CN113674207B - Automatic PCB component positioning method based on graph convolution neural network - Google Patents

Automatic PCB component positioning method based on graph convolution neural network Download PDF

Info

Publication number
CN113674207B
CN113674207B CN202110825126.2A CN202110825126A CN113674207B CN 113674207 B CN113674207 B CN 113674207B CN 202110825126 A CN202110825126 A CN 202110825126A CN 113674207 B CN113674207 B CN 113674207B
Authority
CN
China
Prior art keywords
neural network
pcb
node
convolution neural
component
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
Application number
CN202110825126.2A
Other languages
Chinese (zh)
Other versions
CN113674207A (en
Inventor
郑亚莉
廖文杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202110825126.2A priority Critical patent/CN113674207B/en
Publication of CN113674207A publication Critical patent/CN113674207A/en
Application granted granted Critical
Publication of CN113674207B publication Critical patent/CN113674207B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • 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]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a PCB component automatic positioning method based on a graph convolution neural network, which comprises the steps of collecting a plurality of PCB images, marking the corresponding relation between each PCB image and the position of a component in the image in a manual calibration mode, recording the central coordinate of each component, then sequentially inputting the preprocessed PCB images into the graph convolution neural network, updating the characteristics of each node based on a message transfer mechanism, further completing the iterative training of the graph convolution neural network, and finally inputting the PCB image to be positioned into the graph convolution neural network, wherein the graph convolution neural network directly outputs the central coordinate of each component in the PCB image, thereby realizing automatic positioning.

Description

Automatic PCB component positioning method based on graph convolution neural network
Technical Field
The invention belongs to the technical field of PCB mounting and detection, and particularly relates to a PCB component automatic positioning method based on a graph convolution neural network.
Background
In the SMT manufacturing process, the components need to be mounted according to a given coordinate file, or whether the leakage or the wrong material exists is determined, wherein the PCB component positioning is the first link. In the traditional mounting and detecting process, the mounting and detecting method is mainly realized according to the manual positioning and assembly of components on a drawing. With the development of image processing technology and deep artificial neural networks, the AOI system has gradually realized automatic positioning that relies on manual assistance of PCB components, such as a positioning algorithm based on the back features of IC components and a PCB component detection method based on deep learning. And based on the back characteristic of the IC type component, the positioning algorithm detects the angular points of the pins according to the pin information and the back information, and calculates the minimum external rectangular frame characteristic of the component by using the distance between the angular points to realize the positioning of the component. The deep learning method realizes the component positioning by adopting a deep neural network to realize the PCB position number identification recognition. The published patent "a circuit board component positioning method and device" (CN 202011298850.6) utilizes the pad information of the component and the silk-screen character information in the image processing method to determine the information of the position and direction of the chip of the corresponding component. The utility model provides a convolutional neural network based element positioning identification method and process "(CN 111429431A), utilize convolutional neural network to realize the discernment of silk-screen printing sign. The two methods mainly aim at positioning the chip information of the patch, depend on the characteristics of the identification component and the character identification of the position number of the PCB silk screen, and cannot deal with the situation that the position number of the silk screen character with the missing mark and the silk screen character after the patch are shielded.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a method for automatically positioning a PCB (printed circuit board) component based on a graph convolution neural network, which realizes the automatic positioning of pixel coordinates of a PCB light panel image and a mounted image pixel component through the graph convolution neural network.
In order to achieve the above object, the present invention provides a method for automatically positioning a PCB device based on a graph convolution neural network, which is characterized by comprising the following steps:
(1) PCB image acquisition and pretreatment
(1.1) collecting a plurality of PCB images, marking the corresponding relation between each PCB image and the position of the element in the image in a manual calibration mode, and recording the central coordinate F of each element l L =1,2, …, N represents the number of components in the PCB image;
(1.2) zooming each PCB image to the same resolution, and then carrying out normalization processing on each zoomed PCB image;
(2) Building graph convolution neural network
Adopting n-branch CNN networks to form a graph convolutional neural network, wherein each CNN network structure adopts a structure of convolutional pooling-convolutional-convolution;
(3) Convolutional neural network of training graph
(3.1) sequentially inputting the preprocessed PCB image into a graph convolution neural network, extracting n output features through a CNN network of each branch, and respectively embedding the n output features into n nodes to serve as initial node features, wherein the ith node V is i Is marked as
Figure BDA0003173307070000021
Figure BDA0003173307070000022
(3.2) updating the characteristics of each node based on the message passing mechanism;
setting a maximum iteration number K, and initializing a current iteration number t =1,2, …, K;
node V i Initial characteristics of
Figure BDA0003173307070000023
The node characteristic after updating the iteration k times via the message passing mechanism is recorded as ^ 4>
Figure BDA0003173307070000024
Node V i The message passed over k iterations is marked as ≥ h>
Figure BDA0003173307070000025
Node V i The edge between the adjacent node is marked as->
Figure BDA0003173307070000026
Edge weights between adjacent nodes are recorded as +>
Figure BDA0003173307070000027
Wherein j =1,2, …, n and i ≠ j; finally, constructing a graph structure by adopting a bidirectional relationship based on a message transfer mechanism;
(3.3) updating the characteristics of the n nodes after iteration
Figure BDA0003173307070000028
Performing characteristic channel splicing to obtain output characteristics, and performing deconvolution calculation to up-sample the output characteristics to the same size of the input image, thereby obtaining a predicted coordinate characteristic diagram X;
(3.4) performing sigmoid () function operation on each element in the predicted coordinate feature map X to obtain a probability density distribution map F (X) of PCB image positioning;
(3.5) detecting connected regions on the F (x), recording each connected region as a component, acquiring the centroid coordinate of the connected region, taking the centroid coordinate as the predicted center coordinate of the component, and recording the centroid coordinate as the predicted center coordinate of the component
Figure BDA0003173307070000029
(3.6) calculating the predicted center coordinate F e The error from the center coordinate F, establishes the following loss function:
Figure BDA0003173307070000031
wherein L (θ) is a loss function value; theta is a learning parameter of the graph convolution neural network;
Figure BDA0003173307070000032
when L (theta) is the mostThe predicted central coordinate of the first component under the parameter theta of the graph convolution neural network when the value is small;
(3.7) repeating the steps (3.1) - (3.6) until the atlas neural network converges, thereby obtaining the atlas neural network for automatically positioning the PCB component;
(4) Automatic positioning of PCB components
And (3) preprocessing the PCB image to be positioned according to the method in the step (1.2), inputting the preprocessed PCB image into the graph convolution neural network, and directly outputting the central coordinates of each component in the PCB image by the graph convolution neural network so as to realize automatic positioning.
The invention aims to realize the following steps:
the invention relates to a PCB component automatic positioning method based on a graph convolution neural network, which comprises the steps of collecting a plurality of PCB images, marking a corresponding relation between each PCB image and the position of a component in the image in a manual calibration mode, recording the central coordinate of each component, sequentially inputting the preprocessed PCB images into the graph convolution neural network, updating the characteristics of each node based on a message transfer mechanism, further completing the iterative training of the graph convolution neural network, finally inputting the PCB image to be positioned into the graph convolution neural network, and directly outputting the central coordinate of each component in the PCB image by the graph convolution neural network, thereby realizing automatic positioning.
Meanwhile, the automatic PCB component positioning method based on the graph convolution neural network further has the following beneficial effects:
(1) The accuracy of positioning the components is higher than that of positioning the components based on a convolution neural network, and the effect on positioning accuracy is better.
(2) The component positioning based on the graph convolution nerve is applied to automatic positioning of components on an image with certain lens distortion, such as an image shot by a mobile phone, and is shown in figure 6.
(3) The element positioning method based on the graph convolution nerve is not only suitable for automatic positioning of elements of the PCB light panel, but also suitable for positioning of elements on the mounted printed circuit board image, such as the element positioning on the PCB light panel shown in fig. 5 and the element positioning on the mounted printed circuit board image shown in fig. 6.
Drawings
FIG. 1 is a flow chart of the PCB component automatic positioning method based on the graph convolution neural network of the invention;
FIG. 2 is a graph convolution neural network;
FIG. 3 is a schematic diagram of the figure structure;
FIG. 4 is a graph convolution neural network training flow diagram;
FIG. 5 is a graph of the results of a localization of a neural network on a test set based on graph convolution;
fig. 6 is a graph of the results of a localization based on a convolutional neural network on a handset acquired dataset.
Detailed Description
Specific embodiments of the present invention are described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.
Examples
FIG. 1 is a flow chart of the PCB component automatic positioning method based on the graph convolution neural network.
In this embodiment, as shown in fig. 1, the method for automatically positioning PCB components based on a graph convolution neural network of the present invention includes the following steps:
s1, collecting and preprocessing PCB (printed Circuit Board) images
S1.1, collecting a plurality of PCB images, marking the corresponding relation between each PCB image and the position of a component in the image in a manual calibration mode, and recording the central coordinate F of each component l L =1,2, …, N represents the number of components in the PCB image;
s1.2, scaling each PCB image to 256x256x3 resolution, and then carrying out normalization processing on each scaled PCB image to enable the value range of each pixel of the scaled image to be normalized to be between [0,1 ];
s2, building a graph convolution neural network
In the present embodiment, as shown in fig. 2, a graph convolutional neural network is formed by using 3-branch CNN networks, where each CNN network structure adopts a structure of convolution pooling-convolution;
s3, training graph convolution neural network
S3.1, sequentially inputting the preprocessed PCB image with the size of 256x256x3 into a graph convolution neural network, extracting output features with the feature size of 16x64x64 through a CNN network of each branch as shown in FIG. 2, and respectively embedding the 3 output features into 3 nodes as initial features of the nodes, wherein the ith node V is i Is marked as
Figure BDA0003173307070000041
S3.2, updating the characteristics of each node based on a message transfer mechanism;
setting a maximum iteration number K =3, and initializing a current iteration number K =1,2,3;
as shown in fig. 4, node V i Initial characteristics of
Figure BDA0003173307070000051
The node characteristic after updating the iteration k times through the message passing mechanism is marked as->
Figure BDA0003173307070000052
Figure BDA0003173307070000053
Wherein conv (·) represents a convolution operation;
message generation is primarily dependent on node characteristics and messages between nodes, then node V i The message after k iterations is recorded
Figure BDA0003173307070000054
Figure BDA0003173307070000055
Wherein mean (-) represents a calculationNode V i Is weighted with the adjacent edge->
Figure BDA0003173307070000056
The sum of the products of (c) is taken as the mean value, phi (-) represents the averaged feature and the node V i Is characterized by>
Figure BDA0003173307070000057
Splicing the characteristic channels;
node V i The edges between adjacent nodes are marked as
Figure BDA0003173307070000058
Edge weights between adjacent nodes are recorded as +>
Figure BDA0003173307070000059
Wherein j =1,2,3 and i ≠ j; finally, a graph structure is constructed by adopting a bidirectional relationship based on a message passing mechanism, and in the embodiment, the graph structure formed by three nodes is shown in fig. 3;
s3.3, updating the 3 node characteristics with the size of 64x64x16 after iteration
Figure BDA00031733070700000510
Splicing the characteristic channels to obtain output characteristics with the size of 64X64X48, and then performing deconvolution calculation to up-sample the output characteristics to 256X256X1 with the same size of the input image, so as to obtain a predicted coordinate characteristic diagram X;
s3.4, performing sigmoid () function operation on each element in the predicted coordinate feature map X to obtain a probability density distribution map F (X) for PCB image positioning;
s3.5, detecting connected areas on the F (x), recording each connected area as a component, acquiring the centroid coordinate of the connected area, taking the centroid coordinate as the predicted center coordinate of the component, and recording the centroid coordinate as the predicted center coordinate of the component
Figure BDA00031733070700000511
S3.6, calculating and predicting center coordinate F e The error from the center coordinate F, establishes the following loss function:
Figure BDA00031733070700000512
wherein L (θ) is a loss function value; theta is a learning parameter of the graph convolution neural network;
Figure BDA00031733070700000513
representing the predicted central coordinate of the ith element under the graph convolution neural network parameter theta when L (theta) takes the minimum value;
s3.7, repeating the steps S3.1-S3.6 until the graph convolution neural network is converged, thereby obtaining the graph convolution neural network for automatically positioning the PCB component;
s4, automatically positioning PCB components
And (3) preprocessing the PCB image to be positioned according to the method in the step S1.2, then inputting the preprocessed PCB image to the graph convolution neural network, and directly outputting the central coordinates of each component in the PCB image by the graph convolution neural network so as to realize automatic positioning.
Experimental verification
During the experiment, the graph convolution neural network uses the same setup, where the edge weights remain unchanged during the message passing, i.e. the graph convolution neural network is used for the same purpose
Figure BDA0003173307070000061
And setting the iteration number epoch of network training =800, the learning rate lr =0.0001, and the message transmission iteration number K =3.
Based on the results of the localization of the atlas neural network on the test set, as shown in FIG. 5, where (a) represents the input image and (b) represents the true localization; (c) the proposed convolution network positioning result;
and (4) positioning results of the neural network on the data set acquired by the mobile phone based on the graph convolution. As shown in fig. 6, (a) represents an input image, (b) represents true-value localization; (c) graphically wrapping the network positioning results;
table 1 shows the results of the quantitative analysis of the component placement on the front and back images of 6 PCB light panels, including the number of components in the images. GNN is the method proposed by the present invention, and the positioning results of GNN and CNN methods are compared, as shown in table 1.
TABLE 1 PCB location results quantitative analysis (unit: pixel)
Figure RE-GDA0003276506460000062
It can be seen from table 1 that the accuracy of positioning the components by the graph convolution network is higher than that of the component positioning of the ordinary convolution neural network, the positioning accuracy is better, the accuracy error of the positioning of the result of the component positioning on the PCB _2_top image is the largest, and the accuracy error on the PCB _6_bot image is the smallest.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all inventions utilizing the inventive concept are protected.

Claims (3)

1. A PCB component automatic positioning method based on a graph convolution neural network is characterized by comprising the following steps:
(1) PCB image acquisition and pretreatment
(1.1) collecting a plurality of PCB images, marking the corresponding relation between each PCB image and the position of a component in the image in a manual calibration mode, and recording the central coordinate F of each component l L =1,2, …, N represents the number of components in the PCB image;
(1.2) zooming each PCB image to the same resolution, and then carrying out normalization processing on each zoomed PCB image;
(2) Building graph convolution neural network
Adopting n-branch CNN networks to form a graph convolutional neural network, wherein each CNN network structure adopts a structure of convolutional pooling-convolutional;
(3) Convolutional neural network of training graph
(3.1) sequentially inputting the preprocessed PCB image into a graph convolution neural network, extracting n output features through the CNN network of each branch, and respectively embedding the n output features into n nodes as node initial features, wherein the ith node V is i Is marked as
Figure FDA0003173307060000011
Figure FDA0003173307060000012
(3.2) updating the characteristics of each node based on the message passing mechanism;
setting a maximum iteration number K, and initializing a current iteration number t =1,2, …, K;
node V i Initial characteristics of
Figure FDA0003173307060000013
The node characteristic after updating the iteration k times via the message passing mechanism is recorded as ^ 4>
Figure FDA0003173307060000014
Node V i The message passed over k iterations is marked as ≥ h>
Figure FDA0003173307060000015
Node V i The edge between the adjacent node is recorded as->
Figure FDA0003173307060000016
Edge weights between adjacent nodes are recorded as +>
Figure FDA0003173307060000017
Wherein j =12, …, n and i ≠ j; finally, constructing a graph structure by adopting a bidirectional relationship based on a message transfer mechanism;
(3.3) updating the characteristics of the n nodes after iteration
Figure FDA0003173307060000018
Performing characteristic channel splicing to obtain output characteristics, and performing deconvolution calculation to up-sample the output characteristics to the same size of the input image, thereby obtaining a predicted coordinate characteristic diagram X;
(3.4) carrying out sigmoid () function operation on each element in the predicted coordinate feature diagram X to obtain a probability density distribution diagram F (X) for PCB image positioning;
(3.5) detecting connected regions on the F (x), recording each connected region as a component, acquiring the centroid coordinate of the connected region, taking the centroid coordinate as the predicted center coordinate of the component, and recording the centroid coordinate as the predicted center coordinate of the component
Figure FDA0003173307060000021
(3.6) calculating the predicted center coordinate F e The error from the center coordinate F, establishes the following loss function:
Figure FDA0003173307060000022
wherein L (θ) is a loss function value; theta is a learning parameter of the graph convolution neural network;
Figure FDA0003173307060000023
representing the predicted central coordinate of the ith element under the parameter theta of the graph convolution neural network when the L (theta) takes the minimum value;
(3.7) repeating the steps (3.1) - (3.6) until the atlas neural network converges, thereby obtaining the atlas neural network for automatically positioning the PCB component;
(4) Automatic positioning of PCB components
And (3) preprocessing the PCB image to be positioned according to the method in the step (1.2), inputting the preprocessed PCB image into the graph convolution neural network, and directly outputting the central coordinates of each component in the PCB image by the graph convolution neural network so as to realize automatic positioning.
2. The method for automatically positioning PCB components based on the graph convolution neural network of claim 1, wherein the node V i Message after k times of iteration transmission
Figure FDA0003173307060000024
The calculating method comprises the following steps:
Figure FDA0003173307060000025
wherein mean (-) represents the compute node V i All the adjacent nodes and the adjacent edges of the node are weighted
Figure FDA0003173307060000026
Is averaged, phi (-) represents the averaged feature and node V i In a characteristic>
Figure FDA0003173307060000027
And splicing the characteristic channels.
3. The PCB component automatic positioning method based on the graph convolution neural network of claim 1, characterized in that the node V i Features passed through k iterations
Figure FDA0003173307060000028
The updated formula is as follows
Figure FDA0003173307060000029
Where conv (·) represents a convolution operation.
CN202110825126.2A 2021-07-21 2021-07-21 Automatic PCB component positioning method based on graph convolution neural network Active CN113674207B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110825126.2A CN113674207B (en) 2021-07-21 2021-07-21 Automatic PCB component positioning method based on graph convolution neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110825126.2A CN113674207B (en) 2021-07-21 2021-07-21 Automatic PCB component positioning method based on graph convolution neural network

Publications (2)

Publication Number Publication Date
CN113674207A CN113674207A (en) 2021-11-19
CN113674207B true CN113674207B (en) 2023-04-07

Family

ID=78539741

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110825126.2A Active CN113674207B (en) 2021-07-21 2021-07-21 Automatic PCB component positioning method based on graph convolution neural network

Country Status (1)

Country Link
CN (1) CN113674207B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114347687B (en) * 2022-01-05 2023-05-12 珠海中京电子电路有限公司 Outer offline AOI automatic upper and lower plate recording method, automatic arm machine and application

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017031630A1 (en) * 2015-08-21 2017-03-02 中国科学院自动化研究所 Deep convolutional neural network acceleration and compression method based on parameter quantification
CN109934826A (en) * 2019-02-28 2019-06-25 东南大学 A kind of characteristics of image dividing method based on figure convolutional network
CN110070536A (en) * 2019-04-24 2019-07-30 南京邮电大学 A kind of pcb board component detection method based on deep learning
CN111274903A (en) * 2020-01-15 2020-06-12 合肥工业大学 Cervical cell image classification method based on graph convolution neural network
CN111429431A (en) * 2020-03-24 2020-07-17 深圳市振邦智能科技股份有限公司 Element positioning and identifying method based on convolutional neural network
CN111462088A (en) * 2020-04-01 2020-07-28 深圳前海微众银行股份有限公司 Data processing method, device, equipment and medium based on graph convolution neural network
CN111553873A (en) * 2020-01-17 2020-08-18 西安理工大学 Automatic brain neuron detection method based on multi-scale convolutional neural network
CN111612051A (en) * 2020-04-30 2020-09-01 杭州电子科技大学 Weak supervision target detection method based on graph convolution neural network
CN112132818A (en) * 2020-09-29 2020-12-25 北京航空航天大学 Image processing method for constructing three stages based on graph convolution neural network
CN113095254A (en) * 2021-04-20 2021-07-09 清华大学深圳国际研究生院 Method and system for positioning key points of human body part

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017031630A1 (en) * 2015-08-21 2017-03-02 中国科学院自动化研究所 Deep convolutional neural network acceleration and compression method based on parameter quantification
CN109934826A (en) * 2019-02-28 2019-06-25 东南大学 A kind of characteristics of image dividing method based on figure convolutional network
CN110070536A (en) * 2019-04-24 2019-07-30 南京邮电大学 A kind of pcb board component detection method based on deep learning
CN111274903A (en) * 2020-01-15 2020-06-12 合肥工业大学 Cervical cell image classification method based on graph convolution neural network
CN111553873A (en) * 2020-01-17 2020-08-18 西安理工大学 Automatic brain neuron detection method based on multi-scale convolutional neural network
CN111429431A (en) * 2020-03-24 2020-07-17 深圳市振邦智能科技股份有限公司 Element positioning and identifying method based on convolutional neural network
CN111462088A (en) * 2020-04-01 2020-07-28 深圳前海微众银行股份有限公司 Data processing method, device, equipment and medium based on graph convolution neural network
CN111612051A (en) * 2020-04-30 2020-09-01 杭州电子科技大学 Weak supervision target detection method based on graph convolution neural network
CN112132818A (en) * 2020-09-29 2020-12-25 北京航空航天大学 Image processing method for constructing three stages based on graph convolution neural network
CN113095254A (en) * 2021-04-20 2021-07-09 清华大学深圳国际研究生院 Method and system for positioning key points of human body part

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
A comprehensive survey on graph neural networks;Zonghan Wu等;《IEEE Transactions on Neural Networks and Learning Systems》;4-24 *
Defect detection in printed circuit boards using you-only-look-once convolutional neural networks;Venkat Anil Adibhatla等;《Electronics》;1-16 *
GCN图卷积网络入门详解;雷峰网leiphone;《https://baijiahao.baidu.com/s?id=1678519457206249337&wfr=spider&for=pc》;1-10 *
Graph neural networks: A review of methods and applications;Jie Zhou等;《AI Open》;57-81 *
Using convolutional neural networks for character verification on integrated circuit components of printed circuit boards;Chun-Hui Lin等;《Applied Intelligence》;4022-4032 *
Wire segmentation for printed circuit board using deep convolutional neural network and graph cut model;Kai Qiao等;《IET Image Processing》;793-800 *
基于图神经网络的行人重识别算法;蒋若辉等;《电子技术与软件工程》;116-117 *

Also Published As

Publication number Publication date
CN113674207A (en) 2021-11-19

Similar Documents

Publication Publication Date Title
CN108898086B (en) Video image processing method and device, computer readable medium and electronic equipment
CN106845487B (en) End-to-end license plate identification method
CN110222764A (en) Shelter target detection method, system, equipment and storage medium
CN112115895B (en) Pointer type instrument reading identification method, pointer type instrument reading identification device, computer equipment and storage medium
CN109961009A (en) Pedestrian detection method, system, device and storage medium based on deep learning
CN109190446A (en) Pedestrian's recognition methods again based on triple focused lost function
CN111950453A (en) Optional-shape text recognition method based on selective attention mechanism
CN111178120B (en) Pest image detection method based on crop identification cascading technology
CN111008632B (en) License plate character segmentation method based on deep learning
CN113674207B (en) Automatic PCB component positioning method based on graph convolution neural network
CN115359239A (en) Wind power blade defect detection and positioning method and device, storage medium and electronic equipment
CN110263790A (en) A kind of power plant's ammeter character locating and recognition methods based on convolutional neural networks
CN111861914A (en) Low-illumination color polarization image enhancement method based on deep neural network
CN114723646A (en) Image data generation method with label, device, storage medium and electronic equipment
CN110837860A (en) Paster detection method based on deep learning and related system
CN113362279A (en) Intelligent concentration detection method of immunochromatographic test paper
CN110321869A (en) Personnel's detection and extracting method based on Multiscale Fusion network
CN111881914B (en) License plate character segmentation method and system based on self-learning threshold
CN113112511A (en) Method and device for correcting test paper, storage medium and electronic equipment
CN114694133B (en) Text recognition method based on combination of image processing and deep learning
CN116088264A (en) Determination method and device of process hot spot and photoetching layout design system
CN113420623B (en) 5G base station detection method and system based on self-organizing mapping neural network
CN110956174A (en) Device number identification method
CN110619622B (en) Bread board structure image automatic detection method based on computer vision
CN112712004B (en) Face detection system, face detection method and device and electronic equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant