processing | |
CN110097034B (en) | Intelligent face health degree identification and evaluation method | |
CN115082467B (en) | Building material welding surface defect detection method based on computer vision | |
CN104794491B (en) | Based on the fuzzy clustering Surface Defects in Steel Plate detection method presorted | |
CN103593670B (en) | A kind of copper plate/strip detection method of surface flaw based on online limit of sequence learning machine | |
CN107157447B (en) | Skin surface roughness detection method based on image RGB color space | |
CN112651968B (en) | Wood board deformation and pit detection method based on depth information | |
CN109215009B (en) | Continuous casting billet surface image defect detection method based on deep convolution neural network | |
CN110210448B (en) | Intelligent face skin aging degree identification and evaluation method | |
CN104537651B (en) | Proportion detecting method and system for cracks in road surface image | |
CN113554629A (en) | Strip steel red rust defect detection method based on artificial intelligence | |
CN109540925B (en) | Complex ceramic tile surface defect detection method based on difference method and local variance measurement operator | |
CN114998323B (en) | Deformed steel bar abnormity determination method based on attention mechanism | |
CN116990323B (en) | High-precision printing plate visual detection system | |
CN115861290B (en) | Skin-feel wood door surface defect detection method | |
CN111402236B (en) | Hot rolled strip steel surface defect grading method based on image gray value | |
CN116721391B (en) | Method for detecting separation effect of raw oil based on computer vision | |
CN116452598A (en) | Axle production quality rapid detection method and system based on computer vision | |
CN116703911B (en) | LED lamp production quality detecting system | |
CN114820625A (en) | Automobile top block defect detection method | |
CN115690387A (en) | Alloy surface detection system based on image recognition | |
CN114549441A (en) | Sucker defect detection method based on image processing | |
CN113870202A (en) | Far-end chip defect detection system based on deep learning technology |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |