CN113935999B - Injection molding defect detection method based on image processing - Google Patents
Injection molding defect detection method based on image processing Download PDFInfo
- Publication number
- CN113935999B CN113935999B CN202111541865.5A CN202111541865A CN113935999B CN 113935999 B CN113935999 B CN 113935999B CN 202111541865 A CN202111541865 A CN 202111541865A CN 113935999 B CN113935999 B CN 113935999B
- Authority
- CN
- China
- Prior art keywords
- cluster
- silver
- end region
- clusters
- injection molding
- 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
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Injection Moulding Of Plastics Or The Like (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides an injection molding defect detection method based on image processing, which relates to the technical field of image processing, can distinguish the types of silver line defects and give a process adjustment suggestion according to the types of the defects, and comprises the following steps: collecting an RGB image of the injection molding part, and preprocessing to obtain a silver pattern example binary image; obtaining a special point of the mould; clustering all the silver streak example pixel points to obtain
Point cluster class of silver-like texture instance; acquiring the convergence and divergence direction of the cluster; performing curve fitting on pixel points of all the silver stripe examples in each cluster to obtain corresponding curves, and extending the curves towards the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center; calculating the gathering entropy of a gathering end region and a divergence end region of each cluster; calculating the clustering and scattering ratios of the corresponding various clusters; subjecting various clusters toPerforming primary classification, namely classifying the silver line defect degraded clusters and the clusters to be classified; calculating the distance ratio of the to-be-classified clusters; and performing secondary classification on the clusters to be classified according to the distance comparison.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to an injection molding defect detection method based on image processing.
Background
Injection molding silver streaks in the injection molding industry are common product defects, influence can be caused not only to the appearance of a product, but also to the strength of the product, and the two influences can greatly reduce the product value of an injection molding part. The generation reasons of the crazes are also various, and mainly caused by volatilization of water vapor or degradation gas in the molten material or flowing and shearing of the molten material, the generation reasons correspond to hydrolysis crazes, degradation crazes and shearing crazes respectively. In the injection molding process, the reason for generating the silver streaks is different, and the corresponding processing modes are also different, so that the reason for generating the silver streaks is analyzed while the silver streaks are detected.
In the prior art, the detection method for the silver streak defects is a conventional processing technology such as threshold segmentation, which can only detect the silver streak but cannot distinguish the type of the silver streak, and can deduce the reason for generating the silver streak and adjust and control the process parameters in time by manually distinguishing. The neural network algorithm needs a large number of training pictures and the accuracy rate needs to be improved.
Disclosure of Invention
The invention provides an injection molding defect detection method based on image processing, which comprises the steps of collecting an RGB image of an injection molding piece, and preprocessing the RGB image to obtain a silver pattern example binary image; obtaining a special point of the mould according to the mould parameters of the injection molding part; clustering according to the distribution of all the silver pattern instance pixel points on the binary image to obtain K-type silver pattern instance point clusters; acquiring the convergence and divergence directions of the clusters according to the number and distribution of the pixel points in the clusters; performing curve fitting on pixel points of all the silver streak examples in each cluster to obtain corresponding curves, and extending the curves to the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center corresponding to each cluster; calculating the gathering entropy of the collection end region and the divergence end region of each cluster; calculating the clustering and scattering ratios of the corresponding various clusters according to the clustering entropies of the convergence end region and the divergence end region; performing primary classification on various clusters according to a clustering-scattering ratio, and dividing the various clusters into degraded silver streak defect clusters and to-be-classified clusters; respectively calculating the distance between the collection center of the clusters to be classified and the injection port, the distance between the collection center of the clusters to be classified and the nearest edge rotation center, and the distance ratio; and performing secondary classification on the clusters to be classified according to the distance comparison. Compared with the prior art, the method can adapt to the detection of the silver streak defects of various injection molding parts according to the mold parameters, the silver streak defects cannot be accurately classified in the prior art, only the silver streak defects are classified into one class, the method can classify the silver streak defects based on different generation reasons of the silver streak defects and provide a corresponding defect removing method according to the different generation reasons of the defects, and the defect removing efficiency in the production process is improved.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides an injection molding defect detection method based on image processing, which comprises the following steps:
collecting an injection molding RGB image, and preprocessing the injection molding RGB image to obtain a silver pattern example binary image.
And obtaining a special point of the mold according to the mold parameters of the injection molding part, wherein the special point of the mold is a sprue and an edge rotation center.
And clustering according to the distribution of all the silver pattern instance pixel points on the binary image to obtain K-type silver pattern instance point clusters.
And acquiring the converging and diverging directions of the clusters according to the number and distribution of the pixel points in each cluster.
And performing curve fitting on the pixel points of all the silver streak examples in each cluster to obtain corresponding curves, and extending the curves towards the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center corresponding to each cluster.
And calculating the aggregation entropy of the collection end region and the divergence end region of each cluster.
And calculating the clustering and scattering ratios of the corresponding various clusters according to the clustering entropies of the collection end region and the divergence end region.
And performing primary classification on the clusters according to the clustering-scattering ratio, and dividing the clusters into degraded silver streak defect clusters and to-be-classified clusters.
And respectively calculating the distance between the collection center of the clusters to be classified and the injection port, the distance between the collection center of the clusters to be classified and the nearest edge rotation center and the distance ratio.
And performing secondary classification on the cluster to be classified according to the distance comparison.
Further, the injection molding defect detection method based on image processing calculates the collection entropy of the collection end region and the divergence end region of each cluster, and includes:
for a cluster class, a neighborhood is set.
Introducing labels into the silver print example pixel points and the pixel points in the neighborhoods, counting the total number of the labels of the silver print example pixel points, and calculating the number of the labels of each silver print example pixel point and the average value of the numbers of the labels of the silver print example pixel points in the neighborhoods.
The distribution characteristic expressions of the convergent end region and the divergent end region are respectively as follows:
in the formula
The distribution characteristics of the collection end region are shown,
the distribution characteristics of the divergent terminal area are represented, i represents the number of labels to which the pixel points belong, j represents the average value of the number of labels to which the pixel points belong in the fringe instance in the neighborhood of the pixel points, f (i, j) represents the number of times of occurrence of the binary group (i, j),
representing the total number of silver-line pixel points in the area of the collection end of the cluster,
and the total number of the silver streak pixel points in the divergent end region of the cluster is represented.
The collection entropies of the collection end region and the divergence end region are respectively:
in the formula
Represents a class cluster CkThe entropy of the collection of the end regions is collected,
represents a class cluster CkCollective entropy of the divergent end region, ckAnd (4) representing the number of the silver print instances, namely the total number of the pixel point labels of the silver print instances.
Further, the injection molding defect detection method based on image processing, wherein the obtaining of the converging and diverging directions of the clusters according to the number and distribution of the pixel points in each cluster comprises: for each cluster, obtaining circumscribed circles of all pixel points in the cluster, traversing the diameter of the circumscribed circle, calculating the number of pixel points in the cluster in two semicircles separated by the diameter by taking the horizontal rightward direction as a reference for each diameter, obtaining the offset ratio of the cluster according to the number of the pixel points in the cluster in the two semicircles separated by the diameter, and taking the vertical direction of the maximum offset ratio of the cluster as the converging and diverging direction corresponding to the diameter.
Further, in the injection molding defect detection method based on image processing, the expression of the offset ratio of the cluster is as follows:
in the formula
And represents the offset ratio of the number of pixels in two semicircles separated by the diameter corresponding to the angle theta in the circumscribed circle of the kth cluster, where K is 1, 2, …, K,
indicating the number of such in-cluster pixels in the upper circle divided by this diameter,
indicating the number of pixels in such a cluster in the lower half circle divided by this diameter.
Further, in the injection molding defect detection method based on image processing, the expression of the clustering ratio of the clusters is as follows:
in the formula
Represents a class cluster CkThe ratio of the concentration to the dispersion of (c),
represents a class cluster CkThe entropy of the collection of the end regions is collected,
represents a class cluster CkThe collective entropy of the diverging end regions.
Further, the injection molding defect detection method based on image processing, wherein the clusters are primarily classified according to the clustering-scattering ratio into degraded silver streak defect clusters and to-be-classified clusters, and the method comprises the following steps: normalizing the clustering and scattering ratios of the various clusters, setting a threshold value for the normalized value, and judging that the cluster is a degraded silver streak defect cluster when the normalized value is less than 0.5; and when the normalized value is more than or equal to 0.5, judging the cluster to be classified.
Further, the injection molding defect detection method based on image processing comprises the following steps of performing secondary classification on the clusters to be classified according to the distance comparison: when the distance ratio is less than or equal to 1, judging that the cluster to be classified is a hydrolyzed silver stripe cluster; and when the distance ratio is larger than 1, judging that the cluster to be classified is the cut silver line.
Further, the injection molding defect detection method based on image processing comprises the following steps: graying the injection molding RGB image to obtain the injection molding gray map, detecting the injection molding gray map by adopting a canny operator, detecting silver fringe edge lines, taking each line as a silver fringe example, and carrying out threshold segmentation on the injection molding gray map after edge detection to obtain a silver fringe example binary image after threshold segmentation.
The invention has the beneficial effects that: compared with the prior art, the method can adapt to the detection of the silver streak defects of various injection molding parts according to the mold parameters, the silver streak defects cannot be accurately classified in the prior art, only the silver streak defects are classified into one class, the method can classify the silver streak defects based on different generation reasons of the silver streak defects and provide a corresponding defect removing method according to the different generation reasons of the defects, and the defect removing efficiency in the production process is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of an injection molding defect detection method based on image processing according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of an injection molding defect detection method based on image processing according to an embodiment of the present invention;
FIG. 3 is a schematic view of the edge center of rotation;
FIG. 4 is a schematic view of an example silver streak;
FIG. 5 is a schematic view of a circumscribed circle of a degraded silver streak;
FIG. 6 is a schematic view of the diverging end central axis;
FIG. 7 is a schematic view of a collection end medial axis and a collection center;
FIG. 8 is a diagram of divergent end expansion.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
The embodiment of the invention provides an injection molding defect detection method based on image processing, which comprises the following steps of:
101. collecting an injection molding RGB image, and preprocessing the injection molding RGB image to obtain a silver pattern example binary image.
Among them, injection molding is a method for producing and molding industrial products.
102. And obtaining a special point of the mold according to the mold parameters of the injection molding part, wherein the special point of the mold is a sprue and an edge rotation center.
103. And clustering according to the distribution of all the silver pattern instance pixel points on the binary image to obtain K-type silver pattern instance point clusters.
Clustering is a process of dividing a set of physical or abstract objects into a plurality of classes composed of similar objects, and a class cluster generated by clustering is a set of data objects, wherein the objects are similar to objects in the same class cluster and different from objects in other class clusters.
104. And acquiring the converging and diverging directions of the clusters according to the number and distribution of the pixel points in each cluster.
105. And performing curve fitting on the pixel points of all the silver streak examples in each cluster to obtain corresponding curves, and extending the curves towards the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center corresponding to each cluster.
106. And calculating the aggregation entropy of the collection end region and the divergence end region of each cluster.
107. And calculating the clustering and scattering ratios of the corresponding various clusters according to the clustering entropies of the collection end region and the divergence end region.
108. And performing primary classification on the clusters according to the clustering-scattering ratio, and dividing the clusters into degraded silver streak defect clusters and to-be-classified clusters.
109. And respectively calculating the distance between the collection center of the clusters to be classified and the injection port, the distance between the collection center of the clusters to be classified and the nearest edge rotation center and the distance ratio.
110. And performing secondary classification on the cluster to be classified according to the distance comparison.
Compared with the prior art, the method can adapt to the detection of the silver streak defects of various injection molding parts according to the mold parameters, the silver streak defects cannot be accurately classified in the prior art, only the silver streak defects are classified into one class, the method can classify the silver streak defects based on different generation reasons of the silver streak defects and provide a corresponding defect removing method according to the different generation reasons of the defects, and the defect removing efficiency in the production process is improved.
Example 2
The embodiment of the invention provides an injection molding defect detection method based on image processing, which comprises the following steps of:
201. collecting an injection molding RGB image, and preprocessing the injection molding RGB image to obtain a silver pattern example binary image.
And acquiring an RGB image of the surface of the injection molding part on an automatic production line.
202. The pretreatment comprises the following steps: graying the injection molding RGB image to obtain the injection molding gray map, detecting the injection molding gray map by adopting a canny operator, detecting silver fringe edge lines, taking each line as a silver fringe example, and carrying out threshold segmentation on the injection molding gray map after edge detection to obtain a silver fringe example binary image after threshold segmentation.
Detecting an image after injection molding RGB image graying by using a canny operator, wherein a plurality of silver line edge lines can be detected, each line is used as a silver line example, threshold segmentation based on a gray histogram is carried out on the whole image after edge detection, the lowest point in the gray histogram of the image after canny operator processing is used as a threshold, wherein the silver line is generally close to white and is set to 1, and the background is set to zero to obtain a binary image of the silver line example after threshold segmentation.
203. And obtaining a special point of the mold according to the mold parameters of the injection molding part, wherein the special point of the mold is a sprue and an edge rotation center.
The injection nozzle (i.e. the position of the molten material entering the mold cavity) of the injection molding part and the rotation center of the corner of the mold edge are obtained according to the mold parameters of the injection molding part. The edge rotation center is shown in fig. 3.
And obtaining the preprocessed silver pattern example binary image and each special point position on the mold.
The priori knowledge shows that the silver streaks generated by the same reason are generally distributed at the same position, the defect silver streaks generated by different reasons are distributed differently, the distribution positions of the degraded silver streaks are irregular, and the degraded silver streaks are generally distributed in a way that one side of the degraded silver streaks are more in number and one side of the degraded silver streaks are fewer in number, but the degraded silver streaks are obviously comet-shaped. The shearing silver stripes are mostly distributed at the corners of the edges of the die, and the direction of the silver stripes is vertical to the edges. The hydrolyzed silver streaks are regularly distributed along the direction of melt flow, generally toward the sprue.
Based on the above logic, the present embodiment performs clustering according to the distribution of all the silver streak instance pixel points.
204. And clustering according to the distribution of all the silver pattern instance pixel points on the binary image to obtain K-type silver pattern instance point clusters.
As shown in fig. 4, mean shift clustering is performed on all the silver print example pixel points in the graph based on positions to obtain K types of silver print example point clusters,these clusters are numbered as: c1,C2,…,CK. In this embodiment, K is 3, and these clusters respectively include c1,c2,…,cKAn example of silver streaks.
205. And acquiring the converging and diverging directions of the clusters according to the number and distribution of the pixel points in each cluster.
206. The obtaining of the convergence and divergence directions of the clusters according to the number and distribution of the pixels in each cluster comprises: for each cluster, obtaining circumscribed circles of all pixel points in the cluster, traversing the diameter of the circumscribed circle, calculating the number of pixel points in the cluster in two semicircles separated by the diameter by taking the horizontal rightward direction as a reference for each diameter, obtaining the offset ratio of the cluster according to the number of the pixel points in the cluster in the two semicircles separated by the diameter, and taking the vertical direction of the maximum offset ratio of the cluster as the converging and diverging direction corresponding to the diameter.
Taking the degraded silver streaks as an example, the circumscribed circle of all the silver streak example pixel points of the degraded silver streak cluster is shown in fig. 5, the angle of the diameter relative to the reference direction is theta by taking the horizontal right as the reference, and the theta is in the range of 0 and pi. In the kth cluster, each diameter corresponds to an angle theta, and the number offset ratio of the two semicircle pixel points obtained according to the diameter
After traversing all the diameters, the perpendicular direction of the diameter corresponding to the maximum deviation ratio is the convergent-divergent direction. According to the priori knowledge (the number of the silver streak examples in the collection direction is small), the semicircle side with the large number of the pixel points is the divergence direction, and the semicircle side with the small number of the pixel points is the collection direction.
Thus, the converging and diverging directions of the various clusters can be obtained.
207. The expression of the offset ratio of the class cluster is as follows:
in the formula
And represents the offset ratio of the number of pixels in two semicircles separated by the diameter corresponding to the angle theta in the circumscribed circle of the kth cluster, where K is 1, 2, …, K,
indicating the number of such in-cluster pixels in the upper circle divided by this diameter,
indicating the number of pixels in such a cluster in the lower half circle divided by this diameter.
208. And performing curve fitting on the pixel points of all the silver streak examples in each cluster to obtain corresponding curves, and extending the curves towards the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center corresponding to each cluster.
Because the lengths of the silver wire examples in each cluster are different, in order to compare the extension direction difference of the silver wire examples at the collecting end and the scattering end, curve fitting needs to be carried out on each silver wire to estimate the extension direction of the silver wire
Cluster C of classkC in (K is 1, 2, …, K)kCarrying out curve fitting based on pixel point positions on the silver stripe examples to obtain corresponding ckAnd the curves extend towards the converging direction and the diverging direction respectively, and are respectively a diverging end middle axis as shown in fig. 6, and the diverging end middle axis is a tangent line passing through the intersection point of the circumscribed circle and the diverging direction and making the circumscribed circle. For the converging direction, each curve is extended all the time, intersection points of all curves and the normal lines at all positions in the converging and diverging directions are intercepted, one-dimensional distribution ranges of the intersection points of the normal lines at all the positions and all the curves are counted, the normal line with the minimum distribution range is taken as a converging end central axis, as shown in fig. 7, a converging end central axis and a converging center are taken, and the converging center is the intersection point of the converging end central axis and the converging and diverging directions.
The central axes of the convergence end region and the divergence end region are expanded along the convergence and divergence directions, the expanded width is W (a threshold is set by itself, and one suggested amount is the radius length of a circumscribed circle of the cluster region), and the length L is just to include all the silver stripe example pixel points in the width W, so that the pixel points and the width become a bounding box of a silver stripe example segment, as shown in fig. 8, the expansion diagram of the divergence end is the same as the convergence end, and detailed description is omitted.
209. And calculating the aggregation entropy of the collection end region and the divergence end region of each cluster.
The aggregation entropy is used for measuring the complexity of a silver pattern instance to which a certain silver pattern pixel point belongs and silver pattern instances to which the silver pattern pixel points in the surrounding neighborhood belong, namely, a certain point can belong to one silver pattern instance, and the more the silver pattern instances to which the pixel points in the surrounding neighborhood belong, the closer the pixel point position distribution on different silver pattern instances is, namely, the more aggregation is, the larger the aggregation entropy is.
2091. For a cluster class, a neighborhood is set.
2092. Introducing labels into the silver print example pixel points and the pixel points in the neighborhoods, counting the total number of the labels of the silver print example pixel points, and calculating the number of the labels of each silver print example pixel point and the average value of the numbers of the labels of the silver print example pixel points in the neighborhoods.
Firstly setting a neighborhood range, wherein the neighborhood range gamma is a preset value, and aiming at a cluster CkTo say, it contains ckAn example of a silver streak is
Each silver print instance pixel point corresponds to a label of one silver print instance, and for some common points, the common points have a plurality of labels simultaneously. And (3) introducing a label to which each silver pattern instance pixel point belongs and a label to which the silver pattern instance pixel point in the neighborhood of each silver pattern instance pixel point belongs, calculating the number of the labels to which each silver pattern instance pixel point belongs and the average value of the numbers of the labels to which the silver pattern instance pixel points in the neighborhood of each silver pattern instance pixel point belong, forming a binary group, and marking as (i, j).
2093. The distribution characteristic expressions of the convergent end region and the divergent end region are respectively as follows:
in the formula
The distribution characteristics of the collection end region are shown,
the distribution characteristics of the divergent terminal area are represented, i represents the number of labels to which the pixel points belong, j represents the average value of the number of labels to which the pixel points belong in the fringe instance in the neighborhood of the pixel points, f (i, j) represents the number of times of occurrence of the binary group (i, j),
representing the total number of silver-line pixel points in the area of the collection end of the cluster,
and the total number of the silver streak pixel points in the divergent end region of the cluster is represented.
2094. The collection entropies of the collection end region and the divergence end region are respectively:
in the formula
Represents a class cluster CkThe entropy of the collection of the end regions is collected,
represents a class cluster CkCollective entropy of the divergent end region, ckAnd (4) representing the number of the silver print instances, namely the total number of the pixel point labels of the silver print instances.
210. And calculating the clustering and scattering ratios of the corresponding various clusters according to the clustering entropies of the collection end region and the divergence end region.
211. The expression of the clustering ratio of the cluster is as follows:
in the formula
Represents a class cluster CkThe ratio of the concentration to the dispersion of (c),
represents a class cluster CkThe entropy of the collection of the end regions is collected,
represents a class cluster CkThe collective entropy of the diverging end regions.
212. And performing primary classification on the clusters according to the clustering-scattering ratio, and dividing the clusters into degraded silver streak defect clusters and to-be-classified clusters.
The smaller the convergence-divergence ratio is, the smaller the difference between the convergence entropy of the convergence end region and the divergence end region is, and the smaller the difference between the convergence entropy of the convergence end region and the divergence end region is, the smaller the difference between the convergence end region and the divergence end region is, the more likely the cluster is to be a degraded silver streak defect.
213. The primary classification of the clusters is divided into degradation silver streak defect clusters and clusters to be classified according to the clustering-scattering ratio, and the classification comprises the following steps: normalizing the clustering and scattering ratios of the various clusters, setting a threshold value for the normalized value, and judging that the cluster is a degraded silver streak defect cluster when the normalized value is less than 0.5; and when the normalized value is more than or equal to 0.5, judging the cluster to be classified.
And normalizing the obtained clustering and scattering ratios of the various clusters, wherein the expression is as follows:
in the formula
Represents a class cluster CkResults after normalization of the vergence ratio for cluster CkWhen it corresponds to
Judging as degradation silver streak defect when the content is less than 0.5, and judging as the corresponding
And when the number is more than or equal to 0.5, the cluster is a class cluster to be classified.
214. And respectively calculating the distance between the collection center of the clusters to be classified and the injection port, the distance between the collection center of the clusters to be classified and the nearest edge rotation center and the distance ratio.
For all the clusters to be classified comprising hydrolyzed silver streak clusters and sheared silver streak clusters, a collection end region of the hydrolyzed silver streak clusters is close to a sprue, a collection end region of the sheared silver streak clusters is close to an edge rotation center, and the distance between the collection center of the clusters to be classified and the sprue is calculated based on the logic
Distance between collection center of cluster to be classified and nearest edge rotation center
And a distance ratio.
215. And performing secondary classification on the cluster to be classified according to the distance comparison.
216. The method for performing secondary classification on the cluster to be classified according to the distance comparison comprises the following steps: when the distance ratio is less than or equal to 1, judging that the cluster to be classified is a hydrolyzed silver stripe cluster; and when the distance ratio is larger than 1, judging that the cluster to be classified is the cut silver line.
For class C to be classifiedkCorrespond to
And
by passing
And
the ratio of (2) to (2) determines the class of the class cluster when
And
when the ratio of (1) is less than or equal to 1, judging that the cluster to be classified is a hydrolyzed silver streak cluster; when in use
And
when the ratio of (A) to (B) is more than 1, judging that the cluster to be classified is the cut silver line
Compared with the prior art, the method can adapt to the detection of the silver streak defects of various injection molding parts according to the mold parameters, the silver streak defects cannot be accurately classified in the prior art, only the silver streak defects are classified into one class, the method can classify the silver streak defects based on different generation reasons of the silver streak defects and provide a corresponding defect removing method according to the different generation reasons of the defects, and the defect removing efficiency in the production process is improved.
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 fall within the spirit and principle of the present invention are intended to be included therein.
Claims (6)
1. An injection molding defect detection method based on image processing is characterized by comprising the following steps:
collecting an injection molding RGB image, and preprocessing the injection molding RGB image to obtain a silver pattern example binary image;
obtaining a special point of a mould according to mould parameters of the injection molding part, wherein the special point of the mould is a sprue and an edge rotation center;
clustering is carried out according to the distribution of all silver streak instance pixel points on the binary image to obtain
Point cluster class of silver-like texture instance;
acquiring the convergence and divergence directions of the clusters according to the number and distribution of the pixel points in each cluster;
performing curve fitting on pixel points of all the silver streak examples in each cluster to obtain corresponding curves, and extending the curves to the converging and diverging directions respectively to obtain a converging end region, a diverging end region and a converging center corresponding to each cluster;
calculating the aggregation entropy of the collection end region and the divergence end region of each cluster;
calculating the clustering and scattering ratios of the corresponding various clusters according to the clustering entropies of the collection end region and the divergence end region;
performing primary classification on each cluster according to the clustering-scattering ratio, and dividing the cluster into a degraded silver streak defect cluster and a cluster to be classified;
respectively calculating the distance between the collection center of the clusters to be classified and the injection port, the distance between the collection center of the clusters to be classified and the nearest edge rotation center, and the distance ratio;
performing secondary classification on the cluster to be classified according to the distance comparison;
calculating the aggregation entropy of the collection end region and the divergence end region of each cluster, including:
setting a neighborhood for a cluster class;
introducing labels into the silver print example pixel points and the pixel points in the neighborhoods, counting the total number of the labels of the silver print example pixel points, and calculating the number of the labels of each silver print example pixel point and the average value of the numbers of the labels of the silver print example pixel points in the neighborhoods of each silver print example pixel point;
the distribution characteristic expressions of the convergent end region and the divergent end region are respectively as follows:
in the formula
The distribution characteristics of the collection end region are shown,
the distribution characteristic of the divergent end region is shown,
the number of the labels to which the pixel points belong is represented,
the average value of the number of labels of the silver streak instance pixel points in the pixel point neighborhood is represented,
representing doublets
The number of times of occurrence of the event,
representing the total number of silver-line pixel points in the area of the collection end of the cluster,
representing the total number of silver streak pixel points in the divergent end region of the cluster;
the collection entropies of the collection end region and the divergence end region are respectively:
in the formula
Representing a class cluster
The entropy of the collection of the end regions is collected,
representing a class cluster
The entropy of the collection of the divergent end region,
representing the number of the silver print examples, namely the total number of the pixel point labels of the silver print examples;
the expression of the clustering ratio of the cluster is as follows:
in the formula
Representing a class cluster
The ratio of the concentration to the dispersion of (c),
representing a class cluster
The entropy of the collection of the end regions is collected,
representing a class cluster
The collective entropy of the diverging end regions.
2. The injection molding defect detection method based on image processing according to claim 1, wherein the obtaining of the convergent-divergent directions of the clusters according to the number and distribution of the pixels in each cluster comprises: for each cluster, obtaining circumscribed circles of all pixel points in the cluster, traversing the diameter of the circumscribed circle, calculating the number of pixel points in the cluster in two semicircles separated by the diameter by taking the horizontal rightward direction as a reference for each diameter, obtaining the offset ratio of the cluster according to the number of the pixel points in the cluster in the two semicircles separated by the diameter, and taking the vertical direction of the maximum offset ratio of the cluster as the converging and diverging direction corresponding to the diameter.
3. An injection molding defect detection method based on image processing as claimed in claim 2, wherein the offset ratio of the cluster type is expressed as:
in the formula
Is shown as
The corresponding angle in the circumscribed circle of the cluster-like body is
The number of pixels in two semicircles separated by the diameter of (a) is offset by a ratio,
=1,2,…,
，
indicating the number of such in-cluster pixels in the upper circle divided by this diameter,
indicating the number of pixels in such a cluster in the lower half circle divided by this diameter.
4. The method for detecting injection molding defects based on image processing according to claim 1, wherein the primary classification of the clusters into degraded silver streak defect clusters and to-be-classified clusters according to the clustering-scattering ratio comprises: normalizing the clustering and scattering ratios of the various clusters, setting a threshold value for the normalized value, and judging that the cluster is a degraded silver streak defect cluster when the normalized value is less than 0.5; and when the normalized value is more than or equal to 0.5, judging the cluster to be classified.
5. The injection molding defect detection method based on image processing as claimed in claim 1, wherein the method for secondarily classifying the clusters to be classified according to the distance ratio comprises: when the distance ratio is less than or equal to 1, judging that the cluster to be classified is a hydrolyzed silver stripe cluster; and when the distance ratio is larger than 1, judging that the cluster to be classified is the cut silver line.
6. An injection molding defect detection method based on image processing as claimed in claim 1, wherein the preprocessing comprises: graying the injection molding RGB image to obtain the injection molding gray map, detecting the injection molding gray map by adopting a canny operator, detecting silver fringe edge lines, taking each line as a silver fringe example, and carrying out threshold segmentation on the injection molding gray map after edge detection to obtain a silver fringe example binary image after threshold segmentation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111541865.5A CN113935999B (en) | 2021-12-16 | 2021-12-16 | Injection molding defect detection method based on image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111541865.5A CN113935999B (en) | 2021-12-16 | 2021-12-16 | Injection molding defect detection method based on image processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113935999A CN113935999A (en) | 2022-01-14 |
| CN113935999B true CN113935999B (en) | 2022-03-18 |
Family
ID=79289192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111541865.5A Active CN113935999B (en) | 2021-12-16 | 2021-12-16 | Injection molding defect detection method based on image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935999B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115018828B (en) * | 2022-08-03 | 2022-10-25 | 深圳市尹泰明电子有限公司 | Defect detection method for electronic component |
| CN115019077A (en) * | 2022-08-09 | 2022-09-06 | 江苏思伽循环科技有限公司 | Method for identifying and controlling shaking table separator in waste battery recycling process |
| CN115147429B (en) * | 2022-09-07 | 2022-11-08 | 深圳市欣冠精密技术有限公司 | Streak detection method for optical glass preform |
| CN116689246B (en) * | 2023-08-01 | 2023-10-03 | 深圳平显科技有限公司 | Multi-channel glue injection control method and device for display screen production |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201522490U (en) * | 2009-09-25 | 2010-07-07 | 重庆嘉良塑胶制品有限责任公司 | Instrument plate simulating device for checking plastic welding fluid water content |
| CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
| CN110222681A (en) * | 2019-05-31 | 2019-09-10 | 华中科技大学 | A kind of casting defect recognition methods based on convolutional neural networks |
| CN110889458A (en) * | 2019-12-03 | 2020-03-17 | 广东工业大学 | Injection molding product defect classification method based on machine learning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7391656B2 (en) * | 2019-12-24 | 2023-12-05 | ファナック株式会社 | injection molding system |
| CN113538429B (en) * | 2021-09-16 | 2021-11-26 | 海门市创睿机械有限公司 | Mechanical part surface defect detection method based on image processing |
-
2021
- 2021-12-16 CN CN202111541865.5A patent/CN113935999B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201522490U (en) * | 2009-09-25 | 2010-07-07 | 重庆嘉良塑胶制品有限责任公司 | Instrument plate simulating device for checking plastic welding fluid water content |
| CN104850858A (en) * | 2015-05-15 | 2015-08-19 | 华中科技大学 | Injection-molded product defect detection and recognition method |
| CN110222681A (en) * | 2019-05-31 | 2019-09-10 | 华中科技大学 | A kind of casting defect recognition methods based on convolutional neural networks |
| CN110889458A (en) * | 2019-12-03 | 2020-03-17 | 广东工业大学 | Injection molding product defect classification method based on machine learning |
Non-Patent Citations (4)
| Title |
|---|
| 076 Solutions for Molding Defects (Silver Streaks);misumi官网;《https://www.misumi-techcentral.com/tt/en/mold/2011/04/076-solutions-for-molding-defects-silver-streaks.html》;20190116;第1-2页 * |
| Modeling and Optimization of the Injection-Molding Process: A Review;António J. Pontes等;《https://www.researchgate.net/publication/295503880》;20160229;第1-22页 * |
| What causes silver streaks or marks during injection molding?;Jackie;《https://www.ecomolding.com/silver-streaks》;20191121;第1-5页 * |
| 基于机器视觉的注塑制品缺陷检测系统研究;孙天瑜;《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》;20120331;第I138-2247页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113935999A (en) | 2022-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113935999B (en) | Injection molding defect detection method based on image processing | |
| CN115082683B (en) | Injection molding defect detection method based on image processing | |
| CN109188459B (en) | Ramp small obstacle identification method based on multi-line laser radar | |
| CN115018828A (en) | Defect detection method for electronic component | |
| CN113409313B (en) | Wheel weld surface defect detection method based on computer vision | |
| CN114565614B (en) | Injection molding surface defect analysis method and system based on machine vision | |
| CN114022483B (en) | Injection molding flash area identification method based on edge characteristics | |
| CN115082419A (en) | Blow-molded luggage production defect detection method | |
| CN108846831B (en) | Band steel surface defect classification method based on combination of statistical characteristics and image characteristics | |
| CN114359190B (en) | Plastic product molding control method based on image processing | |
| CN103810474A (en) | Car plate detection method based on multiple feature and low rank matrix representation | |
| CN109886277B (en) | Contour analysis-based fresh tea leaf identification method | |
| CN114882026B (en) | Sensor shell defect detection method based on artificial intelligence | |
| CN116309577B (en) | Intelligent detection method and system for high-strength conveyor belt materials | |
| CN113393428B (en) | Method for detecting shape of air inlet and outlet edges of aero-engine blade | |
| CN114119603A (en) | Image processing-based snack box short shot defect detection method | |
| CN111709934B (en) | Injection molding impeller warping defect detection method based on point cloud characteristic comparison | |
| CN115880699B (en) | Food packaging bag detection method and system | |
| CN115797361B (en) | Aluminum template surface defect detection method | |
| CN115294116A (en) | Method, device and system for evaluating dyeing quality of textile material based on artificial intelligence | |
| CN107784646A (en) | A kind of road self-adapting detecting method to gather materials | |
| CN115631146A (en) | Image-based pantograph carbon slide strip defect detection method | |
| CN116945521B (en) | Injection molding defect detection method | |
| CN116912234B (en) | Cable stranded wire quality detection method based on image features | |
| CN116862871A (en) | Wood counting method based on mixed characteristics |
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 |