CN109741297A - Product quality detection method and device - Google Patents
Product quality detection method and device Download PDFInfo
- Publication number
- CN109741297A CN109741297A CN201811437064.2A CN201811437064A CN109741297A CN 109741297 A CN109741297 A CN 109741297A CN 201811437064 A CN201811437064 A CN 201811437064A CN 109741297 A CN109741297 A CN 109741297A
- Authority
- CN
- China
- Prior art keywords
- product
- detected
- lead
- largest contours
- defect
- 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.)
- Pending
Links
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The embodiment of the present invention provides a kind of product quality detection method and device, this method comprises: detection device first obtains the corresponding images to be recognized of product to be detected.Then, this images to be recognized is input to the first categorization module, so that the first disaggregated model carries out Classification and Identification to image, tentatively really going out product to be detected whether there is lead welding defect.Lead welding defect if it exists, then this images to be recognized can be analyzed and processed by detection device again, and finally determine product to be detected with the presence or absence of lead welding defect based on the analysis results.As it can be seen that product quality detection method provided by the invention includes that two parts first primarily determine product to be detected with the presence or absence of lead welding defect using the first disaggregated model.If it exists, then it carries out second to this product to be detected again to analyze, according to analysis finally to determine product to be detected with the presence or absence of lead welding defect.The detection accuracy of product defect in terms of lead can be improved by above-mentioned classification and analytic process.
Description
Technical field
The present invention relates to automatic measurement technique field more particularly to a kind of product quality detection method and device.
Background technique
In the manufacturing process of electronic device, welding is a kind of common and important process.Since production environment, production are set
The influence of many factors such as standby and production technology, inevitably various defects in the welding process, for example draw
Line tilts, there is offset etc. without lead, lead.And it further results in electronic device and quality problems occurs.
Actually manufacture in, the welded condition of lead be to one of maximum defect type of electronic device performance, therefore,
After carrying out welding procedure, being just particularly important with the presence or absence of lead welding defect for electronic device how is accurately identified.
Summary of the invention
In view of this, the embodiment of the present invention provides a kind of product quality detection method and device, use, to improve detection
As a result accuracy.
In a first aspect, the embodiment of the present invention provides a kind of product quality detection method, comprising:
The corresponding images to be recognized of product to be detected is obtained, includes the product to be detected in the images to be recognized through point
Spot welding region after Welding；
According to the first disaggregated model to the images to be recognized carry out Classification and Identification, with the determination product to be detected whether
There are lead welding defects；
If there are lead welding defects for the product to be detected, the images to be recognized is analyzed and processed, with root
Finally determine the product to be detected with the presence or absence of lead welding defect according to analysis result.
Second aspect, the embodiment of the present invention provide a kind of product quality detection device, comprising:
Module is obtained, includes described in the images to be recognized for obtaining the corresponding images to be recognized of product to be detected
Spot welding region of the product to be detected after spot-welding technology；
Categorization module, for carrying out Classification and Identification to the images to be recognized according to the first disaggregated model, described in determination
Product to be detected whether there is lead welding defect；
Defect determining module, if there are lead welding defects for the product to be detected, to the images to be recognized
It is analyzed and processed, finally to determine the product to be detected with the presence or absence of lead welding defect based on the analysis results.
Product quality detection method provided in an embodiment of the present invention, it is corresponding wait know that detection device first obtains product to be detected
Other image includes the spot welding region that product to be detected is formed after spot-welding technology in images to be recognized.Then, this is to be identified
Image is input to the first categorization module, so that the first disaggregated model carries out Classification and Identification to image, tentatively really to go out production to be detected
Product whether there is lead welding defect.If the first disaggregated model primarily determined product to be detected there are lead welding defect,
Then this images to be recognized can be analyzed and processed by detection device again, and finally whether determine product to be detected based on the analysis results
There are lead welding defects.As it can be seen that product quality detection method provided by the invention includes that two parts first use first point
Class model primarily determines product to be detected with the presence or absence of lead welding defect.If it exists, then the is carried out to this product to be detected again
Secondary analysis, according to analysis finally to determine product to be detected with the presence or absence of lead welding defect.By above-mentioned classification and divide
Analysis process can significantly improve the detection accuracy of product defect in terms of lead.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the flow chart of product quality detection method embodiment one provided in an embodiment of the present invention；
Fig. 2 is a kind of optionally embodiment of step 103 in Fig. 1 embodiment；
Fig. 3 is another optionally embodiment of step 103 in Fig. 1 embodiment；
Fig. 4 is a kind of flow chart of optionally the first disaggregated model method of determination；
Fig. 5 is the structural schematic diagram of product quality detection device embodiment one provided in an embodiment of the present invention；
Fig. 6 is the structural schematic diagram of electronic equipment embodiment one provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
The term used in embodiments of the present invention is only to be not intended to be limiting merely for for the purpose of describing particular embodiments
The present invention.In the embodiment of the present invention and the "an" of singular used in the attached claims, " described " and "the"
It is also intended to including most forms, unless the context clearly indicates other meaning, " a variety of " generally comprise at least two, but not
It excludes to include at least one situation.
It should be appreciated that term "and/or" used herein is only a kind of incidence relation for describing affiliated partner, indicate
There may be three kinds of relationships, for example, A and/or B, can indicate: individualism A, exist simultaneously A and B, individualism B these three
Situation.In addition, character "/" herein, typicallys represent the relationship that forward-backward correlation object is a kind of "or".
It will be appreciated that though XXX may be described in embodiments of the present invention using term first, second, third, etc., but
These XXX should not necessarily be limited by these terms.These terms are only used to for XXX being distinguished from each other out.For example, not departing from implementation of the present invention
In the case where example range, the first XXX can also be referred to as the 2nd XXX, and similarly, the 2nd XXX can also be referred to as the first XXX.
Depending on context, word as used in this " if ", " if " can be construed to " ... when " or
" when ... " or " in response to determination " or " in response to detection ".Similarly, context is depended on, phrase " if it is determined that " or " such as
Fruit detection (condition or event of statement) " can be construed to " when determining " or " in response to determination " or " when detection (statement
Condition or event) when " or " in response to detection (condition or event of statement) ".
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
Include, so that commodity or system including a series of elements not only include those elements, but also including not clear
The other element listed, or further include for this commodity or the intrinsic element of system.In the feelings not limited more
Under condition, the element that is limited by sentence "including a ...", it is not excluded that in the commodity or system for including the element also
There are other identical elements.
Fig. 1 is the flow chart of product quality detection method embodiment one provided in an embodiment of the present invention, the present embodiment provides
The product quality detection method executing subject can be detection device, as shown in Figure 1, this method comprises the following steps:
S101 obtains the corresponding images to be recognized of product to be detected, includes product to be detected in images to be recognized through spot welding
Spot welding region after technique.
After product carries out each road technique processing in the production line, it can all be shot by industrial camera, thus according to the figure clapped
As whether product of the confirmation after each road processing technology be qualified.The scene that the present embodiment and following each embodiments provide can be with
It is the product to be detected for obtain after spot-welding technology to product.Industrial camera can carry out the production to be detected after spot-welding technology to this
Product are shot, to obtain the images to be recognized corresponding to product to be detected.It include spot welding region in images to be recognized, it is optional
Ground can specifically include lead, solder joint and pad in spot welding region.
One specific spot welding scene, welding equipment can be by leads by the pronunciation part in microphone, that is, voice coil spot welding
On pad.Then, the voice coil and pad that industrial camera can be integrally formed a postwelding are shot, thus obtain it is above-mentioned to
Identify image.
S102 carries out Classification and Identification to images to be recognized according to the first disaggregated model, to identify whether product to be detected is deposited
In lead welding defect.
Detection device can directly receive the images to be recognized of industrial camera transmission, and then by this images to be recognized
It is input in the first disaggregated model, so that this first disaggregated model carries out Classification and Identification to images to be recognized, emphasis is to treat knowledge
Spot welding region in other image is identified, so that output category result, shows whether product to be detected is deposited by classification results
In lead welding defect.
Wherein, optionally, the classification results of the first disaggregated model output can show in different forms.One kind is optionally
Form, the first disaggregated model can directly export product to be detected with defect type A.Another optionally form, first point
Class model can export the corresponding at least one defect type of product to be detected, that is to say that output product to be detected has various lack
Fall into the probability value of type, such as at least one defect type exported are as follows: defect type I:95%, defect type II:87% are lacked
Fall into type-iii: 40%, defect type IV:15%.The defect information of such form shows that product to be detected has defect type
I, the probability of defect type II, defect type III and defect type IV are respectively 95%, 87%, 40% and 15%.At this point,
First disaggregated model is properly termed as more disaggregated models.
At this point, accepting the example above, for the first form, defect type A then can directly be determined as by detection device
The defect type that product to be detected has.For second of form, detection device then can be by the highest defect type I of probability value
It is determined as the defect type that product to be detected has.
But just as mentioned in the background art, lead welding defect be on product quality influence most serious defect type it
One, therefore, if the first disaggregated model primarily determined out product to be detected there are on the basis of lead welding defect,
Detection device, which can also treat testing product, to carry out second and detects, with further determine that product to be detected with the presence or absence of lead
Welding defect.
Therefore, after step 102, can with the following steps are included:
S103, if product to be detected is analyzed and processed images to be recognized there are lead welding defect, with basis point
Analysis result finally determines product to be detected with the presence or absence of lead welding defect.
If the first disaggregated model determines product to be detected there are welding defect, detection device can be further to production to be detected
The corresponding images to be recognized of product is analyzed and processed, and be that is to say using pre-set image recognizer and is further determined that production to be detected
Product whether there is lead welding defect.That is for being primarily determined by the first disaggregated model as there are lead welding defects
Product to be detected, detection device can recycle pre-set image recognizer to carry out secondary analysis to image to be detected, with what is obtained
Final judgement result.
If analysis is the result shows that product to be detected is there are lead welding defect, detection device finally determines this production to be detected
There are lead welding defects for product.If analysis is the result shows that lead welding defect is not present in product to be detected, detection device is final
Determining this product to be detected, there is no lead welding defects.
It should be noted that product to be detected is possible to number of drawbacks occur under the above-mentioned spot welding scene referred to,
For example rosin joint, lead are tilted, are revealed, wire sweep (specifically including inclined outside inclined and lead in lead), are welded with product to be detected
Pad on without lead (lead missing), be welded on the pad of product to be detected there are excess lead (specifically include single residual line,
Mostly residual line and residual the end of a thread), solder joint there are the shell of impurity or product to be detected exist damage etc..
Wherein, there are usually may be collectively referred to as lead welding defect without lead on excess lead and pad on pad.In reality
In the production of border, no lead may be considered severe defect.Product to be detected with this severe defect would generally be by production line
Manipulator press from both sides out, after artificially handling, spot-welding technology can also be re-started.Single residual line, double residual lines, residual the end of a thread these types
Defect may be considered moderate defects, would generally be discharged by full page, after artificially repairing, put into production again.
In the present embodiment, detection device first obtains the corresponding images to be recognized of product to be detected, includes in images to be recognized
The spot welding region that product to be detected is formed after spot-welding technology.Then, this images to be recognized is input to the first categorization module,
So that the first disaggregated model carries out Classification and Identification to image, it whether there is lead welding defect tentatively really to go out product to be detected.
If the first disaggregated model has primarily determined product to be detected there are lead welding defect, detection device can be to be identified by this again
Image is analyzed and processed, and finally determines product to be detected with the presence or absence of lead welding defect based on the analysis results.As it can be seen that this
The product quality detection method that invention provides includes that two parts first primarily determine product to be detected using the first disaggregated model
With the presence or absence of lead welding defect.If it exists, then this second of product progress to be detected is analyzed again, according to analysis with final true
Fixed product to be detected whether there is lead welding defect.Product can be significantly improved by above-mentioned classification and analytic process drawing
The detection accuracy of defect in terms of line.
The description in above-described embodiment is accepted, there are wire sweep, Fig. 2 when the first disaggregated model determines product to be detected
For a kind of optionally embodiment of step 103 in the embodiment of the present invention one:
S201 generates the corresponding semantic segmentation figure of images to be recognized, includes product to be detected in semantic segmentation figure through spot welding
Lead and pad after technique.
S202 determines lead and the corresponding largest contours of pad and largest contours respectively in semantic segmentation figure
Center.
S203 determines that the angle between the first side of horizontal line largest contours corresponding with lead, horizontal line revolve counterclockwise
Intersect at first when turning with the first side.
S204 is determined according to the positional relationship between the center of angle and the corresponding largest contours of lead and pad
Product to be detected whether there is wire sweep defect.
If the first sorting device determines product to be detected, there are lead welding defects, and this welding defect is specially lead
Defect is deviated, then detection device can further be the corresponding images to be recognized of product to be detected of wire sweep to this defect type
It is analyzed, finally to determine if that there are wire sweeps.It is specifically finally true using preset image recognition algorithm
Fixed product to be detected there are in lead partially or outside lead it is inclined.
Pre-set image recognizer that is to say that above-mentioned steps 201~204, specific implementation process can be described as:
Mr. detection device is at the semantic segmentation figure for corresponding to images to be recognized, for semantic segmentation map generalization, generally
Threshold method (Thresholding methods), the dividing method (Clustering-based based on pixel cluster can be used
Segmentation methods), figure divide dividing method (Graph partitioning segmentation
) or any one of deep learning (Deep learning) methods.Wherein, may include in the semantic segmentation figure of generation
Lead and pad of the product to be detected after spot-welding technology.It can indicate different in this semantic segmentation figure with different colours respectively
The object of type can indicate pad with pink colour in semantic segmentation figure, indicate lead with orange by taking spot welding scene as an example.
Then, in semantic segmentation figure, detection device determines lead and the corresponding largest contours of pad and every again
The center of a largest contours.One horizontal line is set, and as standard, rotates this horizontal line counterclockwise, obtains this horizontal line
The first side that largest contours corresponding with lead intersect at first, determines the angle between this horizontal line and the first side.Wherein, this
It is on one side usually the bottom of the corresponding largest contours of lead.Finally, corresponding most according to angle and lead and pad
Positional relationship between the center of big profile determines product to be detected with the presence or absence of inclined in lead outside inclined or lead.
It should be noted that the lead, pad and the solder joint that refer in the description of above-mentioned and following embodiments respectively correspond to
Largest contours that is to say the corresponding minimum circumscribed rectangle of each.
It, can be right by the center institute of the corresponding largest contours of lead in semantic segmentation figure in order to which subsequent description is succinct
The pixel answered is known as the first central pixel point, and pixel corresponding to the center of the corresponding largest contours of pad is known as in second
The abscissa of imago vegetarian refreshments, pixel coordinate of the two pixels in semantic segmentation figure is respectively x1And x2, then detection device root
Determine product to be detected with the presence or absence of wire sweep according to the positional relationship between angle and two centers are as follows:
If angle is less than predetermined angle and x1< x2, it is determined that there are inclined outside lead for product to be detected.
If angle is greater than or equal to predetermined angle and x1≥x2, it is determined that there are inclined outside lead for product to be detected.
If angle is greater than or equal to predetermined angle and x1< x2, it is determined that there are inclined in lead for product to be detected.
If angle is less than predetermined angle and x1≥x2, it is determined that there are inclined in lead for product to be detected.
Wherein, predetermined angle can be set to 45 °, the angular range between horizontal line and the first side be usually [- 90 °,
90°].Angle, which is less than predetermined angle, indicates pad left avertence, and angle, which is greater than or equal to predetermined angle, indicates pad right avertence.x1< x2Table
Show left side of the center of the corresponding largest contours of lead at the center of the corresponding largest contours of pad, x1≥x2Indicate lead
Right side of the center of corresponding largest contours at the center of the corresponding largest contours of pad.
In practical applications, in semantic segmentation figure in addition to may include lead of the product to be detected after spot-welding technology and
Pad, can also include solder joint, and solder joint is usually indicated with grey.Based on this, whether there is to further increase determining lead
The accuracy of offset, after step 204, detection device can be to handle below further progress: by the corresponding maximum of lead
Contour fitting determines the intersection point of this first straight line largest contours corresponding with solder joint bottom, is corresponded to by lead at first straight line
Largest contours center and this intersection point determine a second straight line.Further calculate the center of the corresponding largest contours of pad
To the distance of this second straight line.And determine that product to be detected is inclined with the presence or absence of lead according to the relationship between distance and pre-determined distance
Move defect.
Specifically, if distance is greater than or equal to pre-determined distance, lead is finally obtained according to above-mentioned definitive result and is deposited
It is inside partially or outer inclined.If distance is less than pre-determined distance, it is determined that wire sweep is not present in product to be detected.
The corresponding content of complex chart 1, Fig. 2, after getting images to be recognized, the first step, detection device is according to the first classification
Model determines product to be detected with the presence or absence of lead welding defect.Lead welding defect and this welding defect is specially if it exists
Wire sweep defect, then second step, detection device can further divide images to be recognized using pre-set image recognizer
Analysis, with determination product to be detected with the presence or absence of inclined in lead outside inclined or lead.Divided using pre-set image recognizer
After analysing result, third step, detection device can also calculate the center of the corresponding largest contours of pad to the distance of second straight line, and root
Finally determine product to be detected with the presence or absence of wire sweep defect according to the size relation between calculated distance and pre-determined distance.
By above-mentioned judgement layer by layer, the accuracy for determining that product to be detected whether there is lead welding defect can be greatly improved, from
And improve the accuracy of product quality detection.
Similarly, the description in above-described embodiment is accepted, there are leads when the first disaggregated model determines product to be detected
When missing or lead are extra, Fig. 3 is another optionally embodiment of step 103 in the embodiment of the present invention one:
S301 generates the corresponding semantic segmentation figure of images to be recognized, includes product to be detected in semantic segmentation figure through spot welding
Lead after technique.
S302 determines that the quantity of the corresponding largest contours of lead and the area of each largest contours are big in semantic segmentation figure
It is small.
S303 determines that product to be detected is lacked with the presence or absence of lead according to the quantity of largest contours and respective size
Or lead is extra.
Detection device generates the semantic segmentation figure for corresponding to this images to be recognized according to the images to be recognized got,
In, in semantic segmentation figure include lead of the product to be detected after spot-welding technology can also include optionally pad and weldering
Point.In semantic segmentation figure, lead, pad and solder joint can be indicated respectively with different color.Then, it is based on semantic segmentation
Figure, detection device can determine the corresponding largest contours of lead, and further calculate the number of the corresponding largest contours of lead
And the area of each largest contours.
If the number of the corresponding largest contours of lead is zero, it is determined that there are lead missings for product to be detected.Wherein, lead
Missing that is to say on the pad for being welded with product to be detected without lead.
If the number of the corresponding largest contours of lead is not zero and the area of the maximum largest contours of area is greater than or waits
In the first preset value, it is determined that there are first kind lead is extra for product to be detected, wherein this first kind lead is extra to be
It is welded on the pad of product to be detected there are whole residual line, whole residual line can specifically include single residual line or double residual lines.?
In practical application, due to the processing mode to the product to be detected that there is the defects of single residual line, double residual lines be it is identical,
Detection device can actually be the presence of single residual line or double residual lines without accurately determining out product to be detected on earth, only need true
Making it, there are whole residual lines.
If the number of largest contours is not zero and the area of the largest contours of maximum area is less than the first preset value, root
Determine that product to be detected is extra with the presence or absence of the second class lead according to the area of the big largest contours of area time.
Specifically, if the area of the secondary big largest contours of area is greater than or equal to the second preset value, it is determined that be detected
There are first kind lead is extra for product, wherein the first preset value is greater than the second preset value.If the face of the big largest contours of area time
Product shows that the area of this largest contours is smaller less than the second preset value, it is determined that product to be detected there are the second class lead is extra,
Wherein, it can be that there are residual the end of a thread on the pad for be welded with product to be detected that this second class lead is extra.
The corresponding content of complex chart 1, Fig. 3, after getting images to be recognized, the first step, detection device is according to the first classification
Model determines product to be detected with the presence or absence of lead welding defect.Lead welding defect and this welding defect is specially if it exists
Lead missing or lead are extra, then second step, and detection device can carry out images to be recognized using pre-set image recognizer
Further analysis is lacked with the presence or absence of lead with determination product to be detected or lead is extra.Detection device can be tied according to analysis
Fruit finally determines that product to be detected is extra with the presence or absence of lead missing or lead.It, can be significantly by above-mentioned judgement layer by layer
The accuracy for determining that product to be detected whether there is lead welding defect is improved, to improve the accuracy of product quality detection.
In addition, for the first disaggregated model referred in FIG. 1 to FIG. 3 illustrated embodiment, it can be using deep learning
Mode training obtains.The training of first disaggregated model is usually completed by processing equipment, this processing equipment can be built in inspection again
In measurement equipment.Optionally, as shown in figure 4, the first disaggregated model can be obtained in the following ways:
S401, obtaining has the testing product of each defect type is corresponding to identify image.
S402, according to preset quantity to having identified that image is grouped, to obtain at least one set of image.
S403 is successively trained using at least one group of image as training data, obtains the first disaggregated model.
Specifically, processing equipment first gets industrial camera and claps to the testing product with all types of defects
Identify image, that is to say, that processing equipment is available to have identified image to multiple, while can also know every identification figure
As corresponding defect type.Optionally, for having identified image, same defect type can will be belonged to according to defect type
Image is saved to a file, to use in subsequent training process.
Then, processing equipment can will identify that image is grouped according to preset quantity, to obtain at least one set of image.
Finally, successively using by every group of image, as input i.e. training data, and finally, training obtains the first disaggregated model.Optionally, often
It can correspond to all or part of defect type in one group of image.
Can be using in the prior art any for the specific training process of the first disaggregated model, the present invention is not right
This is defined.But it is worth noting that during model training, it will usually use to the accuracy that characterizes classification
Loss function.And following loss function Loss:Loss=can be used in the first disaggregated model of training in the embodiment of the present invention
α(1-p)γLog (p), to improve the classification accuracy of the first disaggregated model.Wherein, α is predetermined coefficient, and p is product to be detected tool
There is the probability value of a certain defect type, log (p) is cross entropy.
For above-mentioned to the packet transaction for having identified image, this is typically to the processing capacity for meeting processing equipment.It is right
The sufficiently strong processing equipment of processing capacity is answered, directly can also have been identified obtained whole without above-mentioned packet transaction
The disposable input processing equipment of image obtains the first disaggregated model with training.
In addition, due to generally can not directly use image data during model training, optionally, in step
It can also include: to convert the image of identification in every group of image as binary data before S403；It will identify that image is corresponding
Binary data is associated with the defect type of testing product.
Specifically, after obtaining at least one set of image, processing equipment can also identify image for every in every group of image
The binary data that model training is supported is converted to, for example is converted to tf-record data file.Wherein, tf-record number
Correspond to one according to every binary data in file and identifies image.After converted, then it will each identify that image is corresponding
Binary data be associated with the defect type of testing product.And this association process is it also will be understood that become tf-record number
According to every binary data in file, one defect type label is set.Finally, it is lacked with having in tf-record data file
The binary data for falling into type label is training data, so that training obtains the first disaggregated model.
Fig. 5 is the structural schematic diagram of product quality detection device embodiment one provided in an embodiment of the present invention, such as Fig. 3 institute
Show, which includes: to obtain module 11, categorization module 12 and defect determining module 13.
Module 11 is obtained, includes institute in the images to be recognized for obtaining the corresponding images to be recognized of product to be detected
State spot welding region of the product to be detected after spot-welding technology.
Categorization module 12, for carrying out Classification and Identification to the images to be recognized according to the first disaggregated model, to determine
Product to be detected is stated with the presence or absence of lead welding defect.
Defect determining module 13, if there are lead welding defects for the product to be detected, to the figure to be identified
As being analyzed and processed, finally to determine the product to be detected with the presence or absence of lead welding defect based on the analysis results.
Optionally, first disaggregated model determines the product to be detected there are wire sweep defects；
The defects of product quality detection device determining module 13 specifically includes:
Generation unit 131 is wrapped in the semantic segmentation figure for generating the images to be recognized corresponding semantic segmentation figure
Include lead and pad of the product to be detected after spot-welding technology.
Rectangle determination unit 132, for determining that the lead and the pad respectively correspond in the semantic segmentation figure
Largest contours and the respective center of the largest contours.
Angle determination unit 133, between the first side for determining horizontal line largest contours corresponding with the lead
Angle intersects when the horizontal line rotates counterclockwise with first side at first.
Defect determination unit 134, for according to the angle and the lead and the corresponding maximum of the pad
Positional relationship between the center of profile determines the product to be detected with the presence or absence of wire sweep defect.
Optionally, the defects of product quality detection device determination unit 134 is specifically used for:
If the angle is less than the cross of predetermined angle and the first central pixel point pixel coordinate in the semantic segmentation figure
Less than the second central pixel point, the abscissa of pixel coordinate or the angle in the semantic segmentation figure are greater than or wait coordinate
In the predetermined angle and the corresponding abscissa of first central pixel point is more than or equal to second central pixel point pair
The abscissa answered, it is determined that there are inclined outside lead for the product to be detected, wherein the center of the corresponding largest contours of the lead
For first central pixel point, the center of the corresponding largest contours of the pad is second central pixel point；
If the angle is greater than or equal to the predetermined angle and the corresponding abscissa of first central pixel point is less than
The corresponding abscissa of second central pixel point or the angle are less than the predetermined angle and first center pixel
The corresponding abscissa of point is greater than or equal to the corresponding abscissa of second central pixel point, it is determined that the product to be detected is deposited
In lead partially.
Optionally, product quality detection device further include: fitting module 21, intersection point determining module 22 and computing module
23。
Fitting module 21 is straight for the corresponding largest contours of lead described in the semantic segmentation figure to be fitted to first
Line.
Intersection point determining module 22, for determining the bottom of first straight line largest contours corresponding with the solder joint
Intersection point.
Computing module 23, for calculating the center of the corresponding largest contours of the pad to the distance of second straight line, wherein
The center of largest contours is corresponded to by the lead and the intersection point constitutes the second straight line.
The defect determining module 13 is also used to determine that the product to be detected is inclined with the presence or absence of lead according to the distance
Move defect.
Optionally, the defects of product quality detection device determining module 13 is specifically also used to:
If the distance is greater than or equal to the pre-determined distance, it is determined that there are wire sweeps to lack for the product to be detected
It falls into；And if the distance is less than the pre-determined distance, it is determined that wire sweep defect is not present in the product to be detected.
Optionally, first disaggregated model determines that the product to be detected is lacked there are lead or lead is extra；
Generation unit 131 in the product quality detection device is also used to generate the corresponding semanteme of the images to be recognized
Segmentation figure includes lead of the product to be detected after spot-welding technology in the semantic segmentation figure.
Computing unit 132, for determine the quantity of the corresponding largest contours of lead described in the semantic segmentation figure and
The size of each largest contours.
Defect determination unit 134 is also used to according to the quantity of the largest contours and the determination of respective size
Product to be detected is lacked with the presence or absence of the lead or lead is extra.
Optionally, the defects of product quality detection device determination unit 134 is specifically used for:
If the number of the corresponding largest contours of the lead is not zero and the area of the maximum largest contours of area is greater than
Or it is equal to the first preset value, it is determined that there are first kind lead is extra for the product to be detected；
If the number of the largest contours is not zero and the area of the largest contours of maximum area is less than the first preset value,
Then the area of secondary big largest contours determines that the product to be detected is extra with the presence or absence of the second class lead according to area.
Optionally, the defects of product quality detection device determination unit 134 is specifically used for:
If the area of the big largest contours of area time is greater than or equal to the second preset value, it is determined that the product to be detected is deposited
It is extra in first kind lead, wherein first preset value is greater than second preset value；And if the most bull wheel that area time is big
Wide area is less than the second preset value, it is determined that there are the second class lead is extra for the product to be detected.
Optionally, the defects of product quality detection device determination unit 134 is specifically used for: if the lead is corresponding
The number of largest contours is zero, it is determined that there are lead missings for the product to be detected.
Optionally, the first kind lead is extra to be welded on the pad of the product to be detected there are whole residual line,
The second class lead is extra to be welded on the pad of the product to be detected there are residual the end of a thread, and the lead lacks as welding
Have on the pad of the product to be detected without lead.
Optionally, product quality detection device further include: grouping module 24 and training module 25.
Module 11 is obtained, the testing product with each defect type is corresponding to have identified image for obtaining.
Grouping module 24, for having identified that image is grouped to described according to preset quantity, to obtain at least one set of figure
Picture.
Training module 25 obtains described first for being successively trained using at least one set of image as training data
Disaggregated model.
The method that Fig. 5 shown device can execute FIG. 1 to FIG. 4 illustrated embodiment, the part that the present embodiment is not described in detail,
It can refer to the related description to FIG. 1 to FIG. 4 illustrated embodiment.The implementation procedure and technical effect of the technical solution referring to Fig. 1~
Description in embodiment illustrated in fig. 4, details are not described herein.
The foregoing describe the built-in function of product quality detection device and structures, in a possible design, product matter
The structure of amount detecting device can be realized as an electronic equipment, such as welded condition detector.Fig. 6 is provided in an embodiment of the present invention
The structural schematic diagram of electronic equipment embodiment one as shown in fig. 6, the electronic equipment includes: memory 31, and connects with memory
The processor 32 connect, memory 31 execute the product quality detection side provided in any of the above-described embodiment for storing electronic equipment
The program of method, processor 32 are configurable for executing the program stored in memory 31.
Program includes one or more computer instruction, wherein one or more computer instruction is executed by processor 32
When can be realized following steps:
The corresponding images to be recognized of product to be detected is obtained, includes the product to be detected in the images to be recognized through point
Spot welding region after Welding；
According to the first disaggregated model to the images to be recognized carry out Classification and Identification, with the determination product to be detected whether
There are lead welding defects；
If there are lead welding defects for the product to be detected, the images to be recognized is analyzed and processed, with root
Finally determine the product to be detected with the presence or absence of lead welding defect according to analysis result.
Optionally, processor 32 is also used to execute all or part of the steps in aforementioned approaches method step.
Wherein, it can also include communication interface 33 in the structure of electronic equipment, for electronic equipment and other equipment or lead to
Communication network communication.
The apparatus embodiments described above are merely exemplary, wherein unit can be as illustrated by the separation member
Or may not be and be physically separated, component shown as a unit may or may not be physical unit, i.e.,
It can be located in one place, or may be distributed over multiple network units.It can select according to the actual needs therein
Some or all of the modules achieves the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creative labor
In the case where dynamic, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It is realized by the mode of required general hardware platform is added, naturally it is also possible to which reality is come in conjunction with by way of hardware and software
It is existing.Based on this understanding, substantially the part that contributes to existing technology can be with product in other words for above-mentioned technical proposal
Form embody, which may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD
Deng, including some instructions use is so that a computer installation (can be personal computer, server or network equipment etc.)
The method for executing certain parts of each embodiment or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (12)
1. a kind of product quality detection method characterized by comprising
The corresponding images to be recognized of product to be detected is obtained, includes the product to be detected in the images to be recognized through a welder
Spot welding region after skill；
Classification and Identification is carried out to the images to be recognized according to the first disaggregated model, whether there is with the determination product to be detected
Lead welding defect；
If there are lead welding defects for the product to be detected, the images to be recognized is analyzed and processed, with basis point
Analysis result finally determines the product to be detected with the presence or absence of lead welding defect.
2. the method according to claim 1, wherein first disaggregated model determines that the product to be detected is deposited
In wire sweep defect；
If there are lead welding defects for the product to be detected, the images to be recognized is analyzed and processed, with root
Finally determine the product to be detected with the presence or absence of lead welding defect according to analysis result, comprising:
The corresponding semantic segmentation figure of the images to be recognized is generated, includes the product to be detected in the semantic segmentation figure through point
Lead and pad after Welding；
In the semantic segmentation figure, the lead and the corresponding largest contours of the pad and the maximum are determined
The respective center of profile；
Determine that the angle between the first side of horizontal line largest contours corresponding with the lead, the horizontal line rotate counterclockwise
When intersect at first with first side；
According to the positional relationship between the angle and the lead and the center of the corresponding largest contours of the pad
Determine the product to be detected with the presence or absence of wire sweep defect.
3. according to the method described in claim 2, it is characterized in that, described according to the angle and the lead and the weldering
Positional relationship between the center of the corresponding largest contours of disk determines that the product to be detected is lacked with the presence or absence of wire sweep
It falls into, comprising:
If the angle is less than the abscissa of predetermined angle and the first central pixel point pixel coordinate in the semantic segmentation figure
Less than the second central pixel point, the abscissa of pixel coordinate or the angle are greater than or equal to institute in the semantic segmentation figure
It states predetermined angle and the corresponding abscissa of first central pixel point is corresponding more than or equal to second central pixel point
Abscissa, it is determined that there are inclined outside lead for the product to be detected, wherein the center of the corresponding largest contours of the lead is institute
State the first central pixel point, the center of the corresponding largest contours of the pad is second central pixel point；
If the angle is greater than or equal to the predetermined angle and the corresponding abscissa of first central pixel point is less than described
The corresponding abscissa of second central pixel point or the angle are less than the predetermined angle and first central pixel point pair
The abscissa answered is greater than or equal to the corresponding abscissa of second central pixel point, it is determined that the product to be detected, which exists, to be drawn
In line partially.
4. according to the method described in claim 3, it is characterized in that, further including the product to be detected in the semantic segmentation figure
Solder joint after spot-welding technology, the method also includes:
The corresponding largest contours of lead described in the semantic segmentation figure are fitted to first straight line；
Determine the intersection point of the bottom of first straight line largest contours corresponding with the solder joint；
The center of the corresponding largest contours of the pad is calculated to the distance of second straight line, wherein corresponding maximum by the lead
The center of profile and the intersection point constitute the second straight line；
Determine that the product to be detected whether there is wire sweep defect according to the distance.
5. according to the method described in claim 4, it is characterized in that, described determine that the product to be detected is according to the distance
It is no that there are wire sweep defects, comprising:
If the distance is greater than or equal to the pre-determined distance, it is determined that there are wire sweep defects for the product to be detected；
If the distance is less than the pre-determined distance, it is determined that wire sweep defect is not present in the product to be detected.
6. the method according to claim 1, wherein first disaggregated model determines that the product to be detected is deposited
It is extra in lead missing or lead；
If there are lead welding defects for the product to be detected, the images to be recognized is analyzed and processed, with root
Finally determine the product to be detected with the presence or absence of lead welding defect according to analysis result, comprising:
The corresponding semantic segmentation figure of the images to be recognized is generated, includes the product to be detected in the semantic segmentation figure through point
Lead after Welding；
Determine the size of the quantity of the corresponding largest contours of lead and each largest contours described in the semantic segmentation figure；
Determine the product to be detected with the presence or absence of the lead according to the quantity of the largest contours and respective size
Missing or lead are extra.
7. according to the method described in claim 6, it is characterized in that, the quantity according to the largest contours and respective face
Product size determines that the product to be detected is abnormal with the presence or absence of lead, comprising:
If the number of the corresponding largest contours of the lead is not zero and the area of the maximum largest contours of area is greater than or waits
In the first preset value, it is determined that there are first kind lead is extra for the product to be detected；
If the number of the largest contours is not zero and the area of the largest contours of maximum area is less than the first preset value, root
Determine that the product to be detected is extra with the presence or absence of the second class lead according to the area of the big largest contours of area time.
8. the method according to the description of claim 7 is characterized in that the area of the big largest contours secondary according to area determines
The product to be detected is extra with the presence or absence of the second class lead, comprising:
If the area of time big largest contours of area is greater than or equal to the second preset value, it is determined that there are for the product to be detected
A kind of lead is extra, wherein first preset value is greater than second preset value；
If the area of the big largest contours of area time is less than the second preset value, it is determined that there are the second classes to draw for the product to be detected
Line is extra.
9. according to the method described in claim 6, it is characterized in that, the method also includes:
If the number of the corresponding largest contours of the lead is zero, it is determined that there are lead missings for the product to be detected.
10. method according to any one of claims 6 to 9, which is characterized in that the first kind lead is extra for welding
Have on the pad of the product to be detected there are whole residual line, the second class lead is extra to be welded with the product to be detected
Pad on there are residual the end of a thread, the lead missing is is welded on the pad of the product to be detected without lead.
11. method described according to claim 1 or 2 or 6, which is characterized in that the method also includes:
Obtaining has the testing product of each defect type is corresponding to identify image；
Identify that image is grouped to described according to preset quantity, to obtain at least one set of image；
It is successively trained using at least one set of image as training data, obtains first disaggregated model.
12. a kind of product quality detection device characterized by comprising
Module is obtained, includes described to be checked in the images to be recognized for obtaining the corresponding images to be recognized of product to be detected
Survey spot welding region of the product after spot-welding technology；
Categorization module, it is described to be checked with determination for carrying out Classification and Identification to the images to be recognized according to the first disaggregated model
Surveying product whether there is lead welding defect；
Defect determining module carries out the images to be recognized if there are lead welding defects for the product to be detected
Analysis processing, finally to determine the product to be detected with the presence or absence of lead welding defect based on the analysis results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811437064.2A CN109741297A (en) | 2018-11-28 | 2018-11-28 | Product quality detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811437064.2A CN109741297A (en) | 2018-11-28 | 2018-11-28 | Product quality detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109741297A true CN109741297A (en) | 2019-05-10 |
Family
ID=66358319
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811437064.2A Pending CN109741297A (en) | 2018-11-28 | 2018-11-28 | Product quality detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109741297A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112711603A (en) * | 2020-12-30 | 2021-04-27 | 广东粤云工业互联网创新科技有限公司 | Workpiece detection method and system based on cloud and computer-readable storage medium |
| CN113723219A (en) * | 2021-08-10 | 2021-11-30 | 浙江大华技术股份有限公司 | Method for identifying breakage of insulating member and related device |
| CN117102630A (en) * | 2023-10-24 | 2023-11-24 | 广东美的制冷设备有限公司 | Arc welding quality monitoring method, device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908636A (en) * | 2005-08-04 | 2007-02-07 | 三井金属矿业株式会社 | Film-carried appearance detection method and device for mounting electronic elements |
| CN101334589A (en) * | 2007-06-26 | 2008-12-31 | 大日本网目版制造株式会社 | Exposure pattern data detecting device, method and program |
| CN101499126A (en) * | 2008-02-03 | 2009-08-05 | Juki株式会社 | Image processing apparatus |
| CN108665452A (en) * | 2018-05-09 | 2018-10-16 | 广东大鹏液化天然气有限公司 | A kind of pipeline-weld film scanning storage and identification of Weld Defects and its system based on big data |
-
2018
- 2018-11-28 CN CN201811437064.2A patent/CN109741297A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908636A (en) * | 2005-08-04 | 2007-02-07 | 三井金属矿业株式会社 | Film-carried appearance detection method and device for mounting electronic elements |
| CN101334589A (en) * | 2007-06-26 | 2008-12-31 | 大日本网目版制造株式会社 | Exposure pattern data detecting device, method and program |
| CN101499126A (en) * | 2008-02-03 | 2009-08-05 | Juki株式会社 | Image processing apparatus |
| CN108665452A (en) * | 2018-05-09 | 2018-10-16 | 广东大鹏液化天然气有限公司 | A kind of pipeline-weld film scanning storage and identification of Weld Defects and its system based on big data |
Non-Patent Citations (2)
| Title |
|---|
| 张洵颖等: "《基于Blob分析的芯片管脚检测方法》", 《兰州大学学报（自然科学版）》 * |
| 彭世亮: "《基于图像处理的电路板卡缺陷检测方法研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112711603A (en) * | 2020-12-30 | 2021-04-27 | 广东粤云工业互联网创新科技有限公司 | Workpiece detection method and system based on cloud and computer-readable storage medium |
| CN113723219A (en) * | 2021-08-10 | 2021-11-30 | 浙江大华技术股份有限公司 | Method for identifying breakage of insulating member and related device |
| CN117102630A (en) * | 2023-10-24 | 2023-11-24 | 广东美的制冷设备有限公司 | Arc welding quality monitoring method, device, electronic equipment and storage medium |
| CN117102630B (en) * | 2023-10-24 | 2024-02-27 | 广东美的制冷设备有限公司 | Arc welding quality monitoring method, device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109727229A (en) | Rosin joint detection method and device | |
| US11488294B2 (en) | Method for detecting display screen quality, apparatus, electronic device and storage medium | |
| CN109741295A (en) | Product quality detection method and device | |
| CN111797890A (en) | Method and system for detecting defects of power transmission line equipment | |
| CN111598825B (en) | Data processing method, flaw detection method, computing device and storage medium | |
| CN109741297A (en) | Product quality detection method and device | |
| CN109741296A (en) | Product quality detection method and device | |
| CN111814850A (en) | Defect detection model training method, defect detection method and related device | |
| CN109840883B (en) | Method and device for training object recognition neural network and computing equipment | |
| CN110796640A (en) | Small target defect detection method and device, electronic equipment and storage medium | |
| US20130163869A1 (en) | Apparatus and method for extracting edge in image | |
| CN109829883A (en) | Product quality detection method and device | |
| CN112070134A (en) | Power equipment image classification method and device, power equipment and storage medium | |
| CN112613569B (en) | Image recognition method, training method and device for image classification model | |
| CN114445746A (en) | Model training method, railway contact net abnormity detection method and related device | |
| JP2020060879A (en) | Learning device, image generator, method for learning, and learning program | |
| CN112634201B (en) | Target detection method and device and electronic equipment | |
| CN117440308B (en) | Intelligent control method and system for microphone production line | |
| CN111523558A (en) | Ship shielding detection method and device based on electronic purse net and electronic equipment | |
| CN117392042A (en) | Defect detection method, defect detection apparatus, and storage medium | |
| CN116385430A (en) | Machine vision flaw detection method, device, medium and equipment | |
| CN115471439A (en) | Method and device for identifying defects of display panel, electronic equipment and storage medium | |
| CN116363136B (en) | On-line screening method and system for automatic production of motor vehicle parts | |
| CN112149693A (en) | Training method of contour recognition model and detection method of target object | |
| CN116246077A (en) | Security state detection method, device, equipment and medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |