CN108053024A - Key tooth flower recognition methods - Google Patents
Key tooth flower recognition methods Download PDFInfo
- Publication number
- CN108053024A CN108053024A CN201711141996.8A CN201711141996A CN108053024A CN 108053024 A CN108053024 A CN 108053024A CN 201711141996 A CN201711141996 A CN 201711141996A CN 108053024 A CN108053024 A CN 108053024A
- Authority
- CN
- China
- Prior art keywords
- layer
- output
- hidden layer
- key tooth
- perceive
- 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
- 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
Abstract
A kind of key tooth flower recognition methods, including perceiving step and training learning procedure.By perceiving step and training learning procedure, it can obtain definite sensor model using machine vision as technological means after the interest-bearing of substantial amounts of learning training, by Model Matching or comparison, identify that key tooth flower corresponds to tooth form.The key tooth flower recognition methods of the present invention can quickly and accurately identify key tooth flower, and identification process completion is automatically performed by computer, manual intervention be not required, and performance is stablized.
Description
Technical field
The present invention relates to a kind of key tooth flower identification technology field, more particularly to a kind of key tooth flower recognition methods.
Background technology
Lockset manufacturing is one of important industry of national economy, the huge market capacity of lockset and cumulative year after year
Demand proposes very high requirement to the automation processing of lockmaking industry and packaging efficiency.Traditional processing technology, manufacture
Equipment and the mode of production are difficult to meet the needs of market.
Key tooth flower identification is one of critical process in lockset assembling, directly affects the efficiency and non-defective unit of lockset assembling
Rate.At present, the method generally used is manual identified and sensor mechanism identification.Manual identified method efficiency is low, poor reliability,
The requirement of industrialized mass production can not be met；Machinery identification has that accuracy of identification is not high, the degree of automation is low, has mechanical loss etc.
Drawback.
The content of the invention
The purpose of the present invention is overcoming shortcoming of the prior art, a kind of key tooth flower recognition methods is provided, it can
Solve the problem of that traditional keys tartar manual identified efficiency is low and mechanical accuracy of identification is unstable, the present invention proposes a kind of base
In the intelligent identification Method of Multilayer Perception study；The invention is using machine vision as technological means, using deep learning algorithm, quickly
And the tartar of key is identified exactly.
The purpose of the present invention is what is be achieved through the following technical solutions：
A kind of key tooth flower recognition methods,
S110：Perceive step：
Defining input layer has n0 to perceive member, and first hidden layer has n1 to perceive member, and second hidden layer has n2 sense
Know member, output layer has n3 to perceive member；Definition W is weight；
Default mathematical model is established, as shown in Figure 1
Establish default input layer functional relation：
First hidden layer：By W(1)It is connected entirely with input layer；
Second hidden layer：By W(2)It is connected entirely with the first hidden layer；
Sigmoid functions are chosen as activation primitive：It obtains
x(1)=f (s(1))=f (W(1)x(0)) (1)
x(2)=f (s(2))=f (W(2)x(1)) (2)
In output layer：By W(3)It is connected to obtain entirely with the second hidden layer：
x(3)=f (s(3))=f (W(3)x(2)) (3)
Multilayer perceptron is according to connectionless, the mode that adjacent two layers neuron connects entirely between in layer, from input to output
The characteristics of forward calculation transferred successively, is propagated
S120：Training learning procedure：
Assuming that q-th of sample is added in above-mentioned sensor model, its performance indicator can be defined, which measures
Error condition between desired output and the output of the model of reality：
It is desired output to define dq, and yq is actual model output
For output layer, the gradient of weight can be directly obtained by formula：
For hidden layer, it can equally be obtained by formula (5)：
It can be obtained by chain type Rule for derivation again：
For the convenience in expression formula, Ke Yiling：
So by formula (7) and expression above, can obtain
Define δj (L)(L represents the number of plies of output layer) is the local error represented on output layer, and δj (l+1)(l=1,2,
3 ...) local error in each hidden layer is then represented.
In a wherein embodiment, in output layer L, by δj (L)=(f (sj (L))-yj (L))f′(sj (L)) can be obtained, root
According to δj (L)=(f (sj (L))-yj (L))f′(sj (L)) other δ j (l) can be solved.
Compared with prior art, the present invention has the following advantages：
The key tooth flower recognition methods of the present invention can quickly and accurately identify key tooth flower, and identification process is complete
It is automatically performed by computer, manual intervention is not required, and performance is stablized.
Description of the drawings
Fig. 1 is Multilayer Perception illustraton of model.
Specific embodiment
For the ease of understanding the present invention, the present invention is described more fully below with reference to relevant drawings.In attached drawing
Give the better embodiment of the present invention.But the present invention can realize in many different forms, however it is not limited to herein
Described embodiment.On the contrary, the purpose of providing these embodiments is that make to understand more the disclosure
Add thorough and comprehensive.
It should be noted that when element is referred to as " being fixed on " another element, it can be directly on another element
Or there may also be elements placed in the middle.When an element is considered as " connection " another element, it can be directly connected to
To another element or it may be simultaneously present centering elements.Term as used herein " vertical ", " horizontal ", " left side ",
For illustrative purposes only, it is unique embodiment to be not offered as " right side " and similar statement.
Unless otherwise defined, all of technologies and scientific terms used here by the article is with belonging to technical field of the invention
The normally understood meaning of technical staff is identical.Term used in the description of the invention herein is intended merely to description tool
The purpose of the embodiment of body, it is not intended that in the limitation present invention.Term as used herein " and/or " include one or more
The arbitrary and all combination of relevant Listed Items.
As shown in Figure 1, it is Multilayer Perception illustraton of model.
1st, a kind of key tooth flower recognition methods, including perceiving step and training learning procedure.
S110：Perceive step：
Defining input layer has n0 to perceive member, and first hidden layer has n1 to perceive member, and second hidden layer has n2 sense
Know member, output layer has n3 to perceive member；Definition W is weight；
Establish default mathematical model：
Establish default input layer functional relation：
First hidden layer：By W(1)It is connected entirely with input layer；
Second hidden layer：By W(2)It is connected entirely with the first hidden layer；
Sigmoid functions are chosen as activation primitive：It obtains
x(1)=f (s(1))=f (W(1)x(0)) (1)
x(2)=f (s(2))=f (W(2)x(1)) (2)
In output layer：By W(3)It is connected to obtain entirely with the second hidden layer：
x(3)=f (s(3))=f (W(3)x(2)) (3)
Multilayer perceptron is according to connectionless, the mode that adjacent two layers neuron connects entirely between in layer, from input to output
The characteristics of forward calculation transferred successively, is propagated
S120：Training learning procedure：
Assuming that q-th of sample is added in above-mentioned sensor model, its performance indicator can be defined, which measures
Error condition between desired output and the output of the model of reality：
It is desired output to define dq, and yq is actual model output
For output layer, the gradient of weight can be directly obtained by formula：
For hidden layer, it can equally be obtained by formula (5)：
It can be obtained by chain type Rule for derivation again：
For the convenience in expression formula, Ke Yiling：
So by formula (7) and expression above, can obtain
Without loss of generality, formula (8) can also be generalized in the model structure of finite layer, and the formula is equally applicable, wherein, definition
δj (L)(L represents the number of plies of output layer) is the local error represented on output layer, and δj (l+1)(l=1,2,3 ...) is then represented
The local error of each hidden layer.
Further, in output layer L, by δj (L)=(f (sj (L))-yj (L))f′(sj (L)) can be obtained, according to δj (L)=(f
(sj (L))-yj (L))f′(sj (L)) other δ j (l) can be solved.
Under normal circumstances, due in output layer L, by δj (L)=(f (sj (L))-yj (L))f′(sj (L)) can directly be obtained,
So other δ can be solved by this formulaj (l), the gradient of each weight can be further calculated, then
According to gradient decline algorithm carry out sensor model error reverse train learning algorithm, so as to according to the requirement of identification with
Precision updates the requirement that weight reaches particular problem.
A kind of reverse train learning method, comprises the following steps：
S1, collecting sample data (x, y) are required according to practical problem,
S2, the weighting w and S that every layer of each neuron is solved in the way of the propagated forward information of sensor model,
S3, calculating solve the δ (L) in output layer；
S4, by formula (8) calculate others δ j (l), and further must each weight gradient；
S5, a suitable Study rate parameter μ is chosen, to carry out the adjustment of weight or update：
In conclusion by this error reverse train learning algorithm, sensor model (i.e. default mathematical model) can guarantee
Performance error Eq minimize, it is hereby achieved that meet key tooth flower identification classificating requirement optimization weight wij (l).
In one embodiment, when key tooth flower recognition methods using the present invention is operated, first, by industrial camera
(CCD) take pictures to key tooth flower, computer is uploaded to by image pick-up card, computer passes through image acquisition bit information
(x, y) establishes sensor model as shown in Figure 1；Then, constantly learn to update corresponding sensor model parameter by image information；
Finally, definite sensor model is obtained after substantial amounts of learning training, by Model Matching or comparison, identifies key tooth flower
Corresponding tooth form.
Compared with prior art, the present invention has the following advantages：
The key tooth flower recognition methods of the present invention can quickly and accurately identify key tooth flower, and identification process is complete
It is automatically performed by computer, manual intervention is not required, and performance is stablized.
Embodiment described above only expresses the several embodiments of the present invention, and description is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that come for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the protection of the present invention
Scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (2)
1. a kind of key tooth flower recognition methods, which is characterized in that
S110：Perceive step：
Defining input layer has n0 to perceive member, and first hidden layer has n1 to perceive member, and second hidden layer has n2 to perceive member,
Output layer has n3 to perceive member；Definition W is weight；
Establish default mathematical model；
Establish default input layer functional relation：
First hidden layer：By W(1)It is connected entirely with input layer；
Second hidden layer：By W(2)It is connected entirely with the first hidden layer；
Sigmoid functions are chosen as activation primitive：It obtains
x(1)=f (s(1))=f (W(1)x(0)) (1)
x(2)=f (s(2))=f (W(2)x(1)) (2)
In output layer：By W(3)It is connected to obtain entirely with the second hidden layer：
x(3)=f (s(3))=f (W(3)x(2)) (3)
Multilayer perceptron is according to connectionless, the mode that adjacent two layers neuron connects entirely between in layer, from input to output successively
The characteristics of forward calculation of transmission, is propagated
S120：Training learning procedure：
Assuming that q-th of sample is added in above-mentioned sensor model, its performance indicator can be defined, which has measured expectation
Error condition between the output of the model of output and reality：
It is desired output to define dq, and yq is actual model output
For output layer, the gradient of weight can be directly obtained by formula：
For hidden layer, it can equally be obtained by formula (5)：
It can be obtained by chain type Rule for derivation again：
For the convenience in expression formula, Ke Yiling：
So by formula (7) and expression above, can obtain
Define δj (L)(L represents the number of plies of output layer) is the local error represented on output layer, and δj (l+1)(l=1,2,3 ...)
Then represent the local error in each hidden layer.
2. key tooth flower recognition methods according to claim 1, which is characterized in that in output layer L, by δj (L)=(f (sj (L))-yj (L))f′(sj (L)) can be obtained, according to δj (L)=(f (sj (L))-yj (L))f′(sj (L)) other δ j (l) can be solved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711141996.8A CN108053024A (en) | 2017-11-17 | 2017-11-17 | Key tooth flower recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711141996.8A CN108053024A (en) | 2017-11-17 | 2017-11-17 | Key tooth flower recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108053024A true CN108053024A (en) | 2018-05-18 |
Family
ID=62118890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711141996.8A Pending CN108053024A (en) | 2017-11-17 | 2017-11-17 | Key tooth flower recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108053024A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020259257A1 (en) * | 2019-06-28 | 2020-12-30 | 深圳数马电子技术有限公司 | Article identification method and device, and computer readable storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001052174A8 (en) * | 2000-01-11 | 2003-01-03 | Mnogoprofilnoe Predpr Ooo Elsy | Method, device and system for biometric identification |
| CN103712567A (en) * | 2013-05-27 | 2014-04-09 | 北京理工大学 | Small module arc gear non-contact optical fine measurement method |
| CN104792522A (en) * | 2015-04-10 | 2015-07-22 | 北京工业大学 | Intelligent gear defect analysis method based on fractional wavelet transform and BP neutral network |
| CN105631445A (en) * | 2014-11-06 | 2016-06-01 | 通号通信信息集团有限公司 | Character recognition method and system for license plate with Chinese characters |
| CN105760883A (en) * | 2016-02-15 | 2016-07-13 | 西安科技大学 | Belt conveyer key component automatic identification method based on infrared thermography |
| CN105938563A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Weld surface defect identification method based on image texture |
| CN106096619A (en) * | 2016-06-21 | 2016-11-09 | 青岛译锁堂安全技术有限公司 | Based on artificial intelligence technology join spoon method and system |
| CN107256009A (en) * | 2017-06-30 | 2017-10-17 | 武汉理工大学 | A kind of Digital product model Intelligent assembly system based on deep learning |
| CN107316054A (en) * | 2017-05-26 | 2017-11-03 | 昆山遥矽微电子科技有限公司 | Non-standard character recognition methods based on convolutional neural networks and SVMs |
-
2017
- 2017-11-17 CN CN201711141996.8A patent/CN108053024A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001052174A8 (en) * | 2000-01-11 | 2003-01-03 | Mnogoprofilnoe Predpr Ooo Elsy | Method, device and system for biometric identification |
| CN103712567A (en) * | 2013-05-27 | 2014-04-09 | 北京理工大学 | Small module arc gear non-contact optical fine measurement method |
| CN105631445A (en) * | 2014-11-06 | 2016-06-01 | 通号通信信息集团有限公司 | Character recognition method and system for license plate with Chinese characters |
| CN104792522A (en) * | 2015-04-10 | 2015-07-22 | 北京工业大学 | Intelligent gear defect analysis method based on fractional wavelet transform and BP neutral network |
| CN105760883A (en) * | 2016-02-15 | 2016-07-13 | 西安科技大学 | Belt conveyer key component automatic identification method based on infrared thermography |
| CN105938563A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Weld surface defect identification method based on image texture |
| CN106096619A (en) * | 2016-06-21 | 2016-11-09 | 青岛译锁堂安全技术有限公司 | Based on artificial intelligence technology join spoon method and system |
| CN107316054A (en) * | 2017-05-26 | 2017-11-03 | 昆山遥矽微电子科技有限公司 | Non-standard character recognition methods based on convolutional neural networks and SVMs |
| CN107256009A (en) * | 2017-06-30 | 2017-10-17 | 武汉理工大学 | A kind of Digital product model Intelligent assembly system based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| D. BLEH等: "A 100 GHz FMCW MIMO radar system for 3D image reconstruction", 《2016 EUROPEAN RADAR CONFERENCE (EURAD)》 * |
| QQ66688857: "多层感知器", 《道客巴巴：HTTPS://WWW.DOC88.COM/P-7039757945070.HTML?R=1》 * |
| 杨竹: "基于DSP的钥匙齿形码识别系统的设计", 《中国优秀硕士学位论文全文数据库·工程科技Ⅰ辑》 * |
| 韩明红等: "多层前馈神经网络研究——单位BP算法", 《振动.测试与诊断》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020259257A1 (en) * | 2019-06-28 | 2020-12-30 | 深圳数马电子技术有限公司 | Article identification method and device, and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022037068A1 (en) | Method for diagnosis of fault in machine tool bearing | |
| JP6616170B2 (en) | Machine learning device, laminated core manufacturing apparatus, laminated core manufacturing system, and machine learning method for learning stacking operation of core sheet | |
| CN109766921B (en) | Vibration data fault classification method based on depth field self-adaption | |
| CN106022380A (en) | Individual identity identification method based on deep learning | |
| CN107492098B (en) | It is a kind of based on PCA and CNN high-temperature forging surface defect in position detecting method | |
| CN110135459B (en) | Zero sample classification method based on double-triple depth measurement learning network | |
| CN105760933A (en) | Method and apparatus for fixed-pointing layer-wise variable precision in convolutional neural network | |
| CN106503802A (en) | A kind of method of utilization genetic algorithm optimization BP neural network system | |
| CN105651457B (en) | Multi-dimension force sensor calibration experiment data fitting method based on genetic algorithm | |
| EP3537346A1 (en) | Method and system for incorporating regression into stacked auto encoder (sae) | |
| CN107390644A (en) | Manufacture adjustment system | |
| CN106408120B (en) | Local area landslide prediction device and method | |
| CN112001903A (en) | Defect detection network construction method, abnormality detection method and system, and storage medium | |
| CN104317269A (en) | Comprehensive forecast and iterative learning control method based on 2D theory | |
| CN109387521A (en) | Image processing system | |
| CN114548368A (en) | Modeling method, prediction model and prediction method of lithium battery temperature field prediction model based on multilayer nuclear overrun learning machine | |
| CN109711530A (en) | A kind of Prediction of Landslide and system | |
| Moisan et al. | Fundamental matrix of a stereo pair, with a contrario elimination of outliers | |
| CN113297994A (en) | Pilot behavior analysis method and system | |
| CN108053024A (en) | Key tooth flower recognition methods | |
| CN116028509A (en) | Optimal working condition index generation method based on causal mechanism | |
| CN110543656A (en) | LED fluorescent powder glue coating thickness prediction method based on deep learning | |
| CN116740567A (en) | Soil moisture content detection method and system for Paulownia seedling cultivation | |
| CN114332176A (en) | End-to-end three-dimensional point cloud registration method | |
| CN115206455A (en) | Deep neural network-based rare earth element component content prediction method and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180518 |