CN105469400B - Method and system for quickly identifying and marking polarity direction of electronic element - Google Patents
Method and system for quickly identifying and marking polarity direction of electronic element Download PDFInfo
- Publication number
- CN105469400B CN105469400B CN201510823564.XA CN201510823564A CN105469400B CN 105469400 B CN105469400 B CN 105469400B CN 201510823564 A CN201510823564 A CN 201510823564A CN 105469400 B CN105469400 B CN 105469400B
- Authority
- CN
- China
- Prior art keywords
- polar orientation
- electronic components
- electronic component
- target
- target electronic
- 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
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 64
- 238000004364 calculation method Methods 0.000 claims abstract description 24
- 238000009826 distribution Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 36
- 230000004913 activation Effects 0.000 claims description 15
- 230000009466 transformation Effects 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 6
- 230000005611 electricity Effects 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004064 recycling Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 239000012141 concentrate Substances 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 239000003990 capacitor Substances 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30141—Printed circuit board [PCB]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a method and a system for quickly identifying and marking the polarity direction of an electronic element. The method comprises the steps of obtaining an image comprising a target electronic element, carrying out forward calculation on the image by utilizing a trained convolutional neural network to obtain the classification characteristics of the polarity direction categories of the target electronic element, obtaining the probability distribution of the target electronic element belonging to various polarity direction categories of various electronic elements, and selecting the polarity direction category with the highest probability as the polarity direction category of the target electronic element. The convolutional neural network is used in the scheme, the polarity direction of the electronic element can be automatically and accurately identified through the convolutional neural network, the convolutional neural network is not specific to a specific electronic element structure, the convolutional neural network is suitable for various electronic elements with polarities, the polarity direction identification of the electronic elements across categories is realized, and the applicability is wide.
Description
Technical field
The present invention relates to automatic optics inspection field, more particularly to a kind of electronic component polar orientation quick identification,
The method and system of mark.
Background technique
Currently, needing carrying out polar orientation judgement to electronic component thereon for pcb board.Now for electronics member
The judgement of part polar orientation, there are mainly two types of methods: first is that the polar orientation based on component structure feature judges that this method is base
It in specific component structure feature, is usually designed just for certain class component, applicability is limited, in addition, part-structure feature is not
It is unstable with (such as illumination is different, shooting angle is different, noise jamming) under environment, cause judging nicety rate lower；Second is that base
Judge in the polar orientation of positive and negative template matching, as long as all setting a template for every kind of element, but in this way may be because more
Class template is mixed and accuracy rate is caused to reduce, as long as and element appearance is slightly different must introduce new template, it is expansible
Property it is limited, calculating the time can also increase because of the increase of template library.
Summary of the invention
Based on this, it is necessary to for the problem that electronic component polar orientation judging nicety rate is low, applicability is limited, provide one
The quick method and system for identifying, marking of kind electronic component polar orientation.
A kind of method for quickly identifying of electronic component polar orientation, comprising the following steps:
Obtain the image comprising target electronic components；
Forward calculation is made to the image data comprising target electronic components using the convolutional neural networks after training, obtains mesh
The characteristic of division for marking the polar orientation classification of electronic component obtains target electronic components according to characteristic of division and belongs to each electron-like member
The probability distribution of the various polar orientation classifications of part；
Choose polar orientation classification of the polar orientation classification of maximum probability as target electronic components.
It is to obtain the image including target electronic components according to above-mentioned method for quickly identifying, the volume after recycling training
Product neural network makees forward calculation to it, obtains the characteristic of division of the polar orientation classification of target electronic components, then obtain target
Electronic component belongs to the probability distribution of the various polar orientation classifications of all kinds of electronic components, chooses the polarity side of wherein maximum probability
Polar orientation classification to classification as target electronic components.Convolutional neural networks have been used in this scheme, have passed through convolutional Neural
Network can automatically and accurately identify the polar orientation of electronic component, and not for specific electronic component structure, be applicable in
In the polar electronic component of various bands, the polar orientation identification of the electronic component across classification is realized, applicability is wider.
In one of the embodiments, using the convolutional neural networks after training to the picture number comprising target electronic components
According to forward calculation is made, the step of obtaining the characteristic of division of the polar orientation classification of target electronic components the following steps are included:
Convolution algorithm is carried out to image data by convolutional layer, nonlinear transformation is then carried out by activation primitive layer, then
Pondization operation is carried out by pond layer, the classification that the polar orientation classification of target electronic components is then obtained by full articulamentum is special
Sign, wherein convolutional neural networks include convolution module and full articulamentum, and convolution module includes sequentially connected convolutional layer, activation
Function layer and pond layer.
Convolutional neural networks after training in one of the embodiments, are obtained by following steps:
Establish the image pattern collection of the various polar orientations of all kinds of electronic components；
Using convolution module interconnected in convolutional neural networks and full articulamentum to each image sample of image pattern collection
Notebook data carries out forward calculation respectively, the characteristic of division of the various polar orientation classifications of all kinds of electronic components is obtained, according to each point
Category feature training convolutional neural networks make convolutional neural networks identify the various polar orientations of all kinds of electronic components.
The step of establishing the sample set of the various polar orientations of all kinds of electronic components in one of the embodiments, include with
Lower step:
Pcb board card graphic and PCB Prototype drawing are obtained, and is reference with PCB Prototype drawing, position is carried out to pcb board card graphic
Registration；
All kinds of electronic component images after interception position registration on pcb board card graphic, with each electron-like in PCB Prototype drawing
Element matches electronic component corresponding in all kinds of electronic component images, obtains corresponding electronics in all kinds of electronic component images
The exact position of element is adjusted corresponding electronic element according to the exact position of corresponding electronic element, makes corresponding electronics member
Part is located at the center of all kinds of electronic component images, obtains the image pattern collection of the various polar orientations of all kinds of electronic components.
Convolution module is 5 in one of the embodiments, wherein the convolution kernel number of 5 convolutional layers is respectively 24,
The convolution kernel size of 64,96,96 and 64,5 convolutional layers is respectively 7 × 7,5 × 5,3 × 3,3 × 3,3 × 3,5 convolutional layers
Step-length is 1；Full articulamentum is 2, wherein the Hidden nodes of 2 full articulamentums are respectively 512 and 4.
In one of the embodiments, after the step of obtaining includes the image of target electronic components, including following step
It is rapid:
The Prototype drawing for obtaining target electronic components, with the Prototype drawing of target electronic components to the target electronic components in image
It is matched, obtains the exact position of the target electronic components in image, according to exact position to the target electronic member in image
Part is adjusted, and so that the target electronic components in image is located at the center of image, image adjusted is for the convolution mind after training
Make forward calculation through network.
A kind of mask method of electronic component polar orientation, comprising the following steps:
According to the method for quickly identifying of above-mentioned electronic component polar orientation determine target electronic components polar orientation,
The polar orientation information of label target electronic component in board-like file, board-like file are used to save each attribute letter of electronic component
Breath.
The mask method of above-mentioned electronic component polar orientation can be applied to the board-like production of AOI, automatically, accurately to PCB
Element on board, which carries out polar orientation mark, improves the efficiency of board-like production to promote the automatization level of board-like production
With accuracy.
A kind of system for rapidly identifying of electronic component polar orientation, including with lower unit:
Acquiring unit obtains the image comprising target electronic components；
Computing unit, before being made using the convolutional neural networks after training to the image data comprising target electronic components
To calculating, the characteristic of division of the polar orientation classification of target electronic components is obtained, target electronic components are obtained according to characteristic of division
Belong to the probability distribution of the various polar orientation classifications of all kinds of electronic components；
Selection unit, for choosing polar orientation class of the polar orientation classification as target electronic components of maximum probability
Not.
It is to obtain the image including target electronic components according to above-mentioned system for rapidly identifying, the volume after recycling training
Product neural network makees forward calculation to it, obtains the characteristic of division of the polar orientation classification of target electronic components, then obtain target
Electronic component belongs to the probability distribution of the various polar orientation classifications of all kinds of electronic components, chooses the polarity side of wherein maximum probability
Polar orientation classification to classification as target electronic components.Convolutional neural networks have been used in this scheme, have passed through convolutional Neural
Network can automatically and accurately identify the polar orientation of electronic component, and not for specific electronic component structure, be applicable in
In the polar electronic component of various bands, the polar orientation identification of the electronic component across classification is realized, applicability is wider.
Computing unit carries out convolution algorithm to image data by convolutional layer in one of the embodiments, then passes through
Activation primitive layer carries out nonlinear transformation, then carries out pondization operation by pond layer, then obtains target electricity by full articulamentum
The characteristic of division of the polar orientation classification of subcomponent, wherein convolutional neural networks include convolution module and full articulamentum, convolution mould
Block includes sequentially connected convolutional layer, activation primitive layer and pond layer.
A kind of labeling system of electronic component polar orientation, including mark unit and above-mentioned electronic component polar orientation it is fast
Fast identifying system, wherein polar orientation information of the mark unit for the label target electronic component in board-like file, board-like text
Part is used to save the various attribute informations of electronic component.
The labeling system of above-mentioned electronic component polar orientation can be applied to the board-like production of AOI, automatically, accurately to PCB
Element on board, which carries out polar orientation mark, improves the efficiency of board-like production to promote the automatization level of board-like production
With accuracy.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for quickly identifying of electronic component polar orientation in one embodiment；
Fig. 2 is the structural schematic diagram of convolution module in one embodiment；
Fig. 3 is the structural schematic diagram of convolutional neural networks in one embodiment；
Fig. 4 is the structural schematic diagram of the system for rapidly identifying of electronic component polar orientation in one embodiment；
Fig. 5 is the structural schematic diagram of the system for rapidly identifying of electronic component polar orientation in one embodiment；
Fig. 6 is the partial structure diagram of the system for rapidly identifying of electronic component polar orientation in one embodiment；
Fig. 7 is the structural schematic diagram of the system for rapidly identifying of electronic component polar orientation in one embodiment；
Fig. 8 is the structural schematic diagram of the system for rapidly identifying of electronic component polar orientation in one embodiment；
Fig. 9 is the structural schematic diagram of the labeling system of electronic component polar orientation in one embodiment.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments, to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
And the scope of protection of the present invention is not limited.
It is shown in Figure 1, it is the method for quickly identifying embodiment of electronic component polar orientation of the invention.As shown in Figure 1,
The method for quickly identifying of electronic component polar orientation in the embodiment the following steps are included:
Step S101: the image comprising target electronic components is obtained；
In this step, the image comprising electronic component, which can be, needs to identify the polar pcb board card graphic of electronic component,
It is also possible to other and needs to identify the polar image of electronic component.
Step S102: to meter before being made using the convolutional neural networks after training to the image data comprising target electronic components
It calculates, obtains the characteristic of division of the polar orientation classification of target electronic components, target electronic components are obtained according to characteristic of division and are belonged to
The probability distribution of the various polar orientation classifications of all kinds of electronic components；
In this step, convolution module interconnected in the convolutional neural networks after training and full connection are mainly utilized
Layer makees forward calculation to the image data comprising target electronic components, and the image data comprising target electronic components passes through convolution mould
The characteristic of division of the polar orientation classification of target electronic components can be obtained after block and full articulamentum；Various polar orientation classifications are
The various polar orientations for referring to various electronic components, include a variety of electronic components, and the probability distribution of acquisition can be adapted for various
Electronic component；Convolutional neural networks after training can operate image, to the polarity side of target electronic components therein
To being identified, probability distribution is obtained.
Step S103: polar orientation classification of the polar orientation classification as target electronic components of maximum probability is chosen.
Above-mentioned steps S101, S102 and S103 are the processes that on-line testing is carried out using the convolutional neural networks after training.
The method for quickly identifying of the electronic component polar orientation of present embodiment is obtained including target electronic components
Image, the convolutional neural networks after recycling training make forward calculation to it, obtain the polar orientation classification of target electronic components
Characteristic of division, then obtain the probability distribution that target electronic components belong to the various polar orientation classifications of all kinds of electronic components, select
Take the polar orientation classification of wherein maximum probability as the polar orientation classification of target electronic components.Convolution has been used in this scheme
Neural network can automatically and accurately identify the polar orientation of electronic component by convolutional neural networks, and not for spy
Fixed electronic component structure is suitable for the polar electronic component of various bands, realizes that the polar orientation of the electronic component across classification is known
Not, applicability is wider.
In one of the embodiments, using the convolutional neural networks after training to the picture number comprising target electronic components
According to forward calculation is made, the step of obtaining the characteristic of division of the polar orientation classification of target electronic components the following steps are included:
Convolution algorithm is carried out to image data by convolutional layer, nonlinear transformation is then carried out by activation primitive layer, then
Pondization operation is carried out by pond layer, the classification that the polar orientation classification of target electronic components is then obtained by full articulamentum is special
Sign, wherein convolutional neural networks include convolution module and full articulamentum, and convolution module includes sequentially connected convolutional layer, activation
Function layer and pond layer.
In the present embodiment, by convolutional layer sequentially connected in convolution module, activation primitive layer and pond layer and with
The full articulamentum of convolution module connection can obtain preferable characteristic of division.
Preferably, activation primitive layer can be ReLU function layer.
As shown in Fig. 2, convolution module includes sequentially connected convolutional layer, ReLU function layer and pond layer, pass through convolutional layer
Convolution algorithm is carried out to the image data of target electronic components, nonlinear transformation is then carried out by ReLU function layer, then pass through
Pond layer carries out pondization operation, and the characteristic of division of the polar orientation classification of target electronic components is then obtained by full articulamentum.
ReLU function in the convolution module of convolutional neural networks is a kind of activation primitive, can be used for nonlinear transformation；
Pond layer can carry out aggregate statistics to the feature of the different location of image pattern for carrying out pondization operation, and will not make
There is the problem of over-fitting in last result.
The ability to express of convolutional neural networks is very strong, can effectively solve the problems, such as that the precision in more classification tasks is not high,
Even if also can reach very high accuracy rate in the electronic component polar orientation identification mission across classification.
Convolutional neural networks after training in one of the embodiments, are obtained by following steps:
Establish the image pattern collection of the various polar orientations of all kinds of electronic components；
Using convolution module interconnected in convolutional neural networks and full articulamentum to each image sample of image pattern collection
Notebook data carries out forward calculation respectively, the characteristic of division of the various polar orientation classifications of all kinds of electronic components is obtained, according to each point
Category feature training convolutional neural networks make convolutional neural networks identify the various polar orientations of all kinds of electronic components.
It can reinforce the identification of convolutional neural networks using external data collection in training convolutional neural networks, but to hardware
Computing capability is more demanding, it usually needs middle and high end GPU is just able to satisfy actual calculating speed demand, therefore improve hardware at
This.Technical solution of the present invention need not use external data collection to be trained convolutional neural networks, can match in the hardware of low cost
Set down, precisely rapidly to electronic component carry out auto polarity walking direction, be suitable for all kinds of electronic components (such as capacitor, socket,
Resistance etc.).Program main feature is not dependent on special hardware mechanism, and cost is relatively low, greatly reduces algorithm to hardware meter
The requirement of calculation ability and memory space solves the problems, such as high precision technology scheme higher cost.
Currently, the recognition methods of existing electronic component polar orientation often utilize be electronic component colouring information etc.
Low layer pictures feature, this conventional method are all too simple for many application scenarios, and robustness is lower, the scope of application also by
Limit, effect are poor；And convolutional neural networks study multilayer feature representation (including low layer, middle layer, high level are used in this programme
Characteristics of image, rather than just low layer pictures feature), and comprehensive such multilayer feature goes identification electronic component polar orientation,
To greatly promote accuracy of identification, the scope of application is also extended.
It, can be according to the image pattern collection of the various polar orientations of all kinds of electronic components to convolutional Neural net by this step
Network carry out off-line training, so that convolutional neural networks is identified the polar orientation information of electronic component in image, so as to
It is identified in on-line testing using polar orientation of the convolutional neural networks after the training to target electronic components.
The step of establishing the sample set of the various polar orientations of all kinds of electronic components in one of the embodiments, include with
Lower step:
Pcb board card graphic and PCB Prototype drawing are obtained, and is reference with PCB Prototype drawing, position is carried out to pcb board card graphic
Registration；
All kinds of electronic component images after interception position registration on pcb board card graphic, with each electron-like in PCB Prototype drawing
Element matches electronic component corresponding in all kinds of electronic component images, obtains corresponding electronics in all kinds of electronic component images
The exact position of element is adjusted corresponding electronic element according to the exact position of corresponding electronic element, makes corresponding electronics member
Part is located at the center of all kinds of electronic component images, obtains the image pattern collection of the various polar orientations of all kinds of electronic components.
The image pattern collection obtained through the above way is directly intercepted from pcb board card graphic, is all practical electricity
The image of subcomponent, with more directiveness.
Preferably, camera, and the pcb board card figure of batch capture different model can be set up on pcb board card production line
Picture, and avoid repeating to shoot a certain pcb board card with board tracking technique, the pcb board card of every kind in this way model includes multiple
Image pattern, each image pattern correspond to certain pcb board card of a certain model；It needs to save phase when acquiring pcb board card graphic
The PCB Prototype drawing answered；
Pcb board card graphic position in multiple images sample may deviate, need using the PCB Prototype drawing of respective model as
With reference to the progress image registration of each image pattern, and using the positions of electronic parts information of the model pcb board card (from plate
Formula file or artificial mark) electronic component picture is intercepted automatically, and board-like file or people (are come from according to electronic component classification information
Work mark) automatic marking is carried out to electronic component picture；
The exact position of electronic component in image pattern is further acquired with the electronic component matching in PCB Prototype drawing, it is right
Electronic component image carries out alignment adjustment, to guarantee that electronic component is located at image center location, obtains the various poles of electronic component
The image pattern collection in property direction.
In one of the embodiments, as shown in figure 3, convolution module is 5, wherein the convolution kernel number of 5 convolutional layers
Respectively 24,64,96,96 and 64, the convolution kernel size of 5 convolutional layers is respectively 7 × 7,5 × 5,3 × 3,3 × 3,3 × 3,5
The step-length of convolutional layer is 1；Full articulamentum is 2, wherein the Hidden nodes of 2 full articulamentums are respectively 512 and 4.
The quick identification of electronic component polar orientation can be preferably realized with the convolutional neural networks in the present embodiment.
In one of the embodiments, after the step of obtaining includes the image of target electronic components, including following step
It is rapid:
The Prototype drawing for obtaining target electronic components, with the Prototype drawing of target electronic components to the target electronic components in image
It is matched, obtains the exact position of the target electronic components in image, according to exact position to the target electronic member in image
Part is adjusted, and target electronic components is made to be located at the center of image, and image adjusted is made for the convolutional neural networks after training
Forward calculation.
The target electronic components in image are adjusted with the Prototype drawing of target electronic components, so that the convolution after training
Neural network is more easier to identify target electronic components.
Choosing polar orientation of the polar orientation classification of maximum probability as electronic component in one of the embodiments,
It is further comprising the steps of after the step of classification:
If the polar orientation classification for the maximum probability chosen is different from preset electronic component polar orientation classification, provide
False alarm information.
Above-mentioned steps are the detection process to polar orientation, can correctly be judged whether electronic component is installed.
The image comprising target electronic components that on-line testing obtains in one of the embodiments, is pcb board card graphic,
With corresponding PCB template come compare matching, obtain the exact position of target electronic components, to the target electronic components in image into
Row alignment adjustment, to guarantee that target electronic components are located at the center of image.
In the present embodiment, corresponding PCB template is used when control matching, which can be image sample
Concentrate the corresponding PCB template of image, be also possible to from the image sample PCB template that concentrate the corresponding PCB template of image different, such as
Fruit is for the former, PCB template corresponding with image sample concentration image has been saved in off-line training, can directly acquire makes
With.If it is the latter, in addition to also needing to obtain corresponding PCB template from the external world when obtaining pcb board card graphic when on-line testing, with
Control matching after realization.Polarity can be carried out to all electronic components in the pcb board card graphic of acquisition by implementing this programme
The identification in direction.
It the step of establishing the sample set of the various polar orientations of all kinds of electronic components in one of the embodiments, and is obtaining
Take after the step of including the image of target electronic components the following steps are included:
The size of image adjusted is normalized.
In the sample set for the various polar orientations for establishing all kinds of electronic components the step of, to the size of image adjusted
It is normalized, convolutional neural networks can be made to the multilayer feature of image pattern convenient for the training to convolutional neural networks
Study it is more accurate.
After the step of obtaining includes the image of target electronic components, the size of image adjusted is normalized
Processing can handle convenient for forward calculation of the convolutional neural networks to the image of target electronic components, accelerate the identification of electronic component
Process.
The method for quickly identifying of electronic component polar orientation of the invention, has used convolutional neural networks, can be with automatic precision
The polar orientation of electronic component is identified quasi-ly, and not for specific electronic component structure, it is polar to be suitable for various bands
Electronic component, as long as the image pattern collection of training convolutional neural networks covers the electronic component of multiple types, so that it may real
Now the polar orientation identification of the electronic component across classification, applicability are wider.Running at the GPU of current mainstream can also be to routine
Pcb board card (comprising about 20-30 with polar element) carry out real-time judge, being averaged for an element is judged on GPU
The time is calculated even within 1 millisecond, therefore meets various real-time judge scenes；Either off-line training step or online survey
The examination stage reduces manual intervention as far as possible, and independent of special hardware mechanism, cost is relatively low；Powerful meter is not needed
Resource is calculated, can even be run in some scenarios in cheap embedded platform.
The present invention also provides a kind of mask methods of electronic component polar orientation, comprising the following steps:
According to the method for quickly identifying of above-mentioned electronic component polar orientation determine target electronic components polar orientation,
The polar orientation information of label target electronic component in board-like file, board-like file are used to save each attribute letter of electronic component
Breath.
The mask method of above-mentioned electronic component polar orientation can be applied to the board-like production of AOI, automatically, accurately to PCB
Element on board, which carries out polar orientation mark, improves the efficiency of board-like production to promote the automatization level of board-like production
With accuracy.
According to the method for quickly identifying of above-mentioned electronic component polar orientation, the present invention also provides a kind of electronic component polarity sides
To system for rapidly identifying, the embodiment of the system for rapidly identifying of electronic component polar orientation of the invention is carried out below detailed
Explanation.
It is shown in Figure 4, it is the embodiment of the system for rapidly identifying of electronic component polar orientation of the invention.The embodiment
In electronic component polar orientation system for rapidly identifying, including in Fig. 4 acquiring unit 210, computing unit 220 and choose single
Member 230；
Acquiring unit 210, for obtaining the image comprising target electronic components；
Computing unit 220, for utilizing the convolutional neural networks after training to the image data comprising target electronic components
Make forward calculation, obtain the characteristic of division of the polar orientation classification of target electronic components, target electronic is obtained according to characteristic of division
Element belongs to the probability distribution of the various polar orientation classifications of all kinds of electronic components；
Selection unit 230, for choosing polar orientation of the polar orientation classification of maximum probability as target electronic components
Classification.
It is to obtain the image including target electronic components according to above-mentioned system for rapidly identifying, the volume after recycling training
Product neural network makees forward calculation to it, obtains the characteristic of division of the polar orientation classification of target electronic components, then obtain target
Electronic component belongs to the probability distribution of the various polar orientation classifications of all kinds of electronic components, chooses the polarity side of wherein maximum probability
Polar orientation classification to classification as target electronic components.Convolutional neural networks have been used in this scheme, have passed through convolutional Neural
Network can automatically and accurately identify the polar orientation of electronic component, and not for specific electronic component structure, be applicable in
In the polar electronic component of various bands, the polar orientation identification of the electronic component across classification is realized, applicability is wider.
Computing unit 220 carries out convolution algorithm to image data by convolutional layer in one of the embodiments, then leads to
It crosses activation primitive layer and carries out nonlinear transformation, then pondization operation is carried out by pond layer, target is then obtained by full articulamentum
The characteristic of division of the polar orientation classification of electronic component, wherein convolutional neural networks include convolution module and full articulamentum, convolution
Module includes sequentially connected convolutional layer, activation primitive layer and pond layer.
In one of the embodiments, as shown in figure 5, the system for rapidly identifying of electronic component polar orientation further includes establishing
Unit 240 and training unit 250；
Establish the image pattern collection of various polar orientations of the unit 240 for establishing all kinds of electronic components；
Training unit 250 is used for using convolution module interconnected in convolutional neural networks and full articulamentum to image sample
Each image sample data of this collection carries out forward calculation respectively, obtains the classification of the various polar orientation classifications of all kinds of electronic components
Feature makes convolutional neural networks identify the various polarity of all kinds of electronic components according to each characteristic of division training convolutional neural networks
Direction.
In one of the embodiments, as shown in fig. 6, establishing unit 240 includes registration unit 241, intercepting process unit
242；
Registration unit 241 is reference for obtaining pcb board card graphic and PCB Prototype drawing, and with PCB Prototype drawing, to pcb board
Card graphic carries out position registration；
Electronic component image of the intercepting process unit 242 on pcb board card graphic after interception position registration, with PCB mould
All kinds of electronic components in plate figure match electronic component corresponding in all kinds of electronic component images, obtain each electron-like member
The exact position of corresponding electronic element in part image, adjusts corresponding electronic element according to the exact position of corresponding electronic element
It is whole, so that corresponding electronic element is located at the center of all kinds of electronic component images, obtains the various polar orientations of all kinds of electronic components
Image pattern collection.
Convolution module is 5 in one of the embodiments, wherein the convolution kernel number of 5 convolutional layers is respectively 24,
The convolution kernel size of 64,96,96 and 64,5 convolutional layers is respectively 7 × 7,5 × 5,3 × 3,3 × 3,3 × 3,5 convolutional layers
Step-length is 1；Full articulamentum is 2, wherein the Hidden nodes of 2 full articulamentums are respectively 512 and 4.
In one of the embodiments, as shown in fig. 7, the system for rapidly identifying of electronic component polar orientation further includes pre- place
Manage unit 260；
Pretreatment unit 260 is used for after acquiring unit obtains the image comprising target electronic components, with target electronic member
The Prototype drawing of part matches the target electronic components in image, obtains the exact position of the target electronic components in image,
The target electronic components in image are adjusted according to exact position, are located at the target electronic components in image in image
The heart, image adjusted make forward calculation for the convolutional neural networks after training.
In one of the embodiments, as shown in figure 8, the system for rapidly identifying of electronic component polar orientation further includes alarm
Unit 270, if the polar orientation classification of the maximum probability for selection is different from preset electronic component polar orientation classification,
Provide false alarm information.
The system for rapidly identifying of electronic component polar orientation of the invention is fast with electronic component polar orientation of the invention
Fast recognition methods corresponds, in the technical characteristic that the embodiment of the method for quickly identifying of above-mentioned electronic component polar orientation illustrates
And its advantages are suitable for the embodiment of the system for rapidly identifying of electronic component polar orientation.
The present invention also provides a kind of labeling systems of electronic component polar orientation, as shown in figure 9, including mark unit 300
And the identifying system of above-mentioned electronic component polar orientation, mark unit 300 are used for according to above-mentioned electronic component polar orientation
System for rapidly identifying determine target electronic components polar orientation, in board-like file mark electronic component polar orientation
Information, board-like file are used to save the various attribute informations of electronic component.
The labeling system of above-mentioned electronic component polar orientation can be applied to the board-like production of AOI, automatically, accurately to PCB
Element on board, which carries out polar orientation mark, improves the efficiency of board-like production to promote the automatization level of board-like production
With accuracy.
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (10)
1. a kind of method for quickly identifying of electronic component polar orientation, which comprises the following steps:
Obtain the image comprising target electronic components；
Forward calculation is made to the image data comprising target electronic components using the convolutional neural networks after training, obtains the mesh
The characteristic of division for marking the polar orientation classification of electronic component obtains the target electronic components according to the characteristic of division and belongs to respectively
The probability distribution of the various polar orientation classifications of electron-like element；Wherein, the various polar orientation classifications include a variety of electronics
The polar orientation classification of element；The forward calculation includes identifying to the polar orientation of the target electronic components；
Choose polar orientation classification of the polar orientation classification of maximum probability as the target electronic components.
2. the method for quickly identifying of electronic component polar orientation according to claim 1, which is characterized in that described to utilize instruction
Convolutional neural networks after white silk make forward calculation to the image data comprising target electronic components, obtain the target electronic components
Polar orientation classification characteristic of division the step of the following steps are included:
Convolution algorithm is carried out to described image data by convolutional layer, nonlinear transformation is then carried out by activation primitive layer, then
Pondization operation is carried out by pond layer, point of the polar orientation classification of the target electronic components is then obtained by full articulamentum
Category feature, wherein the convolutional neural networks include convolution module and full articulamentum, and the convolution module includes sequentially connected
The convolutional layer, the activation primitive layer and the pond layer.
3. the method for quickly identifying of electronic component polar orientation according to claim 1, which is characterized in that after the training
Convolutional neural networks pass through following steps obtain:
Establish the image pattern collection of the various polar orientations of all kinds of electronic components；
Using convolution module interconnected in convolutional neural networks and full articulamentum to each image sample of described image sample set
Notebook data carries out forward calculation respectively, the characteristic of division of the various polar orientation classifications of all kinds of electronic components is obtained, according to each institute
The characteristic of division training convolutional neural networks are stated, the convolutional neural networks is made to identify the various poles of all kinds of electronic components
Property direction.
4. the method for quickly identifying of electronic component polar orientation according to claim 3, which is characterized in that establish all kinds of electricity
The step of sample set of the various polar orientations of subcomponent the following steps are included:
Pcb board card graphic and PCB Prototype drawing are obtained, and is reference with the PCB Prototype drawing, the pcb board card graphic is carried out
Position registration；
All kinds of electronic component images after interception position registration on pcb board card graphic, with each electron-like in the PCB Prototype drawing
Element matches electronic component corresponding in all kinds of electronic component images, obtains in all kinds of electronic component images
The corresponding electronic element is adjusted according to the exact position of the corresponding electronic element in the exact position of corresponding electronic element
It is whole, so that the corresponding electronic element is located at the center of all kinds of electronic component images, obtains the various poles of all kinds of electronic components
The image pattern collection in property direction.
5. the method for quickly identifying of electronic component polar orientation according to claim 2, which is characterized in that the convolution mould
Block is 5, wherein the convolution kernel number of 5 convolutional layers is respectively the volume of 24,64,96,96 and 64,5 convolutional layers
Product core size is respectively that the step-length of 7 × 7,5 × 5,3 × 3,3 × 3,3 × 3,5 convolutional layers is 1；The full articulamentum
It is 2, wherein the Hidden nodes of 2 full articulamentums are respectively 512 and 4.
6. the method for quickly identifying of electronic component polar orientation as claimed in any of claims 1 to 5, feature exist
In after the step of acquisition includes the image of target electronic components, comprising the following steps:
The Prototype drawing of the target electronic components is obtained, with the Prototype drawing of the target electronic components to the target in described image
Electronic component is matched, and the exact position of the target electronic components in described image is obtained, according to the exact position to institute
The target electronic components stated in image are adjusted, and the target electronic components in described image is made to be located at the center of described image,
Image adjusted supplies the convolutional neural networks after the training to make forward calculation.
7. a kind of mask method of electronic component polar orientation, which comprises the following steps:
The target that the method for quickly identifying of electronic component polar orientation as claimed in any of claims 1 to 6 determines
The polar orientation of electronic component marks the polar orientation information of the target electronic components, the board-like text in board-like file
Part is used to save the various attribute informations of electronic component.
8. a kind of system for rapidly identifying of electronic component polar orientation, which is characterized in that including with lower unit:
Acquiring unit obtains the image comprising target electronic components；
Computing unit, before being made using the convolutional neural networks after training to the image data comprising target electronic components based on
It calculates, obtains the characteristic of division of the polar orientation classification of the target electronic components, the target is obtained according to the characteristic of division
Electronic component belongs to the probability distribution of the various polar orientation classifications of all kinds of electronic components；Wherein, the various polar orientation classes
Not Bao Kuo a variety of electronic components polar orientation classification；The forward calculation includes the polar orientation to the target electronic components
It is identified；
Selection unit, for choosing polar orientation class of the polar orientation classification as the target electronic components of maximum probability
Not.
9. the system for rapidly identifying of electronic component polar orientation according to claim 8, which is characterized in that
The computing unit carries out convolution algorithm to described image data by convolutional layer, is then carried out by activation primitive layer non-
Linear transformation, then pondization operation is carried out by pond layer, the polarity of the target electronic components is then obtained by full articulamentum
The characteristic of division of direction classification, wherein the convolutional neural networks include convolution module and full articulamentum, the convolution module packet
Include the sequentially connected convolutional layer, the activation primitive layer and the pond layer.
10. a kind of labeling system of electronic component polar orientation, which is characterized in that including mark unit and such as claim 8 or 9
The system for rapidly identifying of the electronic component polar orientation, wherein the mark unit in board-like file for marking institute
The polar orientation information of target electronic components is stated, the board-like file is used to save the various attribute informations of electronic component.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510823564.XA CN105469400B (en) | 2015-11-23 | 2015-11-23 | Method and system for quickly identifying and marking polarity direction of electronic element |
| PCT/CN2016/098228 WO2017088553A1 (en) | 2015-11-23 | 2016-09-06 | Method and system for rapidly identifying and marking electronic component polarity direction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510823564.XA CN105469400B (en) | 2015-11-23 | 2015-11-23 | Method and system for quickly identifying and marking polarity direction of electronic element |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105469400A CN105469400A (en) | 2016-04-06 |
| CN105469400B true CN105469400B (en) | 2019-02-26 |
Family
ID=55607059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510823564.XA Active CN105469400B (en) | 2015-11-23 | 2015-11-23 | Method and system for quickly identifying and marking polarity direction of electronic element |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN105469400B (en) |
| WO (1) | WO2017088553A1 (en) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105469400B (en) * | 2015-11-23 | 2019-02-26 | 广州视源电子科技股份有限公司 | Method and system for quickly identifying and marking polarity direction of electronic element |
| CN105426917A (en) * | 2015-11-23 | 2016-03-23 | 广州视源电子科技股份有限公司 | Element classification method and device |
| CN105821538B (en) * | 2016-04-20 | 2018-07-17 | 广州视源电子科技股份有限公司 | Spun yarn breakage detection method and system |
| CN107305636A (en) * | 2016-04-22 | 2017-10-31 | 株式会社日立制作所 | Target identification method, Target Identification Unit, terminal device and target identification system |
| CN106530284A (en) * | 2016-10-21 | 2017-03-22 | 广州视源电子科技股份有限公司 | Welding spot type detection and device based on image recognition |
| CN109429473A (en) * | 2017-08-28 | 2019-03-05 | 株洲中车时代电气股份有限公司 | Automatic check method and device with polarity electronic component in circuit board |
| CN107886131A (en) * | 2017-11-24 | 2018-04-06 | 佛山科学技术学院 | One kind is based on convolutional neural networks detection circuit board element polarity method and apparatus |
| CN108446659A (en) * | 2018-03-28 | 2018-08-24 | 百度在线网络技术（北京）有限公司 | Method and apparatus for detecting facial image |
| CN109146919B (en) * | 2018-06-21 | 2020-08-04 | 全球能源互联网研究院有限公司 | Tracking and aiming system and method combining image recognition and laser guidance |
| CN108830850B (en) * | 2018-06-28 | 2020-10-23 | 信利（惠州）智能显示有限公司 | Automatic optical detection picture analysis method and equipment |
| CN110070536B (en) * | 2019-04-24 | 2022-08-30 | 南京邮电大学 | Deep learning-based PCB component detection method |
| CN111310806B (en) * | 2020-01-22 | 2024-03-15 | 北京迈格威科技有限公司 | Classification network, image processing method, device, system and storage medium |
| CN111640088B (en) * | 2020-04-22 | 2023-12-01 | 深圳拓邦股份有限公司 | Electronic element polarity detection method and system based on deep learning and electronic device |
| CN112378908A (en) * | 2020-11-09 | 2021-02-19 | 日立楼宇技术（广州）有限公司 | Capacitor polarity installation detection device and detection method |
| CN114024597A (en) * | 2021-11-03 | 2022-02-08 | 浙江大学湖州研究院 | Laser communication coarse aiming device based on neural network pattern recognition |
| CN115479891A (en) * | 2022-08-12 | 2022-12-16 | 深圳市共进电子股份有限公司 | Automatic detection system and method for circuit board mounted components based on image recognition |
| CN116168040B (en) * | 2023-04-26 | 2023-07-07 | 四川元智谷科技有限公司 | Component direction detection method and device, electronic equipment and readable storage medium |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100668303B1 (en) * | 2004-08-04 | 2007-01-12 | 삼성전자주식회사 | Method for detecting face based on skin color and pattern matching |
| CN103544705B (en) * | 2013-10-25 | 2016-03-02 | 华南理工大学 | A kind of image quality test method based on degree of depth convolutional neural networks |
| CN103927534B (en) * | 2014-04-26 | 2017-12-26 | 无锡信捷电气股份有限公司 | A kind of online visible detection method of coding character based on convolutional neural networks |
| CN103984959B (en) * | 2014-05-26 | 2017-07-21 | 中国科学院自动化研究所 | A kind of image classification method based on data and task-driven |
| CN104459421B (en) * | 2014-12-30 | 2017-07-11 | 广州视源电子科技股份有限公司 | Diode polarity detection method and system |
| CN104850845B (en) * | 2015-05-30 | 2017-12-26 | 大连理工大学 | A kind of traffic sign recognition method based on asymmetric convolutional neural networks |
| CN105469400B (en) * | 2015-11-23 | 2019-02-26 | 广州视源电子科技股份有限公司 | Method and system for quickly identifying and marking polarity direction of electronic element |
-
2015
- 2015-11-23 CN CN201510823564.XA patent/CN105469400B/en active Active
-
2016
- 2016-09-06 WO PCT/CN2016/098228 patent/WO2017088553A1/en active Application Filing
Also Published As
| Publication number | Publication date |
|---|---|
| CN105469400A (en) | 2016-04-06 |
| WO2017088553A1 (en) | 2017-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105469400B (en) | Method and system for quickly identifying and marking polarity direction of electronic element | |
| CN105513046B (en) | Electronic component polarity identification method and system, and labeling method and system | |
| WO2020098250A1 (en) | Character recognition method, server, and computer readable storage medium | |
| CN108460362B (en) | System and method for detecting human body part | |
| CN110070536A (en) | A kind of pcb board component detection method based on deep learning | |
| WO2017088537A1 (en) | Component classification method and apparatus | |
| CN109671103A (en) | Method for tracking target and device | |
| CN105184778A (en) | Detection method and device | |
| CN109100370A (en) | A kind of pcb board defect inspection method based on sciagraphy and connected domain analysis | |
| CN105931269A (en) | Tracking method for target in video and tracking device thereof | |
| CN107886131A (en) | One kind is based on convolutional neural networks detection circuit board element polarity method and apparatus | |
| CN106408510B (en) | A kind of method and system for the U.S. face masking-out obtaining facial image | |
| CN109325456A (en) | Target identification method, device, target identification equipment and storage medium | |
| CN113408662B (en) | Image recognition and training method and device for image recognition model | |
| CN111382743B (en) | License plate character recognition method based on data enhancement and data generation | |
| CN116091470A (en) | Intelligent industrial control method and device based on 5G technology, electronic equipment and storage medium | |
| CN110865077A (en) | Visual inspection system for appearance defects in RFID antenna production | |
| CN109284735A (en) | Mouse feelings monitoring method, device, processor and storage medium | |
| CN117351499B (en) | Split-combined indication state identification method, system, computer equipment and medium | |
| CN110443328A (en) | Fake method, device, equipment and medium are tested in identification based on LED antifalsification label | |
| CN113824874A (en) | Auxiliary shooting method and device, electronic equipment and storage medium | |
| CN116168040B (en) | Component direction detection method and device, electronic equipment and readable storage medium | |
| CN112733723A (en) | Vehicle type identification method and terminal | |
| CN106934339A (en) | A kind of target following, the extracting method of tracking target distinguishing feature and device | |
| CN113918769B (en) | Method, device, equipment and storage medium for marking key actions in video |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |