CN105740950B - The template matching method of neural network based on slip teeth method - Google Patents
The template matching method of neural network based on slip teeth method Download PDFInfo
- Publication number
- CN105740950B CN105740950B CN201610035536.6A CN201610035536A CN105740950B CN 105740950 B CN105740950 B CN 105740950B CN 201610035536 A CN201610035536 A CN 201610035536A CN 105740950 B CN105740950 B CN 105740950B
- Authority
- CN
- China
- Prior art keywords
- error
- data
- template matching
- error range
- neural network
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000001514 detection method Methods 0.000 claims abstract description 8
- 230000007246 mechanism Effects 0.000 claims abstract description 3
- 238000009472 formulation Methods 0.000 claims abstract 2
- 239000000203 mixture Substances 0.000 claims abstract 2
- 230000008569 process Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000012937 correction Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Complex Calculations (AREA)
- Image Analysis (AREA)
Abstract
本发明公开一种基于滑齿法的神经网络的模板匹配方法，包括以下步骤：根据误差反向传播的神经网络的制定规则，将网络结构分为输入层、隐藏层和输出层；在隐藏层中设定第一误差范围、最大训练次数和第二误差范围并初始化；根据数据分块机制，输入的数据进行相似性检测后被分割成多个数据块；利用滑齿法匹配被处理的数据；判断网络节点的误差和模板匹配的误差是否分别落在第一误差范围和第二误差范围内，或者，模板匹配的误差是否在第二误差范围内且达到最大训练次数；如是，则输出结果；否则修正滑齿的权值，重复执行以上步骤，直到输出结果。本发明进一步改善了模板匹配精度，提升了运行时间以及算法稳定性。
The invention discloses a neural network template matching method based on a tooth sliding method, which comprises the following steps: dividing the network structure into an input layer, a hidden layer and an output layer according to the formulation rules of the neural network for error back propagation; Set the first error range, the maximum number of training times, and the second error range and initialize them; according to the data block mechanism, the input data is divided into multiple data blocks after similarity detection; the sliding gear method is used to match the processed data ; Judging whether the error of the network node and the error of template matching fall within the first error range and the second error range respectively, or whether the error of template matching is within the second error range and reaches the maximum number of training times; if so, output the result ; otherwise, correct the weight of the sliding tooth, and repeat the above steps until the result is output. The invention further improves the template matching accuracy, improves the running time and the algorithm stability.
Description
Technical field
The present invention relates to big data processing technology fields, in particular to the template matching side of the neural network based on slip teeth method
Method.
Background technique
The fast development of the emerging technologies such as cloud computing, Internet of Things promotes the scale of data just to increase at an unprecedented rate
Long, big data era has begun arrival.But in the data of the complicated extreme redundancy of magnanimity, it is difficult to quickly timely and effective
Obtain valuable information.Therefore be required to fast and accurately find valuable information, extract effective information and by these
Valuable information is organized, and an effective solution scheme is obtained.Template matching side neural network based in big data
Method is being analyzed data and classified by intelligence, according to the information model that known demand setting needs, then utilizes this
A algorithm quickly carries out the template matching of data, quick obtaining effective information to mass data.However, current major part template
It is that matching search is carried out according to the sequence of event importance with algorithm, since node needs to record multiple identical events, thus
The consistency for keeping sequence causes node to need to be repeated several times the similar information of processing, and calculates and brought greatly to caching
Pressure leads to not be applied to the higher occasion of real-time.
Summary of the invention
Place, the present invention propose a kind of template of neural network based on slip teeth method in view of above-mentioned deficiencies of the prior art
Method of completing the square, using the modified slip teeth method error back propagation of weight neural network template matching algorithm to the data of input into
Row classification, study and template data matching.
The template matching method of neural network based on slip teeth method, comprising the following steps:
Network structure is divided into input layer, hidden by step 1, laying down a regulation according to the neural network of error back propagation
Layer and output layer；
Step 2 sets first error range, maximum frequency of training and the second error range and initial in the hidden layer
Change, the first error range is the acceptable error range of the neural network, and second error range is template matching
It is worth acceptable error range；
Step 3, according to deblocking mechanism, the data of input are divided into multiple data blocks after carrying out similitude detection；
Step 4 matches processed data using slip teeth method；
Whether the error of step 5, the error for judging network node and template matching respectively falls in the first error range
In second error range, alternatively, whether the error of template matching in second error range and reaches maximum instruction
Practice number；If so, then exporting result；It is no to then follow the steps 6；
The network weight of step 6, the amendment input layer；
Step 7 repeats step 4~step 6, until exporting result.
The detailed process of step 3 are as follows: utilize Similarity Detection Algorithm, calculate respective point in input node data and template
Absolute error value adds up the absolute error value of each input data point, until being more than given threshold, stops cumulative and calculates this number
According to the similitude mean value of block, the similitude mean value is greater than desired value at this time, stops segmentation data.
The detailed process of step 4 are as follows: in the network training stage, output signal obtains error letter compared with desired output
Number, the error signal is fed back step by step from the output layer to the hidden layer and the input layer, and hidden described in dynamic corrections
Hide the weight of layer and the input layer.
The present invention is detected by carrying out the similitude of data to mass data, determines the size of the data block of segmentation；It utilizes
The template matching algorithm of the neural network of the slip teeth method error back propagation of modified weight is classified to the data of input, is learnt
It is matched with template data；Using mean absolute difference parameter determining data matching degree, the data information of template matching is obtained, to add
Algorithm speed and precision in fast matching process, and improve the stability of algorithm.Template matching precision is further improved, is promoted
Runing time and algorithm stability.
Detailed description of the invention
Fig. 1 is the network model of the template matching of the neural network based on slip teeth method；
Fig. 2 is the flow chart of the template matching method embodiment of the neural network based on slip teeth method；
Fig. 3 is Fig. 2 embodiment threshold value and convergence number relational graph；
Fig. 4 is Fig. 2 embodiment connection weight and convergence number relational graph；
Fig. 5 is that Fig. 2 embodiment learns index and convergence number relational graph；
Fig. 6 is the relationship that Fig. 2 embodiment restrains number and the number of iterations.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Template matching method neural network based in big data carries out first to needing matched data to be split
Modularized processing is conducive to the speed for accelerating to need matched template data information in this way, utilizes sequential similarity detection algorithm
(Sequential Similarity Detection Algorithm, SSDA) carries out similitude detection such as to the data of input
Formula 1.The absolute error value for calculating respective point in the data and template of input, the error of the data point of input is added up.It is tired every time
If the error added is more than given threshold, stops adding up, the similitude mean value E value of this data block is calculated, if similitude mean value E
Value is greater than the desired value of setting, then terminates and divide this data block；Before similitude mean value E is less than desired value, by the part
Data segment is set as a small data block.Real time data calculation processing in this way, can obtain good data processed result.
WithRespectively represent the mean value of matching mean value and comparing.S (i, j) and T (i, j)
Respectively represent the matching result value of real-time data and the end value of comparing.
Then template matching is carried out, the Nonlinear Mapping characteristic of the neural network of general error back propagation is suitable for big
The unstructuredness of the out-of-order data of amount maps, at the same the model of neural network be influence e-learning performance principal element it
One.The present embodiment replaces data packet using numerical information, creates item with this parameters revision and Fast Learning to neural network
Part.(x is set by the data mode of input layer1,x2x3,…,xk, 1≤k≤n), n is input layer number, exports the number of plies
According to function setup be yk=f (xk), wherein f (xk) indicate network output function.
Network is by the way that sample learning training, the connection weight adjusted between each neuron forces model parameter to realize
Closely, so that network also carries out the study of network structure while connection weight adjusts.Error is calculated first with formula (2).It is hidden
Hide the Sigmoid action function that layer error calculation uses, Sigmoid function expression and its derivative on closed interval [0,1]
Formula (3) are defined as, wherein λ determines the compression degree of Sigmoid function and the compression degree of data, to avoid falling into office
Portion is minimum generally to take 1.
J=xk-yk (2)
Assuming that j-th of node output function is y in networkk, it is denoted as formula (4)
ω in formulaijFor connection weight, h (si) be hidden layer i-th of neuron output, siIt is the algebra of its input
With:
A in formulanFor the weight that input unit is connect with hidden layer, bmFor the connection weight of hidden layer and output unit.
The characteristic parameter of neural network needs dynamic to adjust, and quickly can dynamically adjust nerve net using the method for slip teeth
Network, very big application range can be had by not needing to introduce additional parameter.But slip teeth method requires least square method flat
Poor (hereinafter referred to as adjustment) Normal Distribution, actual conditions are often more complicated to be difficult to meet.Calculating pre- adjustment simultaneously
It needs to carry out data average computation to the difference vector sample value of pre- side when matrix, can there is the On The Choice of data selection size.
If data selection is too small, the pre- adjustment matrix estimated can have biggish error；On the contrary, if data selected
Greatly, it will lead to pre- adjustment Matrix Estimation value to be difficult to reflect the transient response of system.Learning data feature is needed when target data
It when value number is larger, obtains pre- adjustment and pre- mean value is all bigger, therefore the amplitude of neural network adjustment weight should increase accordingly.
According to the state of neural network center position and the size approximate calculation network system of weight.By need to optimize
The state of discrete systemInput state as network system situation.Assuming that the data of input are M dimension datas, then can calculate
The mean value of input isLeast square adjustment is Ak, then the system mode X of neural networki,kIt is calculated by following formula
δ is by Xi,kRelative toThe scalar that determines of diffusion, general value is conducive in this way between 0.001-1
Estimate the system mode of whole network.
X is learnt according to full probability and conditional probabilityi,k|k-1=Xi,k-1, then the calculation formula of the pre- mean value of system and pre- adjustment is such as
Under,
WhereinIt isRelative toValue, Ak|k-1It is AkRelative to Ak-1Value,
When the error fall flat of output, the value information of input data is fed back and modifies in backpropagation, moves
The ability of state regularized learning algorithm.Each neuron node of hidden layer only receives the feedback of itself unit, does not deposit between each node
It is contacted in feedback, so be independent of each other, and the node of hidden layer is made only to receive the tune of certain specific datas (data cell)
Whole information.A kind of approximation method that weight is the track obtained according to the steepest descending method of the weight space is modified simultaneously, it is network-like
State updates calculating such as formula (10)~formula (13) of weight, and the modified weight of network is carried out using slip teeth method, dynamically adjusts network
Learning direction and accelerating algorithm convergence rate.
Akk=h (xk)Ak|k-1h(xk)T (10)
Aik=Ak|k-1h(xk)T (11)
Neural network uses the structure of error back propagation, in the network training stage, BP neural network output signal and phase
Output is hoped to compare to obtain error signal, error is connected to the network through each middle layer layer-by-layer correction from output layer by slip teeth method and is weighed
Value.Mainly from the angle of compensation, network weight is constantly adjusted using the mapping approximation capability and self-learning capability of neural network
To continuous correction result, so that innovatory algorithm, improves the computational accuracy and algorithm the convergence speed of template matching algorithm.
In template matching method, feature abstraction algorithm is that matched data will be needed to be abstracted the template needed
Then matched data matched with original mass data using the data of template matching, the number being mutually matched
It is believed that breath.In the matching process, determine that a critically important parameter of quality of match is mean absolute difference, it is calculation template matching
One standard value of similarity.Using formula (14), data template that can be set with Rapid matching is obtained similar with template data
Data information.
ai,jIt is needed template data, bi,jIt is the template data set of setting.When matching mass data, it can incite somebody to action
Data are divided into many small data blocks and are handled.S × K is the data traffic and dimension of each data block in template data
Summation.
Setting modifies the calculation formula of the mean absolute difference of template matching using polar coordinates, and the data of the inside are turned
The data information in polar coordinates is turned to,
Wherein a (ρ, θ) is needed template data, and b (ρ, θ) is the template data set of setting.Different types of template
Data and matched data, are constantly corrected, formula is as follows according to the data information of starting
A ' (ρ, θ)=a (ρ+1, θ)-a (ρ, θ), ρ=0,1 ..., r-1 (16)
B ' (ρ, θ)=b (ρ+1, θ)-b (ρ, θ), ρ=0,1 ..., r-1 (17)
Wherein a ' (ρ, θ) is needed the differential value of template data, and b ' (ρ, θ) is the differential for needing matched data set
Value.Differential value is that polar radius is carried out differential as parameter to obtain.It is more due to input data when calculating absolute difference
Sample and complexity need to accurately calculate the degree of Data Matching using differential method.In this way, available accurate
The data for spending the template matching more increased make algorithm obtain more stable convergence, and can significantly reduce study number.
Simulation analysis carried out to the template matching method of the neural network based on slip teeth method on emulation platform, and with tradition
Template matching algorithm IEBP compares experiment.
The experimental situation of simulation analysis is the PC of dominant frequency 3.2G, 4G a memory, and software environment is based on Eclipse 4.3.2
To program realization algorithm function.The structure setting of EBP algorithm is 20-50-1 (input layer-hidden layer-output layer), for each
Sample, initial threshold change at random between [1,2].By the multiple study to 10000 groups of data, algorithm structure is found out most
Excellent threshold value, connection weight and study index, and optimal solution is applied in template matching.
Use convergence number, study number, matching accuracy and the runing time of algorithm as the Performance Evaluating Indexes of algorithm.
When one group of data in a large amount of data are by primary matching, i.e. referred to as an iteration, if error within setting range, i.e.,
It is judged to once restraining, the convergence more algorithm accuracies of number are higher under the same terms, obtain a near-optimal solution and are then known as once
It is higher to learn the fewer efficiency of algorithm of number under same algorithm number of run for study.In the identical situation of data volume, matching is just
True rate be equal to match correct number with match the ratio between number, the more high then accuracy of value is higher.
For 10000 groups of data being randomly generated, by change under given initial threshold, connection weight, study index
Amount carries out dynamic adjustment, and improved algorithm is run multiple times, observes the convergence in mean number of algorithm experimental, obtains such as Fig. 3 to Fig. 6
Simulation result.
Fig. 3 is when threshold value changes at random between [0.5,21.5], and the present embodiment method is after 30 operations
Obtained convergence in mean number.With the continuous variation of threshold value, preferable convergence time can have been obtained on section [0.95,1.1]
Number, but there are still fuctuation within a narrow ranges, and be 1 moment in threshold value, maximum convergence number can be obtained, gradually decline is then presented and becomes
Gesture.Therefore, when threshold value be 1 when, the present embodiment method MTA-IEBP obtain convergence in mean number compared to be carved with when other compared with
Big advantage.
In Fig. 4, in the case where data volume is continuously increased, since the present embodiment method can't go out in the matching process
Existing biggish fluctuation, enables algorithmic statement number to maintain more stable level, does not subtract with the increase of network load
It is few.Compared to IEBP, the present embodiment method can not only obtain more convergence number, and the case where data load increases
Under still can guarantee the stability and reliability of algorithm.
In Fig. 5, in the case where data volume is continuously increased, the study number of IEBP is there are biggish fluctuation, and this reality
Neuron of the method due to deleting redundancy is applied, the error of original start can be just able to satisfy under less study number
Range.
In Fig. 6, the present embodiment method is since it is determined Optimal Learning index, so that algorithm will not crash into Local Minimum
Value, therefore the accuracy of 90% or more acquisition can be stablized, matching accuracy is which greatly improved, and have compared to IEBP
There is preferable stability, i.e. accuracy floating variation range is little.
The technical means disclosed in the embodiments of the present invention is not limited only to technological means disclosed in above embodiment, further includes
Technical solution consisting of any combination of the above technical features.
Claims (2)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610035536.6A CN105740950B (en) | 2016-01-19 | 2016-01-19 | The template matching method of neural network based on slip teeth method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610035536.6A CN105740950B (en) | 2016-01-19 | 2016-01-19 | The template matching method of neural network based on slip teeth method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105740950A CN105740950A (en) | 2016-07-06 |
| CN105740950B true CN105740950B (en) | 2019-03-29 |
Family
ID=56247495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610035536.6A Expired - Fee Related CN105740950B (en) | 2016-01-19 | 2016-01-19 | The template matching method of neural network based on slip teeth method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105740950B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106686385B (en) | 2016-12-30 | 2018-09-25 | 平安科技（深圳）有限公司 | Video compress sensing reconstructing method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104951833A (en) * | 2015-07-02 | 2015-09-30 | 上海电机学院 | Neutral network and learning method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020077756A1 (en) * | 1999-11-29 | 2002-06-20 | Scott Arouh | Neural-network-based identification, and application, of genomic information practically relevant to diverse biological and sociological problems, including drug dosage estimation |
-
2016
- 2016-01-19 CN CN201610035536.6A patent/CN105740950B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104951833A (en) * | 2015-07-02 | 2015-09-30 | 上海电机学院 | Neutral network and learning method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105740950A (en) | 2016-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110880036B (en) | Neural network compression method, device, computer equipment and storage medium | |
| US10762426B2 (en) | Multi-iteration compression for deep neural networks | |
| US10984308B2 (en) | Compression method for deep neural networks with load balance | |
| KR102413028B1 (en) | Method and device for pruning convolutional neural network | |
| Qiao et al. | Constructive algorithm for fully connected cascade feedforward neural networks | |
| CN108490965A (en) | Rotor craft attitude control method based on Genetic Algorithm Optimized Neural Network | |
| CN110232448A (en) | It improves gradient and promotes the method that the characteristic value of tree-model acts on and prevents over-fitting | |
| CN109472345A (en) | A kind of weight update method, device, computer equipment and storage medium | |
| Song et al. | Data-driven finite-horizon optimal tracking control scheme for completely unknown discrete-time nonlinear systems | |
| Putra et al. | Estimation of parameters in the SIR epidemic model using particle swarm optimization | |
| EP3948677A1 (en) | Residual semi-recurrent neural networks | |
| CN110059439A (en) | A kind of spacecraft orbit based on data-driven determines method | |
| CN114330644A (en) | Neural network model compression method based on structure search and channel pruning | |
| CN105469142A (en) | Neural network increment-type feedforward algorithm based on sample increment driving | |
| CN116680969A (en) | Filler evaluation parameter prediction method and device for PSO-BP algorithm | |
| Waheeb et al. | Nonlinear autoregressive moving-average (narma) time series forecasting using neural networks | |
| CN114282478A (en) | A Method for Correcting the Dot Product Error of Variable Resistor Device Array | |
| CN116432539A (en) | A method, system, device and medium for time-coherent cooperative guidance | |
| CN105740950B (en) | The template matching method of neural network based on slip teeth method | |
| CN110083676B (en) | Short text-based field dynamic tracking method | |
| CN115793456A (en) | Lightweight sensitivity-based power distribution network edge side multi-mode self-adaptive control method | |
| JP2023550921A (en) | Weight-based adjustment in neural networks | |
| CN112069370B (en) | Neural network structure search method, device, medium and equipment | |
| CN113051820A (en) | Cross-basin pneumatic parameter simulation method based on convolutional neural network | |
| CN108960406B (en) | MEMS gyroscope random error prediction method based on BFO wavelet neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: No. 66, New Model Road, Gulou District, Nanjing City, Jiangsu Province, 210000 Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS Address before: 210023 9 Wen Yuan Road, Ya Dong new town, Nanjing, Jiangsu. Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200421 Address after: 610000 no.1402, block a, No.199, Tianfu 4th Street, Chengdu high tech Zone, China (Sichuan) pilot Free Trade Zone, Chengdu Patentee after: Chengdu Star Innovation Technology Co.,Ltd. Address before: 210000, 66 new model street, Gulou District, Jiangsu, Nanjing Patentee before: NANJING UNIVERSITY OF POSTS AND TELECOMMUNICATIONS |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: Template matching method of neural network based on sliding tooth method Effective date of registration: 20220526 Granted publication date: 20190329 Pledgee: Industrial Bank Limited by Share Ltd. Chengdu branch Pledgor: Chengdu Star Innovation Technology Co.,Ltd. Registration number: Y2022510000141 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190329 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |