CN107481231A - A kind of handware defect classifying identification method based on depth convolutional neural networks - Google Patents
A kind of handware defect classifying identification method based on depth convolutional neural networks Download PDFInfo
- Publication number
- CN107481231A CN107481231A CN201710707883.3A CN201710707883A CN107481231A CN 107481231 A CN107481231 A CN 107481231A CN 201710707883 A CN201710707883 A CN 201710707883A CN 107481231 A CN107481231 A CN 107481231A
- Authority
- CN
- China
- Prior art keywords
- network
- neural networks
- convolutional neural
- handware
- layer
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
A kind of handware defect classifying identification method based on depth convolutional neural networks, comprises the following steps；Network is established, constructs a depth convolutional neural networks；Training network, the image collected is divided into two major classes, i.e. training set and test set, the training set, which accounts for, collects the 70% of total number of images, and the test set, which accounts for, collects the 30% of total number of images；Defect recognition, the network that the handware image input in test set has been trained, checks output result, recognition result is compareed with the label of image, count correct recognition rata and error recognition rate.Depth convolutional neural networks are used in this algorithm, the image processing algorithm of complexity is eliminated, by increasing network depth, extracts the more abstract feature of defect, make that there is between different defect classifications stronger ga s safety degree, discrimination is higher.
Description
Technical field
The present invention relates to the technical field of image processing based on artificial intelligence, more particularly to one kind to be based on depth convolutional Neural
The handware defect classifying identification method of network.
Background technology
Machine vision is also known as computer vision, is that research imitates human eye and brain respectively using camera and computer, with
Just replace people to detect and judge with machine, complete the science of the task such as target identification and industrial detection.Machine vision technique collection
A multi-disciplinary applied technology-oriented discipline such as Digital Image Processing, artificial intelligence, computer graphics is closed, in automatic metaplasia
It is widely used in production.In recent years, with the progress of computer technology and the constantly improve of neural network theory, computer has been promoted
The fast development of vision.China's machine vision industry develops rapidly, is occupied in automated production detection field highly important
Status.
Because handware has the advantages that to be easy to shaping, light weight, material are easily obtained, are adapted to produce in enormous quantities, it is in
The field such as electricity, machinery, chemical industry, aviation is with quite varied.It is more and more wider with the application of handware, rapid shaping process technology
Development it is more and more faster, requirement of the people to the quality of handware also more and more higher.The quality of handware mainly has size, outward appearance
Etc. requirement.Outward appearance is to ensure an important step of handware quality, and generally use artificial detection in actual production
Mode carry out.Manual detection mode efficiency is low, automaticity is not high, and its accuracy rate often passes through with the work of testing staff
Test relevant with attitude.At present, handware production enterprise increasingly focuses on improving production automation level, to production efficiency
It is required that more and more higher, manual detection mode is increasingly unable to meet demand.In addition, in process of manufacture, due to raw material thing
Property Parameters variation, technological parameter be unreasonable and the factors such as processing equipment performance is bad, and hardware, which occurs, damages, sand holes, scrapes
Wound, lack material, deformation, pit, greasy dirt etc. surface defect.These surface defects can not only destroy the outward appearance of hardware, and
Its performance can be influenceed to lead to not use.The surface defects detection of current hardware and identification it is main manually based on,
It is inefficient, automaticity is low.
The development and extensive use of machine vision technique can exactly solve the above problems.Inspection based on machine vision technique
Examining system mainly has the advantage that compared with manual detection mode：
(1) precision is high
The measurement accuracy of NI Vision Builder for Automated Inspection can reach 0.01mm levels of precision, the mankind significantly larger than limited by physical condition
Vision.
(2) it is repeated
NI Vision Builder for Automated Inspection can efficiently and accurately be repeatedly performed Detection task, will not feel tired as testing staff
Labor.And human eye can feel trickle difference in rechecking product because of the influence of various factors, accuracy rate is influenceed.
(3) real-time
NI Vision Builder for Automated Inspection efficiently carries out IMAQ, storage and processing using computer.System carries out image automatically
The transmission of data, can real time reaction production scene situation.
(4) it is untouchable
NI Vision Builder for Automated Inspection detection when need not be contacted with workpiece, therefore will not typically cause workpiece deform and produce it is unfavorable
Influence.Worked in addition, system can substitute testing staff under the adverse circumstances such as poisonous, high temperature.
(5) cost is low
NI Vision Builder for Automated Inspection can for a long time, carry out operation incessantly, can complete equivalent to multiple testing staff appoint
Business.Production cost can be greatly reduced in now cost of labor more and more higher, NI Vision Builder for Automated Inspection.
Technology reproduction that artificial neural network is biological neural network under certain simplification meaning, as a subject, it
Main task be to build practical artificial neural network, design according to the needs of the principle and practical application of biological neural network
Corresponding learning algorithm, certain intelligency activity of human brain is simulated, then technically realized out to solving practical problems.
Application in terms of computer vision is also relatively more, is mainly used to that Classification and Identification will be carried out the defects of extracting.
The outline flowchart of traditional shortcoming detection recognizer is as shown in Figure 1.It carries out figure firstly the need of to input picture
Picture is handled so that defect area to be split from image.Then various defect characteristics are analyzed and selects suitable, area
The higher feature of indexing.And then the artificial extraction of feature is carried out, and the feature of these extractions is inputted into BP neural network or SVM
The common classification device such as (SVMs) is classified, and finally provides Classification and Identification result in output end.As can be seen here, above-mentioned biography
The defects of system the order of accuarcy split highly dependent upon defect area of recognizer, and need artificially to select and extract defect spy
Sign.But for this paper handware product image, divide because it is present than more serious noise jamming, the accurate of defect
The image processing flow needed using complexity is cut, amount of calculation is very big.In addition, effectively choose the higher feature of discrimination and right
It is described often relatively difficult, it is necessary to very professional knowledge and preferable priori.
Convolutional neural networks are a kind of deep neural networks proposed by LeCun, and it directly can make a width two dimensional image
For input, the image preprocessing without making complexity to raw image data.Convolutional neural networks automatically from image extraction,
Assemblage characteristic, visual pattern is identified on the basis of feature is extracted, then provides classification results in output end.Utilize convolutional Neural
Network carry out image recognition brief model below figure 2 show, compared with traditional shortcoming recognizer, its need not artificially choose and
Expressive Features, avoid a large amount of calculating.In addition, convolutional neural networks can identify the pattern changed, have to geometry deformation
Robustness, can admissible chart picture distortion the advantages that.
The content of the invention
It is an object of the invention to solve the above problems to propose a kind of handware defect based on depth convolutional neural networks
Classifying identification method.
In order to reach this purpose, the present invention uses following technical scheme：
A kind of handware defect classifying identification method based on depth convolutional neural networks, comprises the following steps：
A. network is established, constructs a depth convolutional neural networks；
B. training network, image the defects of collecting is divided into two major classes, i.e. training set and test set, the training set accounts for
The 70% of total number of images is collected, the test set, which accounts for, collects the 30% of total number of images；
C. defect recognition, the network that the handware defect image input in test set has been trained, checks output knot
Fruit, recognition result is compareed with the label of image, count correct recognition rata and error recognition rate.
More excellent, training network algorithm comprises the following steps described in step B：
Step 1: first, the weights of network are initialized, and it is 0 weight distribution is submitted to average, variance is
0.01 Gaussian Profile, while number of the weights more than 0 is approximately equal to the number less than 0；
Step 2: training sample goes average, the training sample is colour picture, and each pixel has tri- components of R, G, B
Add and average, when handware sample inputs, three components of sample all pixels are subtracted into average, then input net
Network；
Step 3: network training, the network training uses stochastic gradient descent method.
More excellent, the stochastic gradient descent method comprises the following steps：
Training sample is inputted network by step a, propagated forward, the propagated forward one by one, by convolutional layer, excitation layer with
And grader output result of calculation, label is contrasted, calculates output error；
Step b, calculation error is fed back, the error according to step a, calculates each layer network from output layer successively forward
Error, according to each layer error, calculate right value update amount, update weight w and deviation b；
Step c, after the square-error for obtaining whole sample trainings, summation, then evolution export overall error as network,
If network overall error is more than given threshold, error and counter are recovered into initial value, re -training sample, until error is less than
Given threshold.
More excellent, assume that shared m is to training sample, each training error in the step a
More excellent, the expression formula of the overall error is
More excellent, the training network process GPU speed-up computations.
More excellent, the structure of depth convolutional neural networks described in step A includes six convolutional layers, six excitation layers, and three
Individual pond layer, a full articulamentum；
Excitation layer is connected behind convolutional layer, a pond layer, institute can be met after continuous two convolutional layers and excitation layer
State last layer of link sort device of depth convolutional neural networks.
More excellent, each convolutional layer forms by 3 × 3 convolution unit.
More excellent, the image collected described in step B is coloured image, and the acquisition mode of image is clapped for industrial camera
Take the photograph, and defect image is inputted into network.
More excellent, the grader is softmax graders.
Beneficial effects of the present invention：
1st, a kind of defect classification and identification algorithm, can be adapted to a variety of handwares and other parts；
2nd, Classification and Identification is carried out using convolutional neural networks and softmax graders, it is not necessary to which other complicated images are pre-
Processing Algorithm；
3rd, average is gone before sample training, training process GPU speed-up computations, the training time is reduced, pixel can be handled
Compare more handware defect images；
4th, identified with depth convolutional neural networks, can increase the number of plies according to actual conditions, extraction defect is more abstracted
Feature, recognition correct rate are higher；
5th, it using Leaky ReLU excitation functions, will not be fitted in training process, calculate easy and effective, fast convergence rate；
6th, using stochastic gradient descent method, amount of calculation is smaller than normal gradients method, and can avoid being absorbed in Local Minimum
Value.
Brief description of the drawings
Fig. 1 is the outline flowchart that traditional shortcoming detects recognizer；
Fig. 2 is the brief illustraton of model that image recognition is carried out using convolutional neural networks；
Fig. 3 is the convolutional neural networks structure chart of identification defect；
Fig. 4 is the convolutional neural networks Two-Dimensional Moment system of battle formations；
Fig. 5 is Leaky ReLU functional arrangements；
Fig. 6 is the max-pooling Two-Dimensional Moment systems of battle formations；
Fig. 7 is training network algorithm flow chart.
Embodiment
Further illustrate technical scheme below in conjunction with the accompanying drawings and by specific embodiment mode.
A kind of handware defect classifying identification method based on depth convolutional neural networks, comprises the following steps：
A. network is established, constructs a depth convolutional neural networks；
B. training network, image the defects of collecting is divided into two major classes, i.e. training set and test set, the training set accounts for
The 70% of total number of images is collected, the test set, which accounts for, collects the 30% of total number of images；
C. defect recognition, the network that the handware image input in test set has been trained, checks output result, will
Recognition result is compareed with the label of image, counts correct recognition rata and error recognition rate.
Further description, training network algorithm comprises the following steps described in step B：
Step 1: first, the weights of network are initialized, and it is 0 weight distribution is submitted to average, variance is
0.01 Gaussian Profile, while number of the weights more than 0 is approximately equal to the number less than 0；
Step 2: training sample goes average, the training sample is colour picture, and each pixel has tri- components of R, G, B
Add and average, when handware sample inputs, three components of sample all pixels are subtracted into average, then input net
Network；
Step 3: network training, the network training uses stochastic gradient descent method.
Further description, the stochastic gradient descent method comprise the following steps：
Training sample is inputted network by step a, propagated forward, the propagated forward one by one, by convolutional layer, excitation layer with
And grader output result of calculation, label is contrasted, calculates output error
Step b, calculation error is fed back, the error according to step a, calculates each layer network from output layer successively forward
Error, according to each layer error, calculate right value update amount, update weight w and deviation b；
Step c, after the square-error for obtaining whole sample trainings, summation, then evolution export overall error as network,
If network overall error is more than given threshold, error and counter are recovered into initial value, re -training sample, until error is less than
Given threshold.
Further description, assume that shared m is to training sample, each training error in the step a
Further description, the expression formula of the overall error are
Further description, the training network process GPU speed-up computations.The parallel speed-up computations of GPU, reduce training
Time, the more handware image of pixel ratio can be handled.
Further description, it is characterised in that：The structure of depth convolutional neural networks described in step A includes six volumes
Lamination, six excitation layers, three pond layers, a full articulamentum；Excitation layer is connected behind convolutional layer, by continuous two convolution
A pond layer, last layer of link sort device of the depth convolutional neural networks can be connect after layer and excitation layer.For knowing
The convolutional neural networks structure chart of other defect is as shown in figure 3, convolutional neural networks mainly have convolutional layer, excitation layer, pond layer group
Into.The excitation layer is connected behind the convolutional layer, a pond layer can be connect after continuous two convolutional layers and excitation layer,
After such structure repeats three times, depth convolutional network is obtained.The pixel light that each neuron of last layer of network obtains
Gated, all pixels that will each obtain are arranged in a row.Then, link sort device, defect classification is identified.
The input of excitation layer neuron is similar with hidden layer input, and hidden layer data x is multiplied with weight w along with deviation b is obtained
Inputted to excitation layer, i.e. y=∑siwixi+b.By y value input stimulus function, Leaky ReLU functions are selected here, such as Fig. 5 institutes
Show, calculate export structure.
The effect of pond layer is the output of simplified convolutional layer.For example, each neuron possibility in the layer of pond will be previous
Neuron summation in one 2 × 2 region of layer.And the max-pooling that another is commonly used, the pond unit is simply
Maximum excitation in the input domain of one 2 × 2 is exported, as shown in fig. 6, this method uses max-pooling methods.
Further description, each convolutional layer form by 3 × 3 convolution unit.Under normal circumstances, nerve net
Input layer represents more vivid straight in convolutional neural networks using a series of neurons come what is represented with two-dimensional matrix in network
See.As conventional neural networks, the neuron of input layer needs to connect with the neuron of hidden layer.But convolutional Neural net
Network is not to be connected each input neuron with each hidden neuron, but only in the regional area of an image
Create connection.So that size is 7X7 image as an example, if a 3X3 of the neuron of first hidden layer and input layer area
Domain connects, as shown in Figure 4.This 3X3 region is just called local sensing domain.9 neurons in the local sensing domain and first
The same neuron connection of individual hidden layer, each connect has a weight, therefore local sensing domain shares 3X3 weight.
If by local sensing domain along from left to right, order from top to bottom is slided, and will obtain god different in corresponding hidden layer
Through member, Fig. 4 show only first neuron of first hidden layer and the connection of input layer.Local sensing domain is to the right
A compensation (being set as 2 here) is slided, just gives input layer data to second hidden neuron.Carry out successively, can be with
Complete transmission of the data from input layer to hidden layer.3X3 neuron in first hidden layer obtained above all uses same
3X3 weight, this is referred to as weight and shares principle.In addition, each shared deviation b of hidden neuron, referred to as common
Enjoy deviation.
Further description, the image collected described in step B are coloured image, and the acquisition mode of image is industry
Camera is shot, and entire image is inputted into network.
Further description, the grader are softmax graders.In one embodiment of the present of invention, described point
Class device is softmax graders.Softmax graders only have two layers, i.e. input layer and output layer, lack compared with neutral net
Hidden layer.According to identify a few class defects, set softmax has several output units.The present invention will identify sand holes, scrape
Wound, lack material, deformation, pit, the class defect of greasy dirt six.So softmax has 6 output units, the corresponding defect of each unit
Classification.After image input, by network operations, each output units of Softmax can export a numerical value, represent input and belong to every
The probability of one classification, it is believed that the defects of maximum of output is then input sample classification.
The technical principle of the present invention is described above in association with specific embodiment.These descriptions are intended merely to explain the present invention's
Principle, and limiting the scope of the invention can not be construed in any way.Based on explanation herein, the technology of this area
Personnel would not require any inventive effort the other embodiments that can associate the present invention, and these modes are fallen within
Within protection scope of the present invention.
Claims (10)
1. a kind of handware defect classifying identification method based on depth convolutional neural networks, it is characterised in that including following step
Suddenly：
A. network is established, constructs a depth convolutional neural networks；
B. training network, the image collected is divided into two major classes, i.e. training set and test set, the training set, which accounts for, collects figure
As the 70% of sum, the test set, which accounts for, collects the 30% of total number of images；
C. defect recognition, the network that the handware defect image input in test set has been trained, checks output result, will
Recognition result is compareed with the label of image, counts correct recognition rata and error recognition rate.
2. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 1, its
It is characterised by：Training network algorithm comprises the following steps described in step B：
Step 1: first, the weights of network are initialized, and it is 0 weight distribution is submitted to average, variance is 0.01
Gaussian Profile, while number of the weights more than 0 is approximately equal to the number less than 0；
Step 2: training sample goes average, the training sample is colour picture, each pixel have tri- components of R, G, B add and,
Average, when handware sample inputs, three components of sample all pixels are subtracted into average, then input network；
Step 3: network training, the network training uses stochastic gradient descent method.
3. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 2, its
It is characterised by：The stochastic gradient descent method comprises the following steps：
Training sample is inputted network by step a, propagated forward, the propagated forward one by one, by convolutional layer, excitation layer and is divided
Class device exports result of calculation, contrasts label, calculates output error
Step b, calculation error is fed back, the error according to step a, calculates the mistake of each layer network from output layer successively forward
Difference, according to each layer error, right value update amount is calculated, updates weight w and deviation b；
Step c, after the square-error for obtaining whole sample trainings, summation, then evolution export overall error as network, if
Network overall error is more than given threshold, error and counter is recovered into initial value, re -training sample, until error is less than setting
Threshold value.
4. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 3, its
It is characterised by：Assume that shared m is to training sample, each training error in the step a
5. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 3, its
It is characterised by：The expression formula of the overall error is
6. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 2, its
It is characterised by：The training network process GPU speed-up computations.
7. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 1, its
It is characterised by：The structure of depth convolutional neural networks described in step A includes six convolutional layers, six excitation layers, three ponds
Layer, a full articulamentum；
Excitation layer is connected behind convolutional layer, a pond layer, the depth can be connect after continuous two convolutional layers and excitation layer
Spend last layer of link sort device of convolutional neural networks.
8. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 7, its
It is characterised by：Each convolutional layer forms by 3 × 3 convolution unit.
9. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 1, its
It is characterised by：The image collected described in step B is coloured image, and the acquisition mode of image shoots for industrial camera, and will
Defect image inputs network.
10. a kind of handware defect classifying identification method based on depth convolutional neural networks according to claim 7, its
It is characterised by：The grader is softmax graders.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710707883.3A CN107481231A (en) | 2017-08-17 | 2017-08-17 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710707883.3A CN107481231A (en) | 2017-08-17 | 2017-08-17 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107481231A true CN107481231A (en) | 2017-12-15 |
Family
ID=60600999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710707883.3A Pending CN107481231A (en) | 2017-08-17 | 2017-08-17 | A kind of handware defect classifying identification method based on depth convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107481231A (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074231A (en) * | 2017-12-18 | 2018-05-25 | 浙江工业大学 | Magnetic sheet surface defect detection method based on convolutional neural network |
| CN108333183A (en) * | 2018-01-31 | 2018-07-27 | 西安工程大学 | A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method |
| CN108389197A (en) * | 2018-02-26 | 2018-08-10 | 上海赛特斯信息科技股份有限公司 | Transmission line of electricity defect inspection method based on deep learning |
| CN108460761A (en) * | 2018-03-12 | 2018-08-28 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| CN108491867A (en) * | 2018-03-12 | 2018-09-04 | 苏州卓融新能源科技有限公司 | Image Matching based on artificial intelligence and recognition methods |
| CN108596922A (en) * | 2018-05-10 | 2018-09-28 | 广东工业大学 | A kind of handware Method of Defect Segmentation and system based on convolutional neural networks |
| CN108629370A (en) * | 2018-04-28 | 2018-10-09 | 广东工业大学 | A kind of classification and identification algorithm and device based on depth confidence network |
| CN108764278A (en) * | 2018-04-16 | 2018-11-06 | 苏州富鑫林光电科技有限公司 | A kind of the self study industrial intelligent detecting system and method for view-based access control model |
| CN108876781A (en) * | 2018-06-26 | 2018-11-23 | 广东工业大学 | Surface defect recognition method based on SSD algorithm |
| CN109472790A (en) * | 2018-11-22 | 2019-03-15 | 南昌航空大学 | A kind of machine components defect inspection method and system |
| CN110314854A (en) * | 2019-06-06 | 2019-10-11 | 苏州市职业大学 | A kind of device and method of the workpiece sensing sorting of view-based access control model robot |
| CN110992361A (en) * | 2019-12-25 | 2020-04-10 | 创新奇智(成都)科技有限公司 | Engine fastener detection system and detection method based on cost balance |
| CN111507972A (en) * | 2020-04-20 | 2020-08-07 | 南京航空航天大学 | Tunnel surface defect detection method combining convolutional neural network and support vector machine |
| CN111699496A (en) * | 2018-03-14 | 2020-09-22 | 欧姆龙株式会社 | Neural network type image processing apparatus |
| CN112396580A (en) * | 2020-11-05 | 2021-02-23 | 北京信息科技大学 | Circular part defect detection method |
| CN112700446A (en) * | 2021-03-23 | 2021-04-23 | 常州微亿智造科技有限公司 | Algorithm model training method and device for industrial quality inspection |
| CN112749747A (en) * | 2021-01-13 | 2021-05-04 | 上海交通大学 | Garbage classification quality evaluation method and system |
| CN113298190A (en) * | 2021-07-05 | 2021-08-24 | 四川大学 | Weld image recognition and classification algorithm based on large-size unbalanced samples |
| TWI745767B (en) * | 2019-10-18 | 2021-11-11 | 汎思數據股份有限公司 | Optical inspection secondary image classification method |
| CN113792852A (en) * | 2021-09-09 | 2021-12-14 | 湖南艾科诺维科技有限公司 | Signal modulation mode identification system and method based on parallel neural network |
| US11507801B2 (en) | 2018-07-27 | 2022-11-22 | Samsung Electronics Co., Ltd. | Method for detecting defects in semiconductor device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160163035A1 (en) * | 2014-12-03 | 2016-06-09 | Kla-Tencor Corporation | Automatic Defect Classification Without Sampling and Feature Selection |
| CN106875381A (en) * | 2017-01-17 | 2017-06-20 | 同济大学 | A kind of phone housing defect inspection method based on deep learning |
-
2017
- 2017-08-17 CN CN201710707883.3A patent/CN107481231A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160163035A1 (en) * | 2014-12-03 | 2016-06-09 | Kla-Tencor Corporation | Automatic Defect Classification Without Sampling and Feature Selection |
| CN106875381A (en) * | 2017-01-17 | 2017-06-20 | 同济大学 | A kind of phone housing defect inspection method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| "《Java数字图像处理：编程技巧与应用实践》" * |
| BIXIWEN_LIU: "神经网络八：权重初始化", 《HTTP://BLOG.CSDN.NET/BINXIWEN_LIU/》 * |
| 邓星: "基于深度学习与SVM的电弧熔积表面缺陷检测与分类", 《万方学位论文数据库》 * |
| 颜伟鑫: "深度学习及其在工件缺陷自动检测中的应用研究", 《中国优秀硕士论文全文数据库》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074231B (en) * | 2017-12-18 | 2020-04-21 | 浙江工业大学 | Magnetic sheet surface defect detection method based on convolutional neural network |
| CN108074231A (en) * | 2017-12-18 | 2018-05-25 | 浙江工业大学 | Magnetic sheet surface defect detection method based on convolutional neural network |
| CN108333183A (en) * | 2018-01-31 | 2018-07-27 | 西安工程大学 | A kind of color based on DCGAN and DCNN knits shirt cut-parts defect inspection method |
| CN108333183B (en) * | 2018-01-31 | 2021-03-16 | 西安工程大学 | Yarn-dyed shirt cut piece defect detection method based on DCGAN and DCNN |
| CN108389197A (en) * | 2018-02-26 | 2018-08-10 | 上海赛特斯信息科技股份有限公司 | Transmission line of electricity defect inspection method based on deep learning |
| CN108389197B (en) * | 2018-02-26 | 2022-02-08 | 上海赛特斯信息科技股份有限公司 | Power transmission line defect detection method based on deep learning |
| CN108460761A (en) * | 2018-03-12 | 2018-08-28 | 北京百度网讯科技有限公司 | Method and apparatus for generating information |
| CN108491867A (en) * | 2018-03-12 | 2018-09-04 | 苏州卓融新能源科技有限公司 | Image Matching based on artificial intelligence and recognition methods |
| CN111699496B (en) * | 2018-03-14 | 2023-08-29 | 欧姆龙株式会社 | Neural network type image processing device, appearance inspection device, and appearance inspection method |
| CN111699496A (en) * | 2018-03-14 | 2020-09-22 | 欧姆龙株式会社 | Neural network type image processing apparatus |
| CN108764278A (en) * | 2018-04-16 | 2018-11-06 | 苏州富鑫林光电科技有限公司 | A kind of the self study industrial intelligent detecting system and method for view-based access control model |
| CN108629370A (en) * | 2018-04-28 | 2018-10-09 | 广东工业大学 | A kind of classification and identification algorithm and device based on depth confidence network |
| CN108629370B (en) * | 2018-04-28 | 2022-05-10 | 广东工业大学 | Classification recognition algorithm and device based on deep belief network |
| CN108596922A (en) * | 2018-05-10 | 2018-09-28 | 广东工业大学 | A kind of handware Method of Defect Segmentation and system based on convolutional neural networks |
| CN108876781A (en) * | 2018-06-26 | 2018-11-23 | 广东工业大学 | Surface defect recognition method based on SSD algorithm |
| US11507801B2 (en) | 2018-07-27 | 2022-11-22 | Samsung Electronics Co., Ltd. | Method for detecting defects in semiconductor device |
| CN109472790A (en) * | 2018-11-22 | 2019-03-15 | 南昌航空大学 | A kind of machine components defect inspection method and system |
| CN110314854A (en) * | 2019-06-06 | 2019-10-11 | 苏州市职业大学 | A kind of device and method of the workpiece sensing sorting of view-based access control model robot |
| TWI745767B (en) * | 2019-10-18 | 2021-11-11 | 汎思數據股份有限公司 | Optical inspection secondary image classification method |
| CN110992361A (en) * | 2019-12-25 | 2020-04-10 | 创新奇智(成都)科技有限公司 | Engine fastener detection system and detection method based on cost balance |
| CN111507972A (en) * | 2020-04-20 | 2020-08-07 | 南京航空航天大学 | Tunnel surface defect detection method combining convolutional neural network and support vector machine |
| CN112396580A (en) * | 2020-11-05 | 2021-02-23 | 北京信息科技大学 | Circular part defect detection method |
| CN112396580B (en) * | 2020-11-05 | 2024-02-02 | 北京信息科技大学 | Method for detecting defects of round part |
| CN112749747B (en) * | 2021-01-13 | 2022-11-11 | 上海交通大学 | Garbage classification quality evaluation method and system |
| CN112749747A (en) * | 2021-01-13 | 2021-05-04 | 上海交通大学 | Garbage classification quality evaluation method and system |
| CN112700446A (en) * | 2021-03-23 | 2021-04-23 | 常州微亿智造科技有限公司 | Algorithm model training method and device for industrial quality inspection |
| CN113298190A (en) * | 2021-07-05 | 2021-08-24 | 四川大学 | Weld image recognition and classification algorithm based on large-size unbalanced samples |
| CN113792852A (en) * | 2021-09-09 | 2021-12-14 | 湖南艾科诺维科技有限公司 | Signal modulation mode identification system and method based on parallel neural network |
| CN113792852B (en) * | 2021-09-09 | 2024-03-19 | 湖南艾科诺维科技有限公司 | Signal modulation mode identification system and method based on parallel neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107481231A (en) | A kind of handware defect classifying identification method based on depth convolutional neural networks | |
| Kumar et al. | Resnet-based approach for detection and classification of plant leaf diseases | |
| CN104281853B (en) | A kind of Activity recognition method based on 3D convolutional neural networks | |
| CN106446942A (en) | Crop disease identification method based on incremental learning | |
| CN107977671A (en) | A kind of tongue picture sorting technique based on multitask convolutional neural networks | |
| Ranjan et al. | Detection and classification of leaf disease using artificial neural network | |
| CN107437092A (en) | The sorting algorithm of retina OCT image based on Three dimensional convolution neutral net | |
| CN108876781A (en) | Surface defect recognition method based on SSD algorithm | |
| CN109508655A (en) | The SAR target identification method of incomplete training set based on twin network | |
| CN110807760B (en) | Tobacco leaf grading method and system | |
| CN107590489A (en) | Object detection method based on concatenated convolutional neutral net | |
| CN107016405A (en) | A kind of insect image classification method based on classification prediction convolutional neural networks | |
| Tete et al. | Plant Disease Detection Using Different Algorithms. | |
| CN106920243A (en) | The ceramic material part method for sequence image segmentation of improved full convolutional neural networks | |
| CN105354581B (en) | The color image feature extracting method of Fusion of Color feature and convolutional neural networks | |
| CN106203395A (en) | Face character recognition methods based on the study of the multitask degree of depth | |
| CN108830285A (en) | A kind of object detection method of the reinforcement study based on Faster-RCNN | |
| CN106874929B (en) | Pearl classification method based on deep learning | |
| CN111222519B (en) | Construction method, method and device of hierarchical colored drawing manuscript line extraction model | |
| CN106951836A (en) | Crop cover degree extracting method based on priori threshold optimization convolutional neural networks | |
| CN109214298A (en) | A kind of Asia women face value Rating Model method based on depth convolutional network | |
| CN104484658A (en) | Face gender recognition method and device based on multi-channel convolution neural network | |
| CN115841447A (en) | Detection method for surface defects of magnetic shoe | |
| CN112819096B (en) | Construction method of fossil image classification model based on composite convolutional neural network | |
| Chen et al. | Cell nuclei detection and segmentation for computational pathology using deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171215 |
|
| RJ01 | Rejection of invention patent application after publication |