CN115830022A - Filter screen defect detection method based on machine vision - Google Patents

Filter screen defect detection method based on machine vision Download PDF

Info

Publication number
CN115830022A
CN115830022A CN202310113988.1A CN202310113988A CN115830022A CN 115830022 A CN115830022 A CN 115830022A CN 202310113988 A CN202310113988 A CN 202310113988A CN 115830022 A CN115830022 A CN 115830022A
Authority
CN
China
Prior art keywords
domain
filter hole
filter
area
pixel point
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.)
Granted
Application number
CN202310113988.1A
Other languages
Chinese (zh)
Other versions
CN115830022B (en
Inventor
马广圣
马广含
宋词
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Liangshan Distillery Co ltd
Original Assignee
Shandong Liangshan Distillery Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Liangshan Distillery Co ltd filed Critical Shandong Liangshan Distillery Co ltd
Priority to CN202310113988.1A priority Critical patent/CN115830022B/en
Publication of CN115830022A publication Critical patent/CN115830022A/en
Application granted granted Critical
Publication of CN115830022B publication Critical patent/CN115830022B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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

  • Image Processing (AREA)

Abstract

The invention relates to the technical field of data processing, in particular to a method for detecting defects of a filter screen of a filter based on machine vision, which comprises the steps of obtaining a gray level image corresponding to a surface image of the filter screen, and performing clustering iteration for at least two times on pixel points in the gray level image; in each clustering iteration, acquiring corresponding filter hole communication domains under filter hole clusters, respectively acquiring adjacent filter hole communication domains of each filter hole communication domain, and calculating superposition influence factors corresponding to corresponding interval regions according to gray scale information of the interval regions between the filter hole communication domains and the corresponding adjacent filter hole communication domains; correcting the membership degree of each pixel point in the interval area according to the superposition influence factors to obtain a new membership degree so as to divide the gray level image into a background area and a filter hole area; and detecting the defects of the filter screen according to the pore size of each filter pore area. The invention improves the integrity and accuracy of the segmentation of the background area and the filter hole area, so that the defect detection result of the filter screen is more accurate.

Description

Filter screen defect detection method based on machine vision
Technical Field
The invention relates to the technical field of data processing, in particular to a method for detecting defects of a filter screen of a filter based on machine vision.
Background
In the production process of white spirit, impurities can be separated out from fusel in the white spirit, so that the filtration is needed in the final step of the production, and the produced white spirit meets the standard. The filter of the production line uses a polyethersulfone filter screen to filter the white spirit. The polyether sulfone filter screen has long service time and the smallest filter hole can be reached
Figure SMS_1
And in the using process, the aperture of the surface filter hole of the polyether sulfone filter screen needs to be detected by using a scanning electron microscope at intervals.
The detection of the aperture of the surface filter hole of the polyethersulfone filter screen requires that a surface image of the polyethersulfone filter screen is collected firstly, then the surface image is segmented to segment the filter hole area, and finally the segmented filter hole area is identified according to the aperture size of the filter hole. Because a scanning electron microscope can generate a lot of noises when imaging the polyethersulfone filter screen, and the detailed texture information of the filter holes needs to be extracted, the surface filter holes are obtained by segmenting the surface image of the polyethersulfone filter screen by using an FCM clustering method (FRFCM) based on morphological reconstruction and membership filtering.
However, the surface layer filtering holes of the polyethersulfone filter screen may have lower layer filtering holes, and in the process of detecting the aperture, if the surface layer filtering holes cannot be completely extracted, the misjudgment of the aperture size is caused, and the traditional FRFCM clustering method clusters the gray levels, so that the overlapping situation of the filtering holes cannot be accurately segmented, and the membership filtering in the traditional FRFCM clustering cannot help to correctly segment the filtering holes.
Disclosure of Invention
In order to solve the problem that the traditional FRFCM clustering method cannot accurately divide the filter hole area under the condition of overlapping the filter holes, the invention aims to provide a machine vision-based filter screen defect detection method for a filter, and the adopted technical scheme is as follows:
one embodiment of the invention provides a machine vision-based defect detection method for a filter screen of a filter,
collecting surface images of the filter screen to obtain corresponding gray level images, and performing clustering iteration on pixel points for at least two times according to the distance between the pixel points in the gray level images to obtain filter hole clusters under each clustering iteration;
in each clustering iteration, acquiring corresponding filter hole communication domains under filter hole clusters, respectively acquiring adjacent filter hole communication domains of each filter hole communication domain according to the distance between the centroids of the filter hole communication domains, and calculating the superposition influence factor corresponding to the corresponding interval region according to the gray information of the interval region between the filter hole communication domain and the corresponding adjacent filter hole communication domain; correcting the membership degree of each pixel point in the interval area according to the superposition influence factors to obtain a new membership degree, and dividing the gray level image into a background area and a filter hole area based on the new membership degree;
and detecting the defects of the filter screen according to the pore size of each filter pore area.
Further, the method for acquiring the superposition influence factor includes:
taking any filter hole communicating domain as a target communicating domain, taking any adjacent filter hole communicating domain of the target communicating domain as a first communicating domain, connecting the centroid of the target communicating domain with each edge point on the edge of the first communicating domain to obtain at least two straight lines, obtaining an included angle between any two straight lines, forming two straight lines corresponding to the maximum included angle, the centroid of the target communicating domain and the first communicating domain into a new communicating domain, and taking the region except the first communicating domain in the new communicating domain as an interval region between the target communicating domain and the first communicating domain;
calculating the Euclidean distance between the centroid of the target connected domain and the centroid of the first connected domain;
calculating an average gray value of the interval area as a first value, calculating an average gray value between the target connected domain and the first connected domain as a second value, acquiring a difference value between the first value and the second value, and using a value obtained by normalizing the difference value as a weight value of the interval area;
calculating information entropy according to the membership degree of each pixel point in the interval area;
and taking the reciprocal of the product of the Euclidean distance, the weight value and the information entropy as a superposition influence factor of an interval region between the target connected domain and the first connected domain.
Further, the method for acquiring the new membership degree includes:
respectively obtaining superposition influence factors of an interval area between a target connected domain and each adjacent filter hole connected domain, respectively taking each superposition influence factor corresponding to the target connected domain as an index of a natural constant, taking the obtained result as a first result corresponding to the superposition influence factor, and obtaining an addition result of all the first results;
taking a first result corresponding to an interval region between the target connected domain and the first connected domain as a numerator, and taking an addition result as a denominator to obtain a corresponding ratio; the addition result of the constant 1 and the ratio is used as a correction coefficient;
and regarding any pixel point in a spacing area between the target communication domain and the first communication domain, taking the product of the correction coefficient and the membership degree of the pixel point as the new membership degree of the corresponding pixel point.
Further, the method for segmenting the gray-scale image into the background area and the filter hole area based on the new membership degree comprises the following steps:
for any pixel point in the gray level image, when the pixel point belongs to pixel points in at least two interval areas, acquiring a new membership degree of the pixel point in each interval area, wherein the pixel point belongs to the same cluster, and taking the maximum new membership degree as a final new membership degree of the pixel point belonging to a corresponding cluster; and segmenting the gray level image into a background area and a filter hole area based on the final new membership degree.
Further, the method for respectively obtaining the adjacent filter pore communication domains of each filter pore communication domain according to the distance between the centroids of the filter pore communication domains comprises the following steps:
and for any filter hole communicating domain, calculating the centroid distance between the filter hole communicating domain and other filter hole communicating domains, and selecting the other filter hole communicating domain corresponding to the minimum centroid distance as the adjacent filter hole communicating domain of the filter hole communicating domain.
Further, the method for detecting the defects of the filter screen according to the pore size of each filter pore area comprises the following steps:
and obtaining the minimum circumscribed circle of each filter hole area, obtaining the diameter of the minimum circumscribed circle, taking the diameter as the aperture of the corresponding filter hole area, and determining that the filter screen has defects when the aperture is not in the set aperture range.
Further, the method for performing clustering iteration on pixel points at least twice according to the distance between the pixel points in the gray level image to obtain a filter hole cluster class under each clustering iteration includes:
performing clustering iteration on pixel points in the gray level image at least twice by using an objective function of the traditional FCM clustering to obtain a filter hole cluster class under each clustering iteration, wherein the objective function
Figure SMS_2
Comprises the following steps:
Figure SMS_3
wherein ,
Figure SMS_7
expressing the total number of pixel points in the gray level image;
Figure SMS_9
is an index of a cluster class, represents
Figure SMS_14
A cluster class;
Figure SMS_6
representing the clustering algorithm to divide the gray scale image into a number of clusters for clustering
Figure SMS_10
Class i, also
Figure SMS_13
The maximum value of the values;
Figure SMS_16
for the ith pixel point belonging to the ith
Figure SMS_4
Membership of individual clusters;
Figure SMS_8
is membership fuzzy weighted index;
Figure SMS_12
is shown as
Figure SMS_15
Cluster center points of individual clusters;
Figure SMS_5
is a norm;
Figure SMS_11
is the ith pixel point.
The invention has the following beneficial effects:
the method comprises the steps of obtaining a gray level image corresponding to a surface image of a filter screen, wherein the background and filter holes cannot be completely and accurately segmented under the condition that the filter holes are overlapped due to gray level clustering in the traditional FRFCM clustering, so that at least two times of clustering iteration is carried out on pixel points according to the distance between the pixel points in the gray level image to obtain filter hole clusters under each time of clustering iteration, the problem that the pixel points in the background area and the filter hole overlapping area cannot be accurately and completely segmented is solved, in each time of clustering iteration, the corresponding filter hole communicating areas under the filter hole clusters are obtained, the mistaken segmentation condition caused by filter hole overlapping possibly exists in the background area during each iteration process is considered, the adjacent filter hole communicating areas of each filter hole communicating area are respectively obtained according to the distance between the centroids of the filter hole communicating areas, and the overlapping influence factors corresponding to the corresponding interval areas are calculated according to the gray level information of the interval areas between the filter hole communicating areas and the corresponding adjacent filter hole communicating areas, so that the calculation of the overlapping influence factors is more in line with actual information; in the traditional FRFCM clustering, median filtering is used for correcting a membership matrix, because the median filtering does not contain any semantic information in an image, the median filtering cannot correctly correct the influence area of superposed filter holes, the membership of each pixel point in an interval area is corrected according to superposed influence factors to obtain new membership, and self-adaptive membership filtering of each pixel point is formed through the superposed influence factors, so that the membership matrix in each clustering iteration is corrected to remove the influence of superposed filter holes, the integrity and the accuracy of dividing a gray level image into a background area and a filter hole area based on the new membership are improved, and the result of detecting the defects of a filter screen according to the aperture size of each filter hole area is more accurate.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a schematic illustration of a single filter pore after grey scale morphological reconstruction;
FIG. 2 is a schematic representation of the results of segmenting the filter pores of FIG. 1 using FRFCM clustering;
FIG. 3 is a flow chart illustrating the steps of a method for detecting defects in a filter screen of a filter based on machine vision according to an embodiment of the present invention.
Detailed Description
In order to further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the method for detecting defects of a filter screen of a filter based on machine vision, the specific implementation, structure, features and effects thereof according to the present invention, with reference to the accompanying drawings and preferred embodiments, is provided. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more 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 specific scenes aimed by the invention are as follows: in the process of using a fuzzy C-means clustering method (FRFCM) based on morphological reconstruction and membership filtering to perform filter hole segmentation, the FRFCM consists of three parts: gray level morphological reconstruction + FCM of gray level + traditional membership filtering (median filtering or mean filtering, which plays a corrective role). Since FRFCM clusters the gray levels and corrects the membership matrix by basic filtering, semantic information in the image is not considered, and thus, under the condition of filtering hole superposition, a segmentation error occurs by using FRFCM clustering, and the filtering holes cannot be accurately segmented, as shown in fig. 1 and 2, fig. 1 is a single filtering hole after gray level morphological reconstruction, and fig. 2 is a result of segmenting the filtering hole of fig. 1 by using FRFCM clustering.
The specific scheme of the filter screen defect detection method based on machine vision provided by the invention is specifically described below with reference to the accompanying drawings.
Referring to FIG. 3, a flow chart of the steps of a method for detecting defects in a filter screen of a filter based on machine vision according to an embodiment of the present invention is shown, the method comprising:
and S001, acquiring a surface image of the filter screen to obtain a corresponding gray level image, and performing clustering iteration on pixel points at least twice according to the distance between the pixel points in the gray level image to obtain filter hole clusters under each clustering iteration.
In particular, the minimum filtration pore of the polyethersulfone filter screen can be reached
Figure SMS_17
Therefore, it is necessary to acquire an image of the surface thereof using an electron microscope. And (3) the polyethersulfone filter screen is imaged by a Scanning Electron Microscope (SEM) to obtain a surface image, and the surface image is subjected to morphological reconstruction by utilizing the traditional gray morphological opening operation reconstruction to obtain a gray image after gray morphological reconstruction.
It should be noted that noise in the image can be removed in the gray scale morphology opening operation reconstruction, and the edge of the filter hole is enhanced, and the conventional gray scale morphology opening operation reconstruction is a part of the prior art in the conventional FRFCM clustering, and is not described in detail in this scheme.
The conventional FRFCM clustering is to perform gray level clustering on an image, that is, pixels with the same gray value in the image can only be classified into the same cluster class. In order to solve the problem of incomplete segmentation caused by filtering hole superposition, pixels with the same gray value need to be divided into different clusters because of the positions of the pixels and the information of adjacent pixels, that is, gray level clusters in the traditional FRFCM cluster need to be converted into pixel clusters.
Wherein, the objective function of gray level clustering in the traditional FRFCM clustering
Figure SMS_18
Comprises the following steps:
Figure SMS_19
wherein ,
Figure SMS_25
for the index of the grey level in the image, the first in the image
Figure SMS_24
A gray level;
Figure SMS_37
representing a common in the image for the number of grey levels in the image
Figure SMS_22
A gray scale level of
Figure SMS_34
The maximum value of the value;
Figure SMS_23
for indexing of cluster classes, denote
Figure SMS_33
A cluster class;
Figure SMS_30
representing the clustering algorithm to divide the image into a number of clusters for clustering
Figure SMS_40
Class i and is
Figure SMS_20
The maximum value of the value;
Figure SMS_32
is as follows
Figure SMS_28
The number of pixels contained in each gray level;
Figure SMS_35
is as follows
Figure SMS_26
The gray scale belongs to
Figure SMS_39
Membership of individual clusters;
Figure SMS_29
is membership fuzzy weighted index;
Figure SMS_38
for images after morphological reconstruction
Figure SMS_27
A gray level;
Figure SMS_36
is shown as
Figure SMS_21
Cluster center points of individual clusters;
Figure SMS_31
is a norm.
It should be noted that, in the conventional FRFCM clustering, the target function is to cluster the gray levels of the reconstructed images through gray morphology, that is, to divide the gray levels of the reconstructed images into two categories, that is, the gray value difference distance between the gray level and the gray level in the center of the cluster is measured through the membership and the number of pixels included in the gray level, and the target function is to minimize the gray value difference distance.
For a polyether sulfone filter screen in a white spirit filter, the gray levels in surface layer filter holes are negatively affected by superposition, one type of filter holes are divided based on the clustering of the gray levels, some pixel points with the same gray value belong to a background region, and the pixel points belong to the background region and are divided into the background type, so that the clustering of the gray levels cannot achieve the expected effect, and the clustering of the gray levels needs to be converted into the clustering of the pixel points, specifically: objective function for gray level clustering in conventional FRFCM clustering
Figure SMS_41
Target function for replacing traditional FCM clustering
Figure SMS_42
The traditional FRFCM clustering membership degree calculation method is also based on the traditional FCM clustering objective function and adopts a Lagrange multiplier method to calculate the membership degree
Figure SMS_43
And cluster center point
Figure SMS_44
Respectively derived as an objective function
Figure SMS_45
And taking the membership degree of the local minimum value and the value requirement of the cluster center point.
Wherein the target function of the conventional FCM cluster
Figure SMS_46
Comprises the following steps:
Figure SMS_47
wherein ,
Figure SMS_49
representing the total number of pixel points in the image;
Figure SMS_53
is an index of a cluster class, represents
Figure SMS_57
A cluster class;
Figure SMS_50
representing the clustering algorithm to divide the image into a number of clusters for clustering
Figure SMS_55
Class i and is
Figure SMS_58
The maximum value of the value;
Figure SMS_60
for the ith pixel point belonging to the ith
Figure SMS_48
Membership of individual clusters;
Figure SMS_52
is membership fuzzy weighted index;
Figure SMS_56
is shown as
Figure SMS_59
Cluster center points of individual clusters;
Figure SMS_51
is a norm;
Figure SMS_54
is the ith pixel point.
It is to be noted thatObjective function for gray level clustering in unified FRFCM clustering
Figure SMS_61
Target function replaced by traditional FCM clustering
Figure SMS_62
Compared with the gray level clustering segmentation, the clustering segmentation of the pixel points can divide the pixel points with the same gray level value into different cluster types, so that the aperture segmentation of the polyether sulfone filter screen in white spirit filtration is more delicate, and the problem that the pixel points in the background area and the filter hole superposition area cannot be accurately and completely segmented when the gray levels of the pixel points are the same is solved.
Wherein the objective function of the traditional FCM clustering is utilized
Figure SMS_63
When clustering iteration is carried out, the calculation formula of the membership degree is as follows:
Figure SMS_64
wherein ,
Figure SMS_66
for the ith pixel point to belong to the th
Figure SMS_69
Membership of individual clusters;
Figure SMS_72
is membership fuzzy weighted index;
Figure SMS_67
is shown as
Figure SMS_70
Cluster center points of individual clusters;
Figure SMS_73
is a norm;
Figure SMS_74
is the ith pixel point；
Figure SMS_65
The number of cluster types for clustering;
Figure SMS_68
is as follows
Figure SMS_71
Cluster center points of individual cluster classes.
The method aims to divide the surface image of the polyether sulfone filter screen into a background area and a filter hole area through clustering, so that an objective function of the traditional FCM clustering is set
Figure SMS_75
The constraint of (2): number of clusters
Figure SMS_76
Setting as 2, namely dividing into filter hole cluster and background cluster, fuzzy weighting index of membership degree in clustering process
Figure SMS_77
The iteration stop condition is that the difference of the membership degree matrix between two adjacent iteration processes is less than a threshold value
Figure SMS_78
. So using the objective function of the traditional FCM clustering
Figure SMS_79
And performing clustering iteration for at least two times on the pixel points in the gray level image to obtain a filter hole cluster class under each clustering iteration.
Step S002, in each clustering iteration, acquiring corresponding filter hole communication domains under the filter hole clusters, respectively acquiring adjacent filter hole communication domains of each filter hole communication domain according to the distance between the centroids of the filter hole communication domains, and calculating the superposition influence factor corresponding to the corresponding interval domain according to the gray scale information of the interval domain between the filter hole communication domains and the corresponding adjacent filter hole communication domains; and correcting the membership degree of each pixel point in the interval area according to the superposition influence factors to obtain a new membership degree, and dividing the gray level image into a background area and a filter hole area based on the new membership degree.
Specifically, in the iterative process of clustering, each iteration has a membership matrix corresponding to all the pixel points, and cluster classification is performed through the maximum membership of each pixel point, namely, the deblurring process. After cluster classification is obtained, connected domain processing is carried out on the filter hole clusters, and then membership degree information and gray information of pixel points in interval domains of each connected domain and surrounding connected domains are analyzed, and the method specifically comprises the following steps: and acquiring each filter hole communication domain under the filter hole cluster, calculating the centroid distance between each filter hole communication domain and other filter hole communication domains for any filter hole communication domain, and selecting other filter hole communication domains corresponding to the minimum centroid distance as adjacent filter hole communication domains of the filter hole communication domains.
As an example, the following
Figure SMS_81
Each filter hole is communicated with the area
Figure SMS_84
For example, according to the filter hole communication domain
Figure SMS_85
The center of mass of (1) is obtained and the connected domain of the filter holes is obtained
Figure SMS_82
The centroid of the other filter hole communicating domain closest to the centroid takes the other filter hole communicating domain corresponding to the centroid as the filter hole communicating domain
Figure SMS_83
To obtain a set of connected areas of adjacent filter holes
Figure SMS_86
Inscribing the first in the collection of connected domains of adjacent filter pores
Figure SMS_87
The adjacent filter holes are communicated with the area
Figure SMS_80
。
Through the interval region between the filter hole communicating region and the adjacent filter hole communicating region, namely the interval background region, whether the interval region is the wrong division condition caused by the superposition of the filter holes is analyzed, and the method specifically comprises the following steps:
taking any one filter hole communicated domain as a target communicated domain, taking any one adjacent filter hole communicated domain of the target communicated domain as a first communicated domain, connecting the centroid of the target communicated domain with each edge point on the edge of the first communicated domain to obtain at least two straight lines, obtaining an included angle between any two straight lines, forming a new communicated domain by the two straight lines corresponding to the maximum included angle, the centroid of the target communicated domain and the first communicated domain, and taking the region except the first communicated domain in the new communicated domain as an interval region between the target communicated domain and the first communicated domain; calculating the Euclidean distance between the centroid of the target connected domain and the centroid of the first connected domain; calculating an average gray value of the interval area as a first value, calculating an average gray value between the target connected domain and the first connected domain as a second value, acquiring a difference value between the first value and the second value, and using a value obtained by normalizing the difference value as a weight value of the interval area; calculating information entropy according to the membership degree of each pixel point in the interval area; and taking the reciprocal of the product of the Euclidean distance, the weight value and the information entropy as a superposition influence factor of an interval region between the target connected domain and the first connected domain.
As an example, the domains are connected by filter holes
Figure SMS_89
And its adjacent filter hole communicating region
Figure SMS_93
For example, the filter holes are connected with the domain
Figure SMS_98
Centroid and adjacent filter pore connected domain
Figure SMS_91
Each edge point on the edge of (1) is connected to obtain at least two straight lines, and the included angle between any two straight lines is calculated to obtainTaking two straight lines with the largest included angle as target straight lines, and connecting the target straight lines with filter holes
Figure SMS_95
Centroid and adjacent filter pore connected domain
Figure SMS_99
Forming a new connected domain, and removing the connected domain of the adjacent filter pores in the new connected domain
Figure SMS_102
The rest area of the filter is used as a filter hole communicating area
Figure SMS_88
And adjacent filter hole communicating region
Figure SMS_92
In the interval region between
Figure SMS_96
Spacer region
Figure SMS_100
Namely a background area corresponding to the non-filter hole communication domain; according to the spacing region
Figure SMS_90
Calculating interval area by gray value and membership of each pixel point
Figure SMS_94
By the superposition factor of
Figure SMS_97
Then the influence factors are superposed
Figure SMS_101
The calculation formula of (2) is as follows:
Figure SMS_103
wherein ,
Figure SMS_106
for connecting the filter pores
Figure SMS_111
Centroid and adjacent filter pore connected domain of
Figure SMS_114
The euclidean distance between centroids of (a);
Figure SMS_105
is a normalization function;
Figure SMS_108
is a spacer region
Figure SMS_112
Average gray value of (2) and, filter pore connected domain
Figure SMS_115
And adjacent filter hole communicating region
Figure SMS_104
The difference between the average gray values;
Figure SMS_109
is a spacer region
Figure SMS_113
To middle
Figure SMS_116
The membership degree difference of each pixel point to two clusters needs to be explained, each pixel point has one membership degree to each cluster, and the background clusters and the filter hole clusters are adopted in the scheme, so that two membership degrees exist in one pixel point, and the membership degree difference of the corresponding pixel point is obtained according to the membership degree of each pixel point belonging to each cluster;
Figure SMS_107
according to the interval region
Figure SMS_110
Membership degree difference meter for each pixel pointAnd calculating the information entropy of the membership degree difference.
It should be noted that, for adjacent filter hole connected domains, if the difference between the overall gray-scale value of the interval region and the average gray-scale value of the filter hole connected domains is
Figure SMS_117
The smaller the difference between the membership degrees of the pixel points in the interval area to the two clusters is, the more stable the difference is, namely
Figure SMS_118
The larger the size of the filter hole, the more likely the interval area is influenced by the superposition of the filter holes, and the superposition influence factor of the interval area is
Figure SMS_119
The higher the bit rate; at this time, it is also considered that if an isolated filter hole connected domain occurs, the overlapping influence factor of the spacing region between the isolated filter hole connected domain and the nearest single adjacent filter hole connected domain becomes high, so the euclidean distance between the filter hole connected domains needs to be measured as a part of the overlapping influence factor, that is, the euclidean distance
Figure SMS_120
The larger the overlap factor of the corresponding spacing region
Figure SMS_121
The larger.
Using superimposed influencing factors
Figure SMS_122
The calculation formula of (2) obtains the superposition influence factor of the interval region between each filter hole connected domain and each adjacent filter hole connected domain, and the calculation of the superposition influence factor is carried out through the Euclidean distance between the filter hole connected domains, the gray scale distance and the membership degree information, so that the calculation of the superposition influence factor is more in line with the actual information.
And for each background pixel point in the clustering iteration process, acquiring the superposition influence factor of the background pixel point according to the interval region where the background pixel point is located, and correcting the membership matrix by using median filtering in the traditional FRFCM clustering. Because the median filtering does not contain any semantic information in the image, the median filtering cannot correctly correct the influence area of the superposed filter holes, and therefore self-adaptive membership filtering of each pixel point needs to be formed by superposing influence factors, so that correction for removing the influence of superposed filter holes is carried out on a membership matrix in each clustering iteration, which specifically comprises the following steps: taking any one filter hole communicating domain as a target communicating domain, taking any one adjacent filter hole communicating domain of the target communicating domain as a first communicating domain, respectively obtaining superposition influence factors of an interval region between the target communicating domain and each adjacent filter hole communicating domain, respectively taking each superposition influence factor corresponding to the target communicating domain as an index of a natural constant, taking the obtained result as a first result of the corresponding superposition influence factor, and obtaining the addition result of all the first results; taking a first result corresponding to an interval region between the target connected domain and the first connected domain as a numerator, and taking an addition result as a denominator to obtain a corresponding ratio; the addition result of the constant 1 and the ratio is used as a correction coefficient; and regarding any pixel point in the interval area between the target connected domain and the first connected domain, taking the product of the correction coefficient and the membership degree of the pixel point as the new membership degree of the corresponding pixel point.
As an example, the domains are connected by filter holes
Figure SMS_123
Communicating with its adjacent filter pores
Figure SMS_124
In the interval region between
Figure SMS_125
Taking the ith pixel point as an example, the calculation formula of the new membership degree of the pixel point is as follows:
Figure SMS_126
wherein ,
Figure SMS_129
is a roomSeparation zone
Figure SMS_132
The ith pixel point in (1) belongs to the
Figure SMS_136
New membership of individual clusters;
Figure SMS_130
is a spacer region
Figure SMS_133
The ith pixel point in (1) belongs to the
Figure SMS_138
The membership degree of each cluster class, namely the membership degree before correction;
Figure SMS_140
is a natural constant;
Figure SMS_127
is a spacer region
Figure SMS_131
The superposition influence factor of (2);
Figure SMS_135
for connecting the filter pores
Figure SMS_139
And the adjacent filter hole communicating area
Figure SMS_128
In the interval region between
Figure SMS_134
The superposition influence factor of (2);
Figure SMS_137
for connecting the filter pores
Figure SMS_141
The number of adjacent filter openings communicating with the domain.
It is noted thatIn the filter hole cluster class I
Figure SMS_142
Each filter hole is communicated with the area
Figure SMS_143
Which is provided with
Figure SMS_144
Adjacent filter hole connected region
Figure SMS_145
There will be a superimposed influence factor for each spacer region adjacent to the connected region of the filter hole
Figure SMS_146
The superimposed influence factors of the interval areas are normalized, namely, how to distribute the interval areas in the group to increase the membership degree of the filter hole cluster. If the superposition influence factor of one interval area is maximum compared with a group of interval areas, the membership degree of the pixel points in the interval area to the filter hole clusters is correspondingly increased, namely
Figure SMS_147
The larger the value of (A), the correction coefficient
Figure SMS_148
The larger the value is, the larger the corresponding new membership degree is, so that the pixel point is more likely to be divided into filter hole clusters, and the problem of wrong segmentation caused by filter hole superposition is avoided.
The advantage of utilizing the superposition influence factor of the interval region between the adjacent filter hole connected domains to correct the membership degree of each pixel point in the interval region is as follows: according to the scheme, semantic information of cluster segmentation in an image in the current clustering iteration process is integrated, the semantic information comprises Euclidean distance between filter hole communicating domains, gray scale distance between an interval region and the filter hole communicating domains and membership difference information of the interval region, different correction can be performed on membership of each pixel point in the interval region through superposition influence factors obtained through the semantic information, for traditional membership filtering in the traditional FRFCM clustering, such as median filtering, mean filtering and the like, the essence is that the membership of a central pixel point is corrected through the membership of 8 neighborhood pixel points of the pixel point, the correction is numerical correction and uniform mean solving, the correction method is effective under the condition that Gaussian noise exists in the image, but when the correction of the membership of the pixel point through pixel point distribution characteristics in the image is required, the correction method in the scheme is required, and is a new membership acquisition method for correcting a membership matrix, so that incorrect segmentation caused by filter hole superposition in the clustering process can be corrected.
And acquiring the new membership degree of each pixel point in the interval area between each filter hole communicating domain and each adjacent filter hole communicating domain by using a calculation formula of the new membership degree. Considering the condition that the interval regions have intersection, at least two pixel points in the intersection belong to the second
Figure SMS_149
The new membership of each cluster, therefore, the final new membership of each pixel point in the intersection needs to be confirmed: for any pixel point in the gray level image, when the pixel point belongs to the pixel points in at least two interval areas, acquiring the new membership degree of the pixel point in each interval area belonging to the same cluster, and taking the maximum new membership degree as the final new membership degree of the pixel point belonging to the corresponding cluster.
As an example, when the filter pores communicate with the domain
Figure SMS_153
For connecting the filter pores
Figure SMS_151
Adjacent to the filter hole communication area, and the filter hole communication area
Figure SMS_160
Also a filter pore communicating region
Figure SMS_154
Is adjacent toWhen the filter holes are close to the connected domain, the filter holes are connected with the domain
Figure SMS_161
Communicating with its adjacent filter pores
Figure SMS_156
A spacer region therebetween, and a filter pore communicating region
Figure SMS_164
Communicating with its adjacent filter pores
Figure SMS_155
The interval areas between the two adjacent pixels have intersection, and any one pixel point in the intersection has two pixels belonging to the second pixel
Figure SMS_162
Selecting the maximum new membership degree as the second membership degree of the corresponding pixel point
Figure SMS_150
Final new membership of each cluster; when the filter pores are communicated
Figure SMS_159
Communicating with its adjacent filter pores
Figure SMS_158
A spacer region therebetween, and a filter pore communicating region
Figure SMS_163
Communicating with its adjacent filter pores
Figure SMS_157
When the interval areas between the two areas do not have intersection, each pixel point in the interval areas belongs to the first
Figure SMS_165
The new membership degree of each cluster as the corresponding pixel point belonging to the first
Figure SMS_152
Final new membership of individual cluster classes.
Since the interval region is the region other than the filter hole connected region, that is, the background region, it is obtained that each pixel point except the filter hole cluster belongs to the first pixel point based on the final new membership obtaining method
Figure SMS_166
And the final membership degree of each cluster is obtained by a final membership degree obtaining method, so that the membership degree of each pixel in the gray level image is corrected when the traditional FCM clustering method is used for clustering iteration on the pixels in the gray level image, and further, the objective function clustered by the traditional FCM under the condition of iteration stop is obtained based on the final new membership degree
Figure SMS_167
And (4) completely dividing the gray level image into a background area and a filter hole area according to the clustering and dividing result.
And S003, detecting the defects of the filter screen according to the pore size of each filter pore area.
Specifically, in step S002, the polyethersulfone filter screen is divided into a background area and a complete filtration pore area, and defect detection is performed on each filtration pore area, specifically: and obtaining the minimum circumscribed circle of each filtering hole area, obtaining the diameter of the minimum circumscribed circle, taking the diameter as the aperture of the corresponding filtering hole area, and according to the requirement of white spirit generation on the filtering holes, determining that the polyethersulfone filter screen has defects when the aperture does not meet the requirement, otherwise, determining that the polyethersulfone filter screen has no defects, thereby completing the defect detection of the polyethersulfone filter screen.
As an example, the present solution sets the aperture range to
Figure SMS_168
When the aperture is not within the set aperture range, the filter screen is determined to have defects.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit of the present invention are intended to be included therein.

Claims (7)

1. The machine vision-based filter screen defect detection method is characterized by comprising the following steps of:
acquiring a surface image of the filter screen to obtain a corresponding gray image, and performing clustering iteration on pixel points at least twice according to the distance between the pixel points in the gray image to obtain filter hole clusters under each clustering iteration;
in each clustering iteration, acquiring corresponding filter hole communication domains under filter hole clusters, respectively acquiring adjacent filter hole communication domains of each filter hole communication domain according to the distance between the centroids of the filter hole communication domains, and calculating the superposition influence factor of the corresponding interval region according to the gray information of the interval region between the filter hole communication domain and the corresponding adjacent filter hole communication domain; correcting the membership degree of each pixel point in the interval area according to the superposition influence factors to obtain a new membership degree, and dividing the gray level image into a background area and a filter hole area based on the new membership degree;
and detecting the defects of the filter screen according to the pore size of each filter pore area.
2. The machine vision-based defect detection method for the filter screen of the filter machine as claimed in claim 1, wherein the method for obtaining the superposition influence factor comprises:
taking any one filter hole communicated domain as a target communicated domain, taking any one adjacent filter hole communicated domain of the target communicated domain as a first communicated domain, connecting the centroid of the target communicated domain with each edge point on the edge of the first communicated domain to obtain at least two straight lines, obtaining an included angle between any two straight lines, forming a new communicated domain by the two straight lines corresponding to the maximum included angle, the centroid of the target communicated domain and the first communicated domain, and taking the region except the first communicated domain in the new communicated domain as an interval region between the target communicated domain and the first communicated domain;
calculating the Euclidean distance between the centroid of the target connected domain and the centroid of the first connected domain;
calculating an average gray value of the interval area as a first value, calculating an average gray value between the target connected domain and the first connected domain as a second value, acquiring a difference value between the first value and the second value, and using a value obtained by normalizing the difference value as a weight value of the interval area;
calculating information entropy according to the membership degree of each pixel point in the interval area;
and taking the reciprocal of the product of the Euclidean distance, the weight value and the information entropy as a superposition influence factor of an interval region between the target connected domain and the first connected domain.
3. The machine vision-based defect detection method for filter screens of filters according to claim 2, wherein the new membership obtaining method comprises:
respectively obtaining superposition influence factors of an interval region between a target connected domain and each adjacent filter hole connected domain, respectively taking each superposition influence factor corresponding to the target connected domain as an index of a natural constant, taking the obtained result as a first result corresponding to the superposition influence factors, and obtaining an addition result of all the first results;
taking a first result corresponding to an interval area between the target connected domain and the first connected domain as a numerator, and taking an addition result as a denominator to obtain a corresponding ratio; the addition result of the constant 1 and the ratio is used as a correction coefficient;
and regarding any pixel point in the interval area between the target connected domain and the first connected domain, taking the product of the correction coefficient and the membership degree of the pixel point as the new membership degree of the corresponding pixel point.
4. The machine-vision-based filter screen defect detection method of claim 1, wherein the method for segmenting a gray scale image into a background region and a filter hole region based on new membership comprises:
for any pixel point in the gray level image, when the pixel point belongs to pixel points in at least two interval areas, acquiring a new membership degree of the pixel point in each interval area, wherein the pixel point belongs to the same cluster, and taking the maximum new membership degree as a final new membership degree of the pixel point belonging to a corresponding cluster; and segmenting the gray level image into a background area and a filter hole area based on the final new membership degree.
5. The machine vision-based defect detection method for a filter screen of a filter according to claim 1, wherein the method for respectively obtaining the adjacent connected filter hole domains of each connected filter hole domain according to the distance between the centroids of the connected filter hole domains comprises the following steps:
and for any filter hole communicating domain, calculating the centroid distance between the filter hole communicating domain and other filter hole communicating domains, and selecting the other filter hole communicating domain corresponding to the minimum centroid distance as the adjacent filter hole communicating domain of the filter hole communicating domain.
6. The machine vision-based defect detection method for filter screens of filters according to claim 1, wherein the method for detecting defects of the filter screens according to the pore size of each filter pore area comprises the following steps:
and obtaining the minimum circumcircle of each filter hole area, obtaining the diameter of the minimum circumcircle, taking the diameter as the aperture of the corresponding filter hole area, and confirming that the filter screen has defects when the aperture is not in the set aperture range.
7. The machine vision-based defect detection method for the filter screen of the filter as claimed in claim 1, wherein the method for performing at least two clustering iterations on pixel points according to the distance between the pixel points in the gray image to obtain the filter hole cluster class under each clustering iteration comprises:
using conventional FCMPerforming clustering iteration on pixel points in the gray level image at least twice by using a clustering objective function to obtain a filter hole cluster class under each clustering iteration, wherein the objective function
Figure QLYQS_1
Comprises the following steps:
Figure QLYQS_2
wherein ,
Figure QLYQS_5
expressing the total number of pixel points in the gray level image;
Figure QLYQS_9
is an index of a cluster class, represents
Figure QLYQS_12
A cluster class;
Figure QLYQS_4
representing the clustering algorithm to divide the gray scale image into a number of clusters for clustering
Figure QLYQS_10
Class i and is
Figure QLYQS_13
The maximum value of the value;
Figure QLYQS_15
for the ith pixel point belonging to the ith
Figure QLYQS_3
Membership of individual clusters;
Figure QLYQS_7
is membership fuzzy weighted index;
Figure QLYQS_11
is shown as
Figure QLYQS_14
Cluster center points of individual clusters;
Figure QLYQS_6
is a norm;
Figure QLYQS_8
is the ith pixel point.
CN202310113988.1A 2023-02-15 2023-02-15 Filter screen defect detection method based on machine vision Active CN115830022B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310113988.1A CN115830022B (en) 2023-02-15 2023-02-15 Filter screen defect detection method based on machine vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310113988.1A CN115830022B (en) 2023-02-15 2023-02-15 Filter screen defect detection method based on machine vision

Publications (2)

Publication Number Publication Date
CN115830022A true CN115830022A (en) 2023-03-21
CN115830022B CN115830022B (en) 2023-04-28

Family

ID=85521414

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310113988.1A Active CN115830022B (en) 2023-02-15 2023-02-15 Filter screen defect detection method based on machine vision

Country Status (1)

Country Link
CN (1) CN115830022B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116128880A (en) * 2023-04-18 2023-05-16 东莞市京品精密模具有限公司 Identification method for surface cracking of tab mold
CN116542982A (en) * 2023-07-07 2023-08-04 山东中泳电子股份有限公司 Departure judgment device defect detection method and device based on machine vision
CN116721108A (en) * 2023-08-11 2023-09-08 山东奥晶生物科技有限公司 Stevioside product impurity detection method based on machine vision
CN116740061A (en) * 2023-08-14 2023-09-12 山东淼珠生物科技有限公司 Visual detection method for production quality of explosive beads
CN116912256A (en) * 2023-09-14 2023-10-20 山东大昌纸制品有限公司 Corrugated paper rib defect degree assessment method based on image processing
CN117078688A (en) * 2023-10-17 2023-11-17 江苏普隆磁电有限公司 Surface defect identification method for strong-magnetic neodymium-iron-boron magnet

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109389608A (en) * 2018-10-19 2019-02-26 山东大学 There is the fuzzy clustering image partition method of noise immunity using plane as cluster centre
WO2021007744A1 (en) * 2019-07-15 2021-01-21 广东工业大学 Kernel fuzzy c-means fast clustering algorithm with integrated spatial constraints
CN113538429A (en) * 2021-09-16 2021-10-22 海门市创睿机械有限公司 Mechanical part surface defect detection method based on image processing
CN113723449A (en) * 2021-07-16 2021-11-30 西安邮电大学 Preference information-based agent-driven multi-objective evolutionary fuzzy clustering method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109389608A (en) * 2018-10-19 2019-02-26 山东大学 There is the fuzzy clustering image partition method of noise immunity using plane as cluster centre
WO2021007744A1 (en) * 2019-07-15 2021-01-21 广东工业大学 Kernel fuzzy c-means fast clustering algorithm with integrated spatial constraints
CN113723449A (en) * 2021-07-16 2021-11-30 西安邮电大学 Preference information-based agent-driven multi-objective evolutionary fuzzy clustering method
CN113538429A (en) * 2021-09-16 2021-10-22 海门市创睿机械有限公司 Mechanical part surface defect detection method based on image processing

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
MACIEJ K WOZNIAK等: "Automatic Generation of Individual Fuzzy Cognitive Maps from Longitudinal Data" *
张敏;徐栋;: "局部一致性的信息熵Relief特征加权算法" *
池桂英等: "基于分层的直觉模糊Ｃ均值聚类图像分割算法" *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116128880A (en) * 2023-04-18 2023-05-16 东莞市京品精密模具有限公司 Identification method for surface cracking of tab mold
CN116128880B (en) * 2023-04-18 2023-06-16 东莞市京品精密模具有限公司 Identification method for surface cracking of tab mold
CN116542982A (en) * 2023-07-07 2023-08-04 山东中泳电子股份有限公司 Departure judgment device defect detection method and device based on machine vision
CN116542982B (en) * 2023-07-07 2023-09-29 山东中泳电子股份有限公司 Departure judgment device defect detection method and device based on machine vision
CN116721108A (en) * 2023-08-11 2023-09-08 山东奥晶生物科技有限公司 Stevioside product impurity detection method based on machine vision
CN116721108B (en) * 2023-08-11 2023-11-03 山东奥晶生物科技有限公司 Stevioside product impurity detection method based on machine vision
CN116740061A (en) * 2023-08-14 2023-09-12 山东淼珠生物科技有限公司 Visual detection method for production quality of explosive beads
CN116740061B (en) * 2023-08-14 2023-11-21 山东淼珠生物科技有限公司 Visual detection method for production quality of explosive beads
CN116912256A (en) * 2023-09-14 2023-10-20 山东大昌纸制品有限公司 Corrugated paper rib defect degree assessment method based on image processing
CN116912256B (en) * 2023-09-14 2023-11-28 山东大昌纸制品有限公司 Corrugated paper rib defect degree assessment method based on image processing
CN117078688A (en) * 2023-10-17 2023-11-17 江苏普隆磁电有限公司 Surface defect identification method for strong-magnetic neodymium-iron-boron magnet
CN117078688B (en) * 2023-10-17 2024-01-02 江苏普隆磁电有限公司 Surface defect identification method for strong-magnetic neodymium-iron-boron magnet

Also Published As

Publication number Publication date
CN115830022B (en) 2023-04-28

Similar Documents

Publication Publication Date Title
CN115830022A (en) Filter screen defect detection method based on machine vision
CN115018828B (en) Defect detection method for electronic component
CN115330767B (en) Method for identifying production abnormity of corrosion foil
CN116735612B (en) Welding defect detection method for precise electronic components
CN115187602A (en) Injection molding part defect detection method and system based on image processing
CN114926436A (en) Defect detection method for periodic pattern fabric
CN117197140B (en) Irregular metal buckle forming detection method based on machine vision
CN115311262A (en) Printed circuit board defect identification method
CN116416268B (en) Method and device for detecting edge position of lithium battery pole piece based on recursion dichotomy
CN116703251B (en) Rubber ring production quality detection method based on artificial intelligence
CN117994154B (en) Intelligent image denoising method based on sensor
CN116416252B (en) Method for detecting sedimentation image of wastewater in boehmite production process
CN117934467A (en) Titanium metal welding quality detection method and system based on image processing
CN112288760A (en) Adherent cell image screening method and system and cell image analysis method
CN113673396B (en) Spore germination rate calculation method, device and storage medium
CN112396580A (en) Circular part defect detection method
CN113902667B (en) Thread turning identification method and system for machine vision
CN117314899B (en) Carbon fiber plate quality detection method based on image characteristics
CN113034403A (en) Working method for denoising medical image data through Gaussian filtering
CN114998632B (en) RVIN detection and removal method based on pixel clustering
CN118470004B (en) Method, equipment and medium for detecting concentrated dirt defect of LED semiconductor chip
CN116894841B (en) Visual detection method for quality of alloy shell of gearbox
CN118470654B (en) Distributor quality detection method based on image processing
CN118096766B (en) Surface defect detection method for quickly disassembling and assembling portable sucker
CN117237389B (en) CT image segmentation method for middle ear cholesteatoma

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