CN108022231A - A kind of inside workpiece defect identification method based on firefly neutral net - Google Patents
A kind of inside workpiece defect identification method based on firefly neutral net Download PDFInfo
- Publication number
- CN108022231A CN108022231A CN201610927591.6A CN201610927591A CN108022231A CN 108022231 A CN108022231 A CN 108022231A CN 201610927591 A CN201610927591 A CN 201610927591A CN 108022231 A CN108022231 A CN 108022231A
- Authority
- CN
- China
- Prior art keywords
- firefly
- rbf neural
- value
- inside workpiece
- workpiece defect
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- 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
-
- 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
Abstract
The invention discloses a kind of inside workpiece defect identification method based on firefly neutral net, suitable for the intelligent recognition of inside workpiece defect on workpiece CT sectioning images.Artificial detection may replace to inside workpiece defect recognition using RBF neural method, improve the accuracy rate of defects detection and the efficiency of work piece production and the degree of automation.Include the following steps：1) 14 characteristic values of shape facility and gray feature are extracted by Files from Industrial CT Slicing；2) topological structure and parameter of RBF neural are initialized；3) weights of RBF neural, threshold value are optimized with glowworm swarm algorithm；4) quantity of state representated by optimal firefly is fed forward to RBF neural, determines the initial weight and threshold value of RBF neural, with sample set Training RBF Neural Network.
Description
Technical field
The present invention relates to machine vision, image procossing, area of pattern recognition, more particularly to one kind to be based on firefly nerve net
The inside workpiece defect recognition technology of network.
Background technology
Restricted by production specifications, often there are the defects of crackle, stomata, slag inclusion, these are lacked workpiece interiors of products
Cast properties can be influenced and machine is safe to use by falling into.It is existing at present much to can be used in casting defect in order to ensure casting quality
The detection means in portion, such as ultrasound, microwave, infrared (scanning and imaging), laser hologram and acoustical holography routine and X-ray radiography,
Unconventional lossless detection method such as industry CT.Industrial computer tomography (industry CT) technology is a kind of excellent non-invades
Enter formula imaging technique, at present, which has been widely used for machinery, railway, space flight, aviation, national defence, military project, material, geology
Etc. many key areas.
Inside workpiece defect detecting system is instead of the new of the computer vision technique of human eye defect recognition with machine vision
Using, be certain structure by the character representation of inside workpiece defect using image recognition technology, and with the model structure of storage
Match, the species of defect in image is identified by successful match.The distribution that artificial neural network has information stores, parallel
The features such as processing and self-learning capability, it is possible to achieve the intelligent recognition of inside workpiece defect.Radial basis function (RBF) nerve net
The composition of network only has three layers and only one hidden layers, has simple in structure, and arithmetic speed is fast and local functions approach
Characteristic, the effect of its hidden layer is that input variable is mapped to hidden layer space up, so the structural object of whole network
It is to be determined by the network parameter of hidden layer to determine.Under normal circumstances, the hidden layer neuron node of neutral net is got over
It is more, just there are stronger operational capability, more preferably mapping ability and Function approximation capabilities, while also imply that implicit sheaf space
Dimension is higher, but in actual application, the performance indicator and hidden layer space dimensionality of network have close contact, by
It is too many in the hidden layer neuron interstitial content of neutral net, cause hidden layer space dimensionality excessive, and influence neutral net
Generalization ability.
Glowworm swarm algorithm, which comes from, seeks a spouse fire fly luminescence in nature, and the research of the behavior such as communication, it is a kind of groups
Simulating biology intelligent algorithm, its basic principle are to attract companion or prey, light using the fluorescein induction fire fly luminescence of firefly
Bright more strong more attractive, fluorescence quality is also higher, and firefly is moved to the highest firefly position of fluorescein value；Fluorescein value
Corresponding fitness function value, therefore firefly is determined by finding the position of highest fluorescein firefly in dynamic decision domain
The optimal value of fitness function.It has the ability asked for global extremum well and search for more extreme values, in more extreme value letters
All various aspects such as several solution, the localization of signal source are applied, and achieve good effect.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of inside workpiece defect based on firefly neutral net and know
Other method, suitable for the intelligent recognition of inside workpiece defect on workpiece CT sectioning images.Using RBF neural method to workpiece
Internal flaw identification may replace artificial detection, improve the accuracy rate of defects detection and the efficiency of work piece production and automation journey
Degree.
In order to solve the above technical problems, the technical scheme is that：In a kind of workpiece based on firefly neutral net
Portion's defect identification method, includes the following steps：
1) by Hu not methods of bending moment and grey level histogram, extracted from the Files from Industrial CT Slicing of inside workpiece defect
14 characteristic values of shape facility and gray feature；
2) topological structure of RBF neural is initialized, target error, maximum frequency of training the two ginsengs of network are set
Number；Input layer, the interstitial content of output layer are respectively 14 and 3 nodes, and node in hidden layer is identical with the number of nodes of input layer,
The feature vector that 14 characteristic values are inputted as RBF neural, 3 nodes and the inside workpiece defect of output layer are split
Line, stomata and slag inclusion correspond to.
3) weights of RBF neural, threshold value are optimized with glowworm swarm algorithm, when firefly colony optimization algorithm is initial
The N number of firefly of random distribution in D dimension solution spaces, the current location of firefly i is expressed as Xi(t), i=1,2 ..., N；Initially
When every firefly all carry identical luciferin value l0, and there is identical decision-making radius r0, firefly position is all right
Answer a fitness value f (Xi(t)), the size of fluorescein is related with the fitness value of firefly position, and luciferin value is got over
Greatly, position is better, that is, well adapts to angle value；During firefly is moved, if firefly i is in firefly j
Decision region in, and the fluorescein value of firefly j is higher than firefly i, and firefly i is just with certain probability to firefly j
It is mobile, and update the decision-making radius of oneself；This process is circulated, the maximum iteration until reaching setting, at this moment, Suo Youying
Fireworm can be gathered in around optimal firefly individual present position, this optimal firefly represents the optimal solution of problem；
4) quantity of state representated by optimal firefly is fed forward to RBF neural, determines the initial power of RBF neural
Value and threshold value, with sample set Training RBF Neural Network, until reaching frequency of training or target error, then keep net this moment
Network；Otherwise, continue to train.
The invention has the advantages that the shape facility and gray feature number of the invention by obtaining inside workpiece defect
According to establishing RBF neural, the inside workpiece defect on intelligent recognition workpiece CT sectioning images is in crackle, stomata and slag inclusion
Which kind of, utilize the good global optimizing characteristic optimizing neutral net initial weight of glowworm swarm algorithm, threshold value, enhancing nerve
The generalization ability of network, to INDUSTRIAL CT IMAGE identification quick and precisely, is optimized in terms of defect recognition rate, improves defect knowledge
Not rate.It is a kind of efficient, the intelligentized method for realizing workpiece, defect detection automation.Pass through the class of recognition result, that is, defect
Type, assesses the quality of casting, so as to ensure the safe to use of cast properties and machine.
Brief description of the drawings
Fig. 1 is RBF neural network structure figure；
Fig. 2 is the flow chart of the inside workpiece defect identification method based on firefly neutral net；
Fig. 3 is glowworm swarm algorithm flow chart.
In figure：u1、u2、...、u14- input layer；R1、R2、...、R14- hidden layer node；out1、out2、out3- hidden
Containing node layer.
Embodiment
A kind of inside workpiece defect recognition technology based on firefly neutral net, it is related to RBF neural to work
The intelligent recognition of part internal flaw；RBF neural is initially set up, the topological structure of RBF neural is initialized, passes through the light of firefly
Worm algorithm optimizes the weights of RBF neural, threshold value, determines the optimal initial weight of RBF neural and threshold value, uses
Sample set Training RBF Neural Network, until reaching target error.It is comprised the following steps that：
(1) the defects of obtaining inside workpiece defect data.
The implementation data of the present invention are obtained by Files from Industrial CT Slicing, extract shape facility and gray feature to defect into
Row identification.For the gray value of dreg defect apparently higher than crackle and the gray value of stomata, crackle and gas hole defect can be by shape facilities
Distinguish.By Hu, bending moment and grey level histogram do not carry the inside workpiece defect characteristic in Files from Industrial CT Slicing
Take, obtain 14 characteristic values.
Bending moment does not refer to the expression formula of 7 moment invariants constructed using second order and third central moment, this 7 expression formulas
With translation invariance, rotational invariance and constant rate.
For the digital picture of a width N × M, its gray scale is represented with f (x, y), its (p+q) rank moment of the orign MpqIt is defined as：
Place normalization processing is carried out to above-mentioned square, the invariant features of square can be obtained, obtain the digital picture of N × M
Central moment μpqFor：
Wherein, xc、ycFor picture centre coordinate.
7 constructed using second order and third central moment have translation invariance, rotational invariance and constant rate
The expression formula of moment invariants is as follows：
Under discrete state, bending moment can not influenced be subject to scaling, and in order to make it meet scaling consistency, remake following pumping
As：
Above-mentioned formula is the characteristic formula obtained after extending, meets structure translation scaling and rotation invariant, workpiece is lacked
Fall into image and carry out feature extraction, extract the feature vector that 10 shape facility values are inputted as neutral net.
In addition to above-mentioned 10 shape facility values, effectively to distinguish defect kind, we choose aspect ratio and circularity conduct
The feature vector that other 2 shape facility values of image are inputted as neutral net.
Grey level histogram is by all pixels in digital picture, according to the size of gray value, counts its frequency occurred,
It is a kind of statistical nature of image.Gray average and gray variance are chosen as the grey value characteristics value of image as neutral net
The feature vector of input.
Described on end, we choose the feature vector that 14 characteristic values are inputted as neutral net altogether.By all samples
Originally it is divided into training sample set and test sample collection.
(2) RBF neural is established.
The forward direction that the neutral net is made of input layer, a hidden layer (radial direction basic unit) and a linear convergent rate layer
Neutral net.It is actually needed according to the present invention, input layer, the interstitial content of output layer are respectively 14 and 3 nodes, hidden layer
Number of nodes is identical with the number of nodes of input layer, and crackle, stomata and folder can be represented respectively with 001,010,100 by exporting result in theory
Slag, since, all the time there are error, we select scope of the absolute value of the error range of output valve 0.25 in actual operation
Interior numeral is as theoretical output valve.
(3) weights of RBF neural, threshold value are optimized with glowworm swarm algorithm.
Each firefly represents a candidate solution in solution space in glowworm swarm algorithm, by constantly moving, updating position
Put and update decision-making radius, each firefly independently searches for optimal solution in solution space.When firefly search one it is more excellent
Xie Shi, can volatilize certain luciferin, and the performance of solution is better, and luciferin volatilizees more, and the bigger solution of luciferin value is chosen
In possibility it is bigger.With the iteration of algorithm, become larger compared with the luciferin value on good position, algorithm tends to restrain gradually, directly
Reach maximum to iterations, terminate algorithm, find optimal firefly, the quantity of state representated by optimal firefly is fed forward to
RBF neural, determines the initial weight and threshold value of RBF neural.It is comprised the following steps that：
A. the parameters such as ρ, γ, β, s, rs, nt are initialized, wherein, ρ represents the evaporation rate of the luciferin at t-1 moment；γ tables
Show luciferin turnover rate；β is neighborhood change rate；S is moving step length；Rs is the threshold values for controlling firefly sensing range；Nt is control
The threshold value of neighbours firefly quantity processed.
B. firefly colony optimization algorithm is initialized, the N number of firefly of random distribution in D dimension solution spaces, firefly i's is current
Positional representation is Xi(t), i=1,2 ..., N.Every firefly all carries identical luciferin value l when initial0, and there is phase
Same decision-making radius r0,
C. the specific more new formula of luciferin renewal is as follows：
li(t)=(1- ρ) li(t-1)+γf(Xi(t))
In examination, li(t-1) luciferin values of the firefly i at the t-1 moment is represented；f(Xi(t)) represent in t moment firefly i
The fitness value of position.Fitness value is bigger, the brighter display of firefly, and luciferin is higher.Calculate the t times iteration institute
Locate the luciferin value corresponding to the fitness value on position.
D. the neighborhood formula of firefly is as follows：
Ni(t)={ j：||Xj(t)-Xi(t) | | ＜ ri(t)；li(t) ＜ lj(t)}
Ni(t) set of the firefly i in the neighbours firefly of t moment is represented；| | | | represent Euclidean distance；ri(t) represent
In the decision-making radius of t moment firefly i.
It is as follows as the new probability formula of neighbours firefly that firefly i selects firefly j：
Pij(t) bigger, the probability that firefly i is moved toward firefly j directions is bigger.
Location update formula is as follows after firefly movement：
Decision-making radius more new formula is as follows：
ri(t)=min { rs, max [0, ri(t)+β(nt-|Ni(t)|)]}
Wherein, ri(t) be t moment firefly i sensing range, and have 0 ＜ ri(t) ＜ rs；|Ni(t) | represent neighbours' collection
Size.
Each firefly finds firefly of the fluorescein value higher than oneself in certain neighborhood and is moved to it, more
Position and the perception radius of oneself after new movement.
E. an iteration is completed, judges whether iterations has reached the maximum iteration of setting, if satisfied, terminating
Algorithm, records the optimal value searched, if not satisfied, continuing the iterative process of a new round.
(4) neutral net is trained and tested.
Neutral net is trained and tested by the use of above-mentioned data as training sample, study mechanism is established, works as input
During one group of data, by the automatic computing of network, one group of output valve is obtained, determines that this group of output valve (is manually known with desired output
Other workpiece, defect species) identical rate be how many, which is the discrimination of defect.If defect recognition rate, which is more than, specifies essence
Degree, then study terminate.Otherwise, into next group of study, until defect recognition rate within the specified range, export at this time optimal
Weights.
Claims (1)
1. a kind of inside workpiece defect identification method based on firefly neutral net, it is characterised in that include the following steps：
1) by Hu not methods of bending moment and grey level histogram, shape is extracted from the Files from Industrial CT Slicing of inside workpiece defect
14 characteristic values of feature and gray feature；
2) topological structure of RBF neural is initialized, the target error of network, maximum the two parameters of frequency of training are set；
Input layer, the interstitial content of output layer are respectively 14 and 3 nodes, and node in hidden layer is identical with the number of nodes of input layer, institute
State the feature vector that 14 characteristic values are inputted as RBF neural, 3 nodes of output layer and inside workpiece defect crack,
Stomata and slag inclusion correspond to.
3) weights of RBF neural, threshold value are optimized with glowworm swarm algorithm, in D when firefly colony optimization algorithm is initial
The N number of firefly of random distribution in solution space is tieed up, the current location of firefly i is expressed as Xi(t), i=1,2 ..., N；When initial
Every firefly all carries identical luciferin value l0, and there is identical decision-making radius r0, firefly position all corresponds to
One fitness value f (Xi(t)), the size of fluorescein is related with the fitness value of firefly position, and luciferin value is bigger,
Position is better, that is, well adapts to angle value；During firefly is moved, if firefly i determining in firefly j
In the range of plan, and the fluorescein value of firefly j is higher than firefly i, and firefly i is just moved with certain probability to firefly j,
And update the decision-making radius of oneself；This process is circulated, the maximum iteration until reaching setting, at this moment, all fireflies
It can be gathered in around optimal firefly individual present position, which represents the optimal solution of problem；
4) quantity of state representated by optimal firefly is fed forward to RBF neural, determine RBF neural initial weight and
Threshold value, with sample set Training RBF Neural Network, until reaching frequency of training or target error, then keeps network this moment；It is no
Then, continue to train.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610927591.6A CN108022231A (en) | 2016-10-31 | 2016-10-31 | A kind of inside workpiece defect identification method based on firefly neutral net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610927591.6A CN108022231A (en) | 2016-10-31 | 2016-10-31 | A kind of inside workpiece defect identification method based on firefly neutral net |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108022231A true CN108022231A (en) | 2018-05-11 |
Family
ID=62070247
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610927591.6A Pending CN108022231A (en) | 2016-10-31 | 2016-10-31 | A kind of inside workpiece defect identification method based on firefly neutral net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108022231A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109657731A (en) * | 2018-12-28 | 2019-04-19 | 长沙理工大学 | A kind of anti-interference classification method of droplet digital pcr instrument |
| CN109900501A (en) * | 2019-01-31 | 2019-06-18 | 石家庄铁道大学 | High-speed EMUs vehicle artificial intelligence lossless detection method |
| CN112955733A (en) * | 2018-09-19 | 2021-06-11 | 爱德华兹生命科学公司 | Medical implant examination using imaging |
| CN113076895A (en) * | 2021-04-09 | 2021-07-06 | 太原理工大学 | Conveyor belt longitudinal damage vibration sensing method based on infrared computer vision |
| CN114169385A (en) * | 2021-09-28 | 2022-03-11 | 北京工业大学 | MSWI process combustion state identification method based on mixed data enhancement |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080058596A (en) * | 2006-12-22 | 2008-06-26 | 한국생산기술연구원 | Detection method of internal castig defects by using industrial computer tompgraphy |
| CN101216880A (en) * | 2007-12-28 | 2008-07-09 | 中国人民解放军海军航空工程学院 | An initiating explosive device defect identification method |
| CN103018383A (en) * | 2012-12-07 | 2013-04-03 | 四川电力科学研究院 | Oil chromatogram on-line monitoring noise data correction method |
| KR101361922B1 (en) * | 2012-10-19 | 2014-02-21 | 서울여자대학교 산학협력단 | Automatic detection system of porosity defect using density and shape information in industrial ct and controlling method therefor |
| CN105550711A (en) * | 2015-12-16 | 2016-05-04 | 合肥工业大学 | Firefly algorithm based selective ensemble learning method |
| CN105808845A (en) * | 2016-03-08 | 2016-07-27 | 中山大学 | Structural damage identification method based on glowworm group intelligent algorithm |
| CN105891215A (en) * | 2016-03-31 | 2016-08-24 | 浙江工业大学 | Welding visual detection method and device based on convolutional neural network |
| CN105976051A (en) * | 2016-04-29 | 2016-09-28 | 武汉大学 | Wavelet transformation and improved firefly-optimized extreme learning machine-based short-term load prediction method |
-
2016
- 2016-10-31 CN CN201610927591.6A patent/CN108022231A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080058596A (en) * | 2006-12-22 | 2008-06-26 | 한국생산기술연구원 | Detection method of internal castig defects by using industrial computer tompgraphy |
| CN101216880A (en) * | 2007-12-28 | 2008-07-09 | 中国人民解放军海军航空工程学院 | An initiating explosive device defect identification method |
| KR101361922B1 (en) * | 2012-10-19 | 2014-02-21 | 서울여자대학교 산학협력단 | Automatic detection system of porosity defect using density and shape information in industrial ct and controlling method therefor |
| CN103018383A (en) * | 2012-12-07 | 2013-04-03 | 四川电力科学研究院 | Oil chromatogram on-line monitoring noise data correction method |
| CN105550711A (en) * | 2015-12-16 | 2016-05-04 | 合肥工业大学 | Firefly algorithm based selective ensemble learning method |
| CN105808845A (en) * | 2016-03-08 | 2016-07-27 | 中山大学 | Structural damage identification method based on glowworm group intelligent algorithm |
| CN105891215A (en) * | 2016-03-31 | 2016-08-24 | 浙江工业大学 | Welding visual detection method and device based on convolutional neural network |
| CN105976051A (en) * | 2016-04-29 | 2016-09-28 | 武汉大学 | Wavelet transformation and improved firefly-optimized extreme learning machine-based short-term load prediction method |
Non-Patent Citations (4)
| Title |
|---|
| FEI ZHAO: "Learning-based automatic defect recognition with computed tomographic imaging", 《2013 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 * |
| 侯越: "基于萤火虫优化的BP神经网络算法研究", 《万方数据知识服务平台》 * |
| 彭伟: "基于萤火虫算法的神经网络CPI预测模型", 《万方数据知识服务平台》 * |
| 王婧: "基于神经网络的焊缝缺陷自动识别技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112955733A (en) * | 2018-09-19 | 2021-06-11 | 爱德华兹生命科学公司 | Medical implant examination using imaging |
| CN109657731A (en) * | 2018-12-28 | 2019-04-19 | 长沙理工大学 | A kind of anti-interference classification method of droplet digital pcr instrument |
| CN109900501A (en) * | 2019-01-31 | 2019-06-18 | 石家庄铁道大学 | High-speed EMUs vehicle artificial intelligence lossless detection method |
| CN113076895A (en) * | 2021-04-09 | 2021-07-06 | 太原理工大学 | Conveyor belt longitudinal damage vibration sensing method based on infrared computer vision |
| CN113076895B (en) * | 2021-04-09 | 2022-08-02 | 太原理工大学 | Conveyor belt longitudinal damage vibration sensing method based on infrared computer vision |
| CN114169385A (en) * | 2021-09-28 | 2022-03-11 | 北京工业大学 | MSWI process combustion state identification method based on mixed data enhancement |
| CN114169385B (en) * | 2021-09-28 | 2024-04-09 | 北京工业大学 | MSWI process combustion state identification method based on mixed data enhancement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jin et al. | Weed identification using deep learning and image processing in vegetable plantation | |
| CN108022231A (en) | A kind of inside workpiece defect identification method based on firefly neutral net | |
| Kumar et al. | Resnet-based approach for detection and classification of plant leaf diseases | |
| Lee et al. | Deep-plant: Plant identification with convolutional neural networks | |
| Mohamad et al. | Bacteria identification from microscopic morphology: a survey | |
| CN111127385B (en) | Medical information cross-modal Hash coding learning method based on generative countermeasure network | |
| CN108510467A (en) | SAR image target recognition method based on variable depth shape convolutional neural networks | |
| CN108426994A (en) | Digital holographic microscopy data are analyzed for hematology application | |
| Ahmed et al. | Classification of bacterial contamination using image processing and distributed computing | |
| CN108363970A (en) | A kind of recognition methods of fingerling class and system | |
| CN108537102A (en) | High Resolution SAR image classification method based on sparse features and condition random field | |
| CN108765475B (en) | Building three-dimensional point cloud registration method based on deep learning | |
| CN108363967A (en) | A kind of categorizing system of remote sensing images scene | |
| Chaturvedi et al. | Efficient Method for Tomato Leaf Disease Detection and Classification based on Hybrid Model of CNN and Extreme Learning Machine | |
| Gaur et al. | Comparative assessment of artificial intelligence (AI)-based algorithms for detection of harmful bloom-forming algae: an eco-environmental approach toward sustainability | |
| CN108491923B (en) | Pest image recognition method based on improved wolf colony algorithm optimized Bayesian network | |
| Tzampazaki et al. | Machine Vision—Moving from Industry 4.0 to Industry 5.0 | |
| Liang et al. | Figure-ground image segmentation using genetic programming and feature selection | |
| CN103824279A (en) | Image segmentation method based on organizational evolutionary cluster algorithm | |
| Rasi et al. | YOLO Based Deep Learning Model for Segmenting the Color Images | |
| GB2604093A (en) | System for clustering data points | |
| Kutlu et al. | An Intelligent Software for Measurements of Biological Materials: BioMorph | |
| Kerssies et al. | Neural architecture search for visual anomaly segmentation | |
| Dhindsa et al. | An efficient microbes detection system using microscopic images via morphological and correlation based features | |
| Wikarsa et al. | Detection of the types of consumable saltwater fish in the coastal area of likupang: Uses the convolutional neural network method |
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: 20180511 |