CN109978014A - A kind of flexible base board defect inspection method merging intensive connection structure - Google Patents
A kind of flexible base board defect inspection method merging intensive connection structure Download PDFInfo
- Publication number
- CN109978014A CN109978014A CN201910166742.4A CN201910166742A CN109978014A CN 109978014 A CN109978014 A CN 109978014A CN 201910166742 A CN201910166742 A CN 201910166742A CN 109978014 A CN109978014 A CN 109978014A
- Authority
- CN
- China
- Prior art keywords
- frame model
- layer
- picture
- base board
- flexible base
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000007689 inspection Methods 0.000 title claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 24
- 230000002950 deficient Effects 0.000 claims abstract description 11
- 238000013526 transfer learning Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 3
- 238000000605 extraction Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 19
- 238000010586 diagram Methods 0.000 description 3
- 238000011179 visual inspection Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000003252 repetitive effect Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 239000000758 substrate Substances 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000007800 oxidant agent Substances 0.000 description 1
- 230000003647 oxidation Effects 0.000 description 1
- 238000007254 oxidation reaction Methods 0.000 description 1
- 230000001590 oxidative effect Effects 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06N3/08—Learning methods
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of flexible base board defect inspection methods for merging intensive connection structure, including obtaining with defective FICS picture, it is unified for standard size after pre-processing to picture, marks position and the classification of defect, the training sample as SSD frame model；SSD frame model, the infrastructure network using VGG-16 as SSD frame are constructed, and increases N layers of convolutional layer, which further includes intensive connection structure；Training sample inputs SSD frame model, introduces transfer learning method and is trained to frame model, obtains trained SSD frame model；Picture to be detected is inputted into trained SSD frame model, exports the defective locations and type of picture to be detected.The quick positioning that flexible base board open defect may be implemented in the present invention judges solve the problem of traditional shortcoming detection method speed is slow, and accuracy is low, can not detect number of drawbacks simultaneously with type.
Description
Technical field
The present invention relates to machine vision surface defects detection technical field, and in particular to a kind of to merge intensive connection structure
Flexible base board defect inspection method.
Background technique
Flexible base board (Flexible Integrated Circuit Substrate, abbreviation FICS), is that one kind can be used as
The flexible print wiring board of IC package substrate.In the production process of FICS, since the precision of technical process control is insufficient, it can produce
Raw apparent defect.By high-precision vision-based detection, the various appearances defect of the FICS produced is quickly positioned
It is the key that realize quality control in FICS manufacturing process with type identification.Most of flexible base board manufacturers still adopt at present
Manual Visual Inspection is taken to carry out the detection of open defect to FICS.The detection efficiency of Manual Visual Inspection is relatively low, will cause human resources
Significant wastage, and as worker-hours increase, false detection rate can increase, it is difficult to guarantee the quality of detection.Some experts
The detection method based on digital picture characteristic processing is proposed, but these method speed are slower, it is difficult to meet the reality of FICS detection
Border application demand；Meanwhile these methods can only detect the special defect of single, it can not be to a plurality of types of open defects on FICS
It is identified simultaneously.
Summary of the invention
In order to overcome shortcoming and deficiency of the existing technology, the present invention provides a kind of flexibility for merging intensive connection structure
Base board defect detection method.
The present invention adopts the following technical scheme:
A kind of flexible base board defect inspection method merging intensive connection structure, comprising:
It obtains with defective FICS picture, is unified for standard size after pre-processing to picture, marks the position of defect
It sets and classification, the training sample as SSD frame model；
SSD frame model, the infrastructure network using VGG-16 as SSD frame are constructed, and increases N layers of convolutional layer,
The infrastructure network further includes intensive connection structure；
Training sample inputs SSD frame model, introduces transfer learning method and is trained to frame model, is trained
SSD frame model；
Picture to be detected is inputted into trained SSD frame model, exports the defective locations and type of picture to be detected.
The pretreatment includes carrying out data enhancing to FICS picture, then carries out different angle rotation to picture, finally
The distortion of plus noise and form is carried out to postrotational image.
The VGG-16 is made of 15 convolutional layers, 5 pond layers and 1 softmax classification layer.
SSD frame model of the present invention utilizes the characteristic pattern on 5 convolutional layers in 15+N convolutional layer, the extraction of multi-layer
Then feature is attached with the classification layer of softmax below, is converted into the feature vector for being identified and being classified.
The transfer learning method, specifically network model carry out pre-training on the data set of PASCAL VOC, then
Network parameter is taken out to initialize to network model；In the network model that the input of training sample data is built, carry out
Repetitive exercise, until frequency of training reaches the maximum number of iterations of setting, training terminates, and obtains trained SSD frame model
For detecting.
SSD frame model of the present invention utilizes the characteristic pattern on 5 convolutional layers in 15+N convolutional layer, the extraction of multi-layer
Then feature is attached with the classification layer of softmax below, is converted into the feature vector that can be identified and be classified, specifically
Are as follows:
It is the setting centered on each point of characteristic pattern using the characteristic pattern on 5 convolutional layers in 15+N convolutional layer
3 Aspect Ratio aspect_, one of Aspect Ratio are 1:1, then generate a series of concentric square prediction blocks, square
The a length of min_size of prediction block minimum edge of shape, maximal side are
Other two Aspect Ratio is 2:1 and 3:1, each to generate two rectangle prediction blocks than regular meeting, a length ofWidth is
The min_size and max_size of each characteristic pattern are determined by formula below:
In formula, m indicates to carry out the number of plies of feature extraction, for first layer characteristic pattern, min_size=s1, max_
Size=s2；The characteristic pattern of feature, min_size=s are extracted for the second layer2, max_size=s3, and so on；
After obtaining a series of prediction blocks, the corresponding feature vector of each prediction block includes general to the prediction of c defect classification
Four parameters (cx, cy, w, h) of rate and defective locations, determine position of the prediction block in figure, cx represents prediction block center
Opposite abscissa, cy represent the opposite ordinate at prediction block center, and w represents the width of prediction block, and h represents the length of prediction block；
By feature vector input softmax classification layer, the calculating lost according to loss function, so as to adjust network ginseng
Number.
The N takes 6.
Beneficial effects of the present invention:
(1) present invention can be used for quickly detecting flexible base board open defect, realize to the substitution of Manual Visual Inspection, mention
The working efficiency of high detection process；
(2) detection method can be quickly detected from different types of defect on flexible base board automatically, and to defect
Position and type are marked, solve the defect in the case of the insoluble random site of traditional images algorithm quickly position with
Test problems；
(3) after the detection model for having trained picture to be detected input, position and the classification letter of defect can directly be exported
Breath, to skip this time-consuming step of image characteristic analysis, solution traditional images processing method speed is slow, and nothing
The problem of method detects multiple types defect simultaneously.
Detailed description of the invention
Fig. 1 is work flow diagram of the invention；
Fig. 2 is VGG-16 schematic network structure of the invention；
Fig. 3 is model training schematic diagram of the present invention；
Fig. 4 (a), Fig. 4 (b), Fig. 4 (c) and Fig. 4 (d) are respectively the present embodiment detection short circuit, open circuit, draw oxidant holes
Defect schematic diagram.
Specific embodiment
Below with reference to examples and drawings, the present invention is described in further detail, but embodiments of the present invention are not
It is limited to this.
Embodiment
As shown in Figure 1, Figure 2 and Figure 3, a kind of flexible base board defect inspection method merging intensive connection structure, comprising:
In the factory, using micro-photographing apparatus, and FICS picture is shot using various light sources, is not shared the same light
According to the defective FICS picture of lower band, picture is pre-processed, carries out data enhancing, then every FICS image is carried out not
It with the rotation of angle, is rotated by 90 ° respectively, rotates 180 °, then the image after 270 ° of rotation carries out postrotational image
The transformation of a variety of digital picture features, the distortion including plus noise and form, has expanded sample size, has avoided the occurrence of over-fitting
The problem of.Then all images are unified into 300 × 300 standard size, and to the position of FICS defect and classification in image
Information is labeled, and is made into training sample.
SSD frame model, the infrastructure network using VGG-16 as SSD frame are constructed, and increases N layers of convolutional layer,
The infrastructure network further includes intensive connection structure, reduces number of parameters, simplifies and calculates.Joined in the present embodiment 5 it is close
Collect connection structure, is intensive connection structure 1, intensive connection structure 2, intensive connection structure 3, intensive connection structure 4 and close respectively
Collect connection structure 5, and increases N=6 convolutional layer.The infrastructure network be to multiple characteristic layers carry out feature extraction, and
The means of transfer learning are introduced in training process, to make deep neural network study more information into FICS picture.
The detailed process of training SSD frame model in the present embodiment are as follows:
VGG-16 of the present invention is by 15 convolutional layers, and 5 pond layers and 1 softmax classification layer are constituted, on this basis
Increase by 6 convolutional layers, while introducing intensive connection structure in VGG-16 basic network.Using the 12nd layer in network model, the
Characteristic pattern on 17 layers, the 19th layer, the 21st layer, the 23rd layer, this 5 layers, extracts feature to multi-layer, with rearmost softmax
Classification layer is attached, and is converted into the feature vector that can be identified and be classified.
The method for introducing transfer learning, first allows the network architecture to carry out pre-training on the data set of PASCAL VOC, then
Network parameter is taken out to initialize to network model.In the network model that the input of training sample data is built, carry out
Repetitive exercise, until frequency of training reaches the maximum number of iterations of setting, training terminates, and obtains detection model
Multi-layer extracts the process of feature in the present embodiment, specific as follows:
Using the characteristic pattern on the 12nd layer in network model, the 17th layer, the 19th layer, the 21st layer, the 23rd layer, this 5 layers, with spy
Centered on each point for levying figure, 3 Aspect Ratio aspect_ are set, one of Aspect Ratio is 1:1, then generates a series of
Concentric square prediction block, the square a length of min_size of prediction block minimum edge, maximal side areOther two Aspect Ratio is 2:1 and 3:1, each to generate two rectangles than regular meeting
Prediction block, it is a length ofWidth isEach
The min_size and max_size of characteristic pattern are determined by formula below:
In formula, m represents the number of plies of feature extraction to be carried out.smin=0.2, smax=0.9, m=6, for first
Layer characteristic pattern, min_size=s1, max_size=s2.The characteristic pattern of feature, min_size=s are extracted for the second layer2,
Max_size=s3, and so on.
After generating a series of prediction blocks according to above step, feature vector corresponding to each prediction block is contained to c
The prediction probability of classification and 4 parameters (cx, cy, w, h) of defective locations, cx represent the opposite abscissa at prediction block center,
Cy represents the opposite ordinate at prediction block center, and w represents the width of prediction block, and h represents the length of prediction block.Feature vector is inputted
Softmax classification layer, according to the calculating that loss function can be lost, so as to adjust network parameter.
Loss function contain to position loss calculating and classification loss calculating, wherein position loss be prediction block and
SmoothL1 loses between actual position, and calculation formula is as follows:
Wherein,
This part includes to prediction block center with respect to abscissa cx, and prediction block center is with respect to ordinate cy, prediction block
The calculating of the smoothL1 loss of width w, prediction block height h.N is exactly the number for the prediction block that can be matched with object actual position
Amount.
Classification loss is lost using a multi-class softmax, and calculation formula is as follows:
Wherein,If i-th of prediction block can match with j-th of true frame
It is right, thenOtherwise,N is exactly the quantity for the prediction block that can be matched with object actual position.Previous item
Indicate that the frame for foreground classification belongs to the penalty values of the probabilistic forecasting of each classification to object.Latter is indicated for background point
The frame of class, to the penalty values of each class probability prediction.
Picture to be detected is uniformly adjusted to size 300*300 after pretreatment, inputs trained SSD frame mould
Type obtains the defective locations and type of picture to be detected.
Open defect detection method of the present invention is used for the short circuit, open circuit of route in FICS image, damaged, oxidation
The defects of detection, detection effect figure such as Fig. 4 (a), Fig. 4 (b), shown in Fig. 4 (c) and Fig. 4 (d).The method of the present invention passes through training
Detection model out based on deep learning SSD frame can be exported in FICS image input model to be detected quickly
The position of defect on FICS image and its classification information.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by the embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (7)
1. a kind of flexible base board defect inspection method for merging intensive connection structure characterized by comprising
Obtain with defective FICS picture, be unified for standard size after pre-processing to picture, mark the position of defect with
Classification, the training sample as SSD frame model；
SSD frame model, the infrastructure network using VGG-16 as SSD frame are constructed, and increases N layers of convolutional layer, the base
Plinth network structure further includes intensive connection structure；
Training sample inputs SSD frame model, introduces transfer learning method and is trained to frame model, obtains trained
SSD frame model；
Picture to be detected is inputted into trained SSD frame model, exports the defective locations and type of picture to be detected.
2. flexible base board defect inspection method according to claim 1, which is characterized in that the pretreatment includes to FICS
Picture carries out data enhancing, then carries out different angle rotation to picture, finally carries out plus noise and shape to postrotational image
The distortion of state.
3. flexible base board defect inspection method according to claim 1, which is characterized in that the VGG-16 is by 15 convolution
Layer, 5 pond layers and 1 softmax classification layer are constituted.
4. flexible base board defect inspection method according to claim 3, which is characterized in that SSD frame model utilizes 15+N
Then the characteristic pattern on 5 convolutional layers in a convolutional layer, the extraction feature of multi-layer are carried out with the classification layer of softmax below
Connection, is converted into the feature vector for being identified and being classified.
5. flexible base board defect inspection method according to claim 1, which is characterized in that the transfer learning method, tool
Body is that network model carries out pre-training on the data set of PASCAL VOC, and network parameter is then taken out to carry out network model
Initialization；In the network model that the input of training sample data is built, it is iterated training, until frequency of training reaches setting
Maximum number of iterations, training terminates, and obtains trained SSD frame model for detecting.
6. flexible base board defect inspection method according to claim 4, which is characterized in that SSD frame model utilizes 15+N
Then the characteristic pattern on 5 convolutional layers in a convolutional layer, the extraction feature of multi-layer are carried out with the classification layer of softmax below
Connection, is converted into the feature vector that can be identified and be classified, specifically:
It is to be arranged 3 centered on each point of characteristic pattern using the characteristic pattern on 5 convolutional layers in 15+N convolutional layer
Aspect Ratio aspect_ratio, one of Aspect Ratio are 1:1, then generate a series of concentric square prediction blocks, just
The rectangular a length of min_size of prediction block minimum edge, maximal side are
Other two Aspect Ratio is 2:1 and 3:1, each to generate two rectangle prediction blocks than regular meeting, a length ofWidth is
The min_size and max_size of each characteristic pattern are determined by formula below:
In formula, m indicates to carry out the number of plies of feature extraction, for first layer characteristic pattern, min_size=s1, max_size
=s2；The characteristic pattern of feature, min_size=s are extracted for the second layer2, max_size=s3, and so on；
After obtaining a series of prediction blocks, the corresponding feature vector of each prediction block include to the prediction probability of c defect classification with
And four parameters (cx, cy, w, h) of defective locations, determine position of the prediction block in figure, cx represents the opposite of prediction block center
Abscissa, cy represent the opposite ordinate at prediction block center, and w represents the width of prediction block, and h represents the length of prediction block；
By feature vector input softmax classification layer, the calculating lost according to loss function, so as to adjust network parameter.
7. flexible base board defect inspection method according to claim 1 or 6, which is characterized in that the N takes 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166742.4A CN109978014A (en) | 2019-03-06 | 2019-03-06 | A kind of flexible base board defect inspection method merging intensive connection structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166742.4A CN109978014A (en) | 2019-03-06 | 2019-03-06 | A kind of flexible base board defect inspection method merging intensive connection structure |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109978014A true CN109978014A (en) | 2019-07-05 |
Family
ID=67077992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910166742.4A Pending CN109978014A (en) | 2019-03-06 | 2019-03-06 | A kind of flexible base board defect inspection method merging intensive connection structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109978014A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127417A (en) * | 2019-12-20 | 2020-05-08 | 江苏理工学院 | Soft package coil stock printing defect detection method based on SIFT feature matching and improved SSD algorithm |
| CN111640089A (en) * | 2020-05-09 | 2020-09-08 | 武汉精立电子技术有限公司 | Defect detection method and device based on feature map center point |
| CN111723690A (en) * | 2020-06-03 | 2020-09-29 | 北京全路通信信号研究设计院集团有限公司 | Circuit equipment state monitoring method and system |
| CN112686887A (en) * | 2021-01-27 | 2021-04-20 | 上海电气集团股份有限公司 | Method, system, equipment and medium for detecting concrete surface cracks |
| CN113077430A (en) * | 2021-03-30 | 2021-07-06 | 太原理工大学 | Laser chip defect detection and classification method and system based on SSD algorithm |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229658A (en) * | 2017-12-27 | 2018-06-29 | 深圳先进技术研究院 | The implementation method and device of object detector based on finite sample |
| CN108345911A (en) * | 2018-04-16 | 2018-07-31 | 东北大学 | Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics |
| CN108805166A (en) * | 2018-05-03 | 2018-11-13 | 全球能源互联网研究院有限公司 | It is a kind of to establish image classification neural network model and image classification method, device |
| CN108846446A (en) * | 2018-07-04 | 2018-11-20 | 国家新闻出版广电总局广播科学研究院 | The object detection method of full convolutional network is merged based on multipath dense feature |
| CN109035184A (en) * | 2018-06-08 | 2018-12-18 | 西北工业大学 | A kind of intensive connection method based on the deformable convolution of unit |
| CN109034210A (en) * | 2018-07-04 | 2018-12-18 | 国家新闻出版广电总局广播科学研究院 | Object detection method based on super Fusion Features Yu multi-Scale Pyramid network |
| CN109142371A (en) * | 2018-07-31 | 2019-01-04 | 华南理工大学 | High density flexible exterior substrate defect detecting system and method based on deep learning |
| CN109325947A (en) * | 2018-09-20 | 2019-02-12 | 北京航空航天大学 | A kind of SAR image steel tower object detection method based on deep learning |
| CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
| CN109376589A (en) * | 2018-09-07 | 2019-02-22 | 中国海洋大学 | ROV deformation target and Small object recognition methods based on convolution kernel screening SSD network |
-
2019
- 2019-03-06 CN CN201910166742.4A patent/CN109978014A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229658A (en) * | 2017-12-27 | 2018-06-29 | 深圳先进技术研究院 | The implementation method and device of object detector based on finite sample |
| CN108345911A (en) * | 2018-04-16 | 2018-07-31 | 东北大学 | Surface Defects in Steel Plate detection method based on convolutional neural networks multi-stage characteristics |
| CN108805166A (en) * | 2018-05-03 | 2018-11-13 | 全球能源互联网研究院有限公司 | It is a kind of to establish image classification neural network model and image classification method, device |
| CN109035184A (en) * | 2018-06-08 | 2018-12-18 | 西北工业大学 | A kind of intensive connection method based on the deformable convolution of unit |
| CN108846446A (en) * | 2018-07-04 | 2018-11-20 | 国家新闻出版广电总局广播科学研究院 | The object detection method of full convolutional network is merged based on multipath dense feature |
| CN109034210A (en) * | 2018-07-04 | 2018-12-18 | 国家新闻出版广电总局广播科学研究院 | Object detection method based on super Fusion Features Yu multi-Scale Pyramid network |
| CN109142371A (en) * | 2018-07-31 | 2019-01-04 | 华南理工大学 | High density flexible exterior substrate defect detecting system and method based on deep learning |
| CN109344821A (en) * | 2018-08-30 | 2019-02-15 | 西安电子科技大学 | Small target detecting method based on Fusion Features and deep learning |
| CN109376589A (en) * | 2018-09-07 | 2019-02-22 | 中国海洋大学 | ROV deformation target and Small object recognition methods based on convolution kernel screening SSD network |
| CN109325947A (en) * | 2018-09-20 | 2019-02-12 | 北京航空航天大学 | A kind of SAR image steel tower object detection method based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| GAO HUANG，等: "Densely Connected Convolutional Networks", 《2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》 * |
| WEI LIU,等: "SSD：Single Shot MultiBox Detector", 《EUROPEAN CONFERENCE ON COMPUTER VISION.AMSTERDAM:SPRINGER 2016》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111127417A (en) * | 2019-12-20 | 2020-05-08 | 江苏理工学院 | Soft package coil stock printing defect detection method based on SIFT feature matching and improved SSD algorithm |
| CN111127417B (en) * | 2019-12-20 | 2024-02-06 | 江苏理工学院 | Printing defect detection method based on SIFT feature matching and SSD algorithm improvement |
| CN111640089A (en) * | 2020-05-09 | 2020-09-08 | 武汉精立电子技术有限公司 | Defect detection method and device based on feature map center point |
| CN111640089B (en) * | 2020-05-09 | 2023-08-15 | 武汉精立电子技术有限公司 | Defect detection method and device based on feature map center point |
| CN111723690A (en) * | 2020-06-03 | 2020-09-29 | 北京全路通信信号研究设计院集团有限公司 | Circuit equipment state monitoring method and system |
| CN111723690B (en) * | 2020-06-03 | 2023-10-20 | 北京全路通信信号研究设计院集团有限公司 | Method and system for monitoring state of circuit equipment |
| CN112686887A (en) * | 2021-01-27 | 2021-04-20 | 上海电气集团股份有限公司 | Method, system, equipment and medium for detecting concrete surface cracks |
| CN113077430A (en) * | 2021-03-30 | 2021-07-06 | 太原理工大学 | Laser chip defect detection and classification method and system based on SSD algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978014A (en) | A kind of flexible base board defect inspection method merging intensive connection structure | |
| Huang et al. | Fabric defect segmentation method based on deep learning | |
| CN108982508B (en) | Plastic package IC chip defect detection method based on feature template matching and deep learning | |
| Cai et al. | SMT solder joint inspection via a novel cascaded convolutional neural network | |
| CN109829903B (en) | Chip surface defect detection method based on convolution denoising autoencoder | |
| CN110070536A (en) | A kind of pcb board component detection method based on deep learning | |
| CN106570511A (en) | Keyboard character defect online detection system based on computer vision and keyboard character defect online detection method thereof | |
| CN109859207B (en) | Defect detection method of high-density flexible substrate | |
| CN104268538A (en) | Online visual inspection method for dot matrix sprayed code characters of beverage cans | |
| Gao et al. | A line-based-clustering approach for ball grid array component inspection in surface-mount technology | |
| CN109544522A (en) | A kind of Surface Defects in Steel Plate detection method and system | |
| CN110490842A (en) | A kind of steel strip surface defect detection method based on deep learning | |
| CN108985337A (en) | A kind of product surface scratch detection method based on picture depth study | |
| CN106874913A (en) | A kind of vegetable detection method | |
| CN112686894A (en) | FPCB (flexible printed circuit board) defect detection method and device based on generative countermeasure network | |
| CN109272060A (en) | A kind of method and system carrying out target detection based on improved darknet neural network | |
| Wang et al. | Attention-based deep learning for chip-surface-defect detection | |
| CN110865077A (en) | Visual inspection system for appearance defects in RFID antenna production | |
| CN108709500B (en) | Circuit board element positioning and matching method | |
| Xie et al. | A method for surface defect detection of printed circuit board based on improved YOLOv4 | |
| CN107478152A (en) | TR chip positionings method and detection method | |
| Ling et al. | Precise Detection for Dense PCB Components Based on Modified YOLOv8 | |
| CN112132798B (en) | Method for detecting complex background PCB mark point image based on Mini ARU-Net network | |
| Aggarwal et al. | A survey on automatic printed circuit board defect detection techniques | |
| CN109815957A (en) | A kind of character recognition method based on color image under complex background |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190705 |