CN107563385A - License plate character recognition method based on depth convolution production confrontation network - Google Patents
License plate character recognition method based on depth convolution production confrontation network Download PDFInfo
- Publication number
- CN107563385A CN107563385A CN201710781905.0A CN201710781905A CN107563385A CN 107563385 A CN107563385 A CN 107563385A CN 201710781905 A CN201710781905 A CN 201710781905A CN 107563385 A CN107563385 A CN 107563385A
- Authority
- CN
- China
- Prior art keywords
- convolutional layer
- license plate
- car plate
- model
- picture
- 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.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
- Character Discrimination (AREA)
Abstract
The present invention proposes a kind of license plate character recognition method based on depth convolution production confrontation network, and specific implementation step is as follows：(1) car plate picture to be identified is extracted；(2) build and train depth convolution production confrontation network DCGAN；(3) car plate picture is generated；(4) sample set of character recognition network is built；(5) build and train Recognition of License Plate Characters network C NN；(6) Recognition of License Plate Characters.The present invention is using the license plate character recognition method based on depth convolution production confrontation network, can effectively overcome in the prior art car plate data it is seriously deficient, training sample causes the shortcomings that network over-fitting less, effectively enhance data sample, and make it that the generalization ability of character recognition network and robustness are stronger, improve Recognition of License Plate Characters rate.
Description
Technical field
The invention belongs to technical field of image processing, the one kind further related in deep learning is given birth to based on depth convolution
An accepted way of doing sth resists the character identifying method of network.The present invention is directed in traffic system, obtained by the high definition photographing device set from crossing
To picture in extract license plate image, and then substantial amounts of license plate image is generated with a small amount of license plate image, by the license plate image of generation
Handle as training sample, train Recognition of License Plate Characters network, realize Recognition of License Plate Characters.
Background technology
With the continuous improvement of social and economic level and the popularization of vehicle, the communication that scale constantly expands is to more intelligence
The technology of energyization and the demand of system are bigger, and intelligent transportation system has become the hot issue of social life.Vehicle identification system
The important component united as intelligent transportation system, remembers automatically in expressway access, parking lot unattended, vehicles peccancy
The fields such as record have this to be widely applied, and important composition of the Recognition of License Plate Characters as vehicle identification system, its realization has
Very big economic value and realistic meaning.
With the raising of computer hardware capacity and the speed of service, deep learning algorithm can carry out real-time to image
Processing, so having there is increasing deep learning algorithm to apply in Car license recognition.Apply at present in Car license recognition
Deep learning method is such as BP neural network, CNN, but deep learning needs substantial amounts of labeled data as training sample,
And often only have partial data to use in the data of magnanimity in reality, and the research of domestic license plate recognition technology is existing
Shape is business and privatization, and researcher is difficult to find the research that open data carry out license plate recognition technology, in data trade net
Stand, car plate data are also fairly expensive, so in the license plate recognition technology of application deep learning, the rare of car plate data is asked
The deep application for having influence on deep learning method of topic.
The patent document " a kind of license plate character recognition method " that Jiuzhou Electrical Appliance Group Co., Ltd., Sichuan applies at it
(number of patent application：201210587347.1 publication number：CN103065137B a kind of license plate character recognition method is proposed in).
The method that this method is combined using canny algorithms and bianry image first identifies character, and the side of character has been extracted during identification
Edge and saltus step information, then the saltus step of character edge pixel is matched with the saltus step of template character set, find matching degree highest
Template character, and then obtain the recognition result of character.This character recognition based on saltus step, is preferably solved under various interference
Character recognition problem, keep relatively stable high discrimination.But the weak point that this method still has is recognition speed
Relatively slow, discrimination and recognition speed are difficult to meet simultaneously, and for vehicle identification of the part without car plate or car plate serious damage
Rate is at a fairly low, it is impossible to the fine real-time met in practical application, accuracy requirement.
Dong Jun prince wife, Zheng Baichuan, Yang Zejing are in its paper delivered " Recognition of License Plate Characters based on convolutional neural networks "
Propose a kind of license plate character recognition method based on convolutional neural networks.This method carries out size to characters on license plate image first
The pretreatments such as normalization, denoising, binaryzation, refinement, character zone be placed in the middle, remove complex background, obtain simple character shape
Structure, proposed CNN models are then utilized to be trained, identify to pretreated characters on license plate collection.Test result indicates that
This method can reach higher correct recognition rata.But the weak point that this method still has is, first, being based on depth
The problem of car plate data are seriously deficient in the Vehicle License Plate Recognition System of study.Secondly, it is necessary to substantial amounts of labeled data is as training sample
This.Third, training sample is very few to cause network over-fitting.
The content of the invention
The purpose of the present invention is to be directed to above-mentioned the shortcomings of the prior art, it is proposed that one kind is based on depth convolution production
The license plate character recognition method of network is resisted, present invention sample compared with other technology of vehicle license plate character identification in the prior art is rich
Richness, recognition accuracy is high, and speed is fast, strong adaptability.
Realizing the thinking of the object of the invention is:First build and train depth convolution production to resist network, will be to be identified
Car plate picture resists network by the depth convolution production trained and generates substantial amounts of car plate picture, then the car plate figure by generation
Piece carries out denoising, binaryzation and Character segmentation respectively, and the numeral obtained by splitting and letter are built into character recognition network
Sample set, then build and train Recognition of License Plate Characters network C NN, the car plate word for finally having trained the input of test alphabetic collection
Symbol identification network C NN is classified, and obtains final character identification result.
Realize comprising the following steps that for the object of the invention：
(1) car plate picture to be identified is extracted：
Using car plate picture extracting method, in the picture from traffic intersection obtained by high definition photographing device, extract and wait to know
Other car plate picture；
(2) build and train depth convolution production confrontation network DCGAN：
(2a) structure is containing the depth convolutional neural networks being of five storeys as generation model；
(2b) structure is containing the convolutional neural networks being of five storeys as discrimination model；
(2c) is trained generation model and discrimination model, by the generation mould trained and sentenced using individually alternating training method
The depth convolution production confrontation network DCGAN of other model composition generation car plate picture；
(3) car plate picture is generated：
Car plate photo resolution to be identified is normalized to 256 × 64 pixel sizes by (3a), the car plate after being handled
Picture；
(3b) by the car plate picture after processing, the depth convolution production for being input to generation car plate picture resists network
DCGAN, export the car plate picture of generation；
(4) sample set of character recognition network is built：
In the picture of (4a) from traffic intersection obtained by high definition photographing device, extract car plate picture to be identified and enter respectively
Row denoising, binaryzation and Character segmentation, 7 character pictures are obtained, remove chinese character, retain letter and number character, formed
Training data sample set A；
(4b) obtains 700 figures from the car plate picture of generation, and denoising, binaryzation and word are carried out respectively to 700 figures
Symbol segmentation, retains letter and number character composing training sample set B；
(4c) chooses 5% sample after set of data samples A and set of data samples B is mixed, remaining as test sample collection
95% sample is as training sample set C；
(5) build and train Recognition of License Plate Characters network C NN：
(5a) structure contains 7 layers of convolutional neural networks CNN；
Training sample set C is input in convolutional Neural net CNN by (5b), training convolutional nerve net CNN, until its output layer
Loss function value be less than or equal to 0.0001, the Recognition of License Plate Characters network C NN trained；
(6) Recognition of License Plate Characters：
Test sample collection is input in the Recognition of License Plate Characters network C NN trained, exports the alphabetical sum identified
Word character.
The present invention has advantages below compared with prior art：
First, because the present invention is using the method for building and training depth convolution production confrontation network DCGAN, generate car
Board picture, the diversity and randomness of training sample set are enriched, enhances the training sample set in character recognition network.Overcome
Car plate data are seriously deficient in the Vehicle License Plate Recognition System based on deep learning in the prior art, and training sample is very few can cause net
Network over-fitting is, it is necessary to the problems such as substantial amounts of labeled data is as training sample so that the present invention is avoided that over-fitting, causes simultaneously
The generalization ability and robustness of network are stronger, it is only necessary to which low volume data saves the consumption of artificial labeled data as training sample
When.
Second, due to the method that the present invention uses Recognition of License Plate Characters, assembled for training using the sample of structure character recognition network
Practice convolutional neural networks CNN and do character recognition, it is slower to overcome recognition speed in existing conventional identification techniques, discrimination and identification
The problems such as speed is difficult to meet simultaneously so that the speed of present invention identification letter and numerical character, recognition correct rate are high.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention.
Embodiment
1 the present invention will be further described below in conjunction with the accompanying drawings.
Step 1, car plate picture to be identified is extracted.
Using car plate picture extracting method, in the picture from traffic intersection obtained by high definition photographing device, extract and wait to know
Other car plate picture, car plate picture extracting method refer to, the picture obtained by high definition photographing device in traffic intersection is used and cut
Figure instrument, 256 × 64 area sizes are intercepted, include the car plate figure to be identified of high-visible car plate numeral and alphabetic character
Piece.
Step 2, build and train depth convolution production confrontation network DCGAN.
Structure is set containing the depth convolutional neural networks being of five storeys as generation model, this 5 layers of depth convolutional neural networks
Putting is, is from left to right followed successively by full articulamentum, micro-stepping width convolutional layer, convolutional layer, micro-stepping width convolutional layer, convolutional layer.
Structure contains the convolutional neural networks being of five storeys, from
It is left-to-right to be followed successively by convolutional layer, stride convolutional layer, convolutional layer, stride convolutional layer, full articulamentum.
Using independent alternately training method, generation model and discrimination model are trained, by the generation mould trained and differentiates mould
The depth convolution production confrontation network DCGAN of type composition generation car plate picture；
The individually alternating training method comprises the following steps that：
1st step, fixed discrimination model, trains generation model, and 1 100 dimension random Gaussian is input into generation model
Full articulamentum, obtain the full articulamentum output characteristic figure of generation model of 512 16 × 4 pixel sizes；
2nd step, the full articulamentum output characteristic figure of generation model is input to generation model micro-stepping width convolutional layer, to it successively
Batch standardization and linear R elu conversion, deconvolution operation are carried out, obtains the generation model micro-stepping of 256 32 × 8 pixel sizes
Width convolutional layer output characteristic figure；
3rd step, generation model micro-stepping width convolutional layer output characteristic figure is input to generation model convolutional layer, it is entered successively
Row batch standardization and linear R elu conversion, deconvolution operation, obtain the generation model convolutional layer of 128 64 × 16 pixel sizes
Output characteristic figure；
4th step, generation model convolutional layer output characteristic figure is input to generation model micro-stepping width convolutional layer, it is entered successively
Row batch standardization and linear R elu conversion, deconvolution operation, obtain the generation model micro-stepping width of 64 128 × 32 pixel sizes
Convolutional layer output characteristic figure；
5th step, generation model micro-stepping width convolutional layer output characteristic figure is input to generation model convolutional layer, it is entered successively
Row batch standardization and linear R elu conversion, deconvolution operation, obtain the generation car plate picture of 256 × 64 pixel sizes；
6th step, fixed generation model, trains discrimination model, by the generation car plate picture of 256 × 64 pixel sizes and waits to know
Other car plate picture is input to discrimination model convolutional layer jointly, and convolution operation is carried out to it, exports as 128 × 32 pixel sizes
Discrimination model convolutional layer output characteristic figure；
7th step, discrimination model convolutional layer output characteristic figure is input to discrimination model stride convolutional layer, it is carried out successively
Standardization, convolution and linear R elu conversion are criticized, exports the discrimination model stride convolutional layer output characteristic of 64 × 16 pixel sizes
Figure；
8th step, discrimination model stride convolutional layer output characteristic figure is input to discrimination model convolutional layer, it is carried out successively
Standardization, convolution and linear R elu conversion are criticized, exports the discrimination model convolutional layer output characteristic figure of 32 × 8 pixel sizes；
9th step, discrimination model convolutional layer output characteristic figure is input to discrimination model stride convolutional layer, it is carried out successively
Standardization, convolution and linear R elu conversion are criticized, exports the discrimination model stride convolutional layer output characteristic of 16 × 4 pixel sizes
Figure；
10th step, discrimination model stride convolutional layer output characteristic figure is input to the full articulamentum of discrimination model, it is carried out
Nonlinear transformation, obtain the full articulamentum output probability of discrimination model；
11st step, the full articulamentum output probability of discrimination model is delivered to generation model, is optimized using optimizer and differentiate mould
Type and generation model, obtain loss function value；
12nd step, judge loss function value whether and meanwhile meet that the loss function value of generation model output layer is less than or equal to
10.000 and discrimination model output layer loss function value be less than or equal to 0.0001, if so, then perform the 13rd step, otherwise, perform
1st step；
13rd step, obtain the depth convolution production confrontation network DCGAN trained.
Step 3, car plate picture is generated.
Car plate photo resolution to be identified is normalized to 256 × 64 pixel sizes, the car plate picture after being handled.
Again by the car plate picture after processing, the depth convolution production confrontation network DCGAN of generation car plate picture is input to, exports generation
Car plate picture.
Step 4, the sample set of character recognition network is built.
In picture from traffic intersection obtained by high definition photographing device, extract car plate picture to be identified and gone respectively
Make an uproar, binaryzation and Character segmentation, obtain 7 character pictures, remove chinese character, retain letter and number character, composing training
Set of data samples A.
700 figures are obtained from the car plate picture of generation, denoising, binaryzation and character point are carried out respectively to 700 figures
Cut, retain letter and number character composing training sample set B.5% is chosen after set of data samples A and set of data samples B is mixed
Sample, as test sample collection, the sample of residue 95% is as training sample set C.
Step 5, build and train Recognition of License Plate Characters network C NN.
Structure contains 7 layers of convolutional neural networks CNN.7 layers of convolutional neural networks CNN setting is, from left to right according to
Secondary is convolutional layer Conv1, pond layer Pool2, convolutional layer Conv3, pond layer Pool4, full articulamentum Fc5, full articulamentum Fc6,
Classify layer Softmax7.Training sample set C is input to convolutional Neural net CNN, training convolutional nerve net CNN is until its output layer
Loss function be less than or equal to 0.0001, the Recognition of License Plate Characters network C NN trained.
Step 6, Recognition of License Plate Characters.
Test sample collection is input in the Recognition of License Plate Characters network C NN trained, exports the alphabetical sum identified
Word character.
The effect of the present invention is further described with reference to emulation experiment.
1. emulation experiment condition：
The hardware environment that the present invention emulates is：Intel Core (TM) i7-6700@3.40GHZ × 8, GPU NVIDIA
GeForce GTX TITAN X, 8GB internal memories；Software environment：Ubuntu 14.04, Ipython2.7；Windows 7, Matlab
R2015b。
2. emulation content and result：
Car plate picture to be identified is input to depth convolution production confrontation net DCGAN by the present invention first, generates 700
Car plate picture.Training sample set C is built again, trains Recognition of License Plate Characters network C NN with training sample set C, test set is inputted
Classify in the Recognition of License Plate Characters network C NN graders trained, show that the final accuracy rate of Recognition of License Plate Characters is
94.66%.Table 1 is to be input to test set in the Recognition of License Plate Characters network C NN trained to identify, after iteration 50 times, statistics
The correct quantity of each character recognition in 34 characters on license plate, discrimination are the correct character quantity of identification and single character
The ratio between total quantity.
Table 1, Recognition of License Plate Characters rate table
| Characters on license plate | 0 | 1 | 2 | 3 | 4 | 5 | 6 |
| Discrimination | 96.5% | 95.42% | 93.67% | 94.18% | 92.42% | 93.47% | 95.42% |
| Characters on license plate | 7 | 8 | 9 | A | B | C | D |
| Discrimination | 96.52% | 93.5% | 95.74% | 96.41% | 91.75% | 97.5% | 91.8% |
| Characters on license plate | E | F | G | H | J | K | L |
| Discrimination | 96.25% | 96.72% | 93.2% | 96.72% | 94.3% | 95.27% | 95.42% |
| Characters on license plate | M | N | P | Q | R | S | T |
| Discrimination | 94.1% | 95.18% | 95.42% | 93.42% | 92.25% | 91.32% | 95.67% |
| Characters on license plate | U | V | W | X | Y | Z | |
| Discrimination | 93.36% | 93.12% | 95.42% | 96.27% | 93.83% | 92.9% |
The table of comparisons 1, two being included in form, characters on license plate and discrimination, characters on license plate have 34, are followed successively by 0,1,2,3,
4,5,6,7,8,9, A, B, C, D, E, F, G, H, J, K, L, M, N, P, Q, R, S, T, U, V, W, X, Y, Z, the discrimination of each character
It is followed successively by:96.5%, 95.42%, 93.67%, 94.18%, 92.42%, 93.47%, 95.42%, 96.52%, 93.5%,
95.74%, 96.41%, 91.75%, 97.5%, 91.8%, 96.25%, 96.72%, 93.2%, 96.72%, 94.3%,
95.27%, 95.42%, 94.1%, 95.18%, 95.42%, 93.42%, 92.25%, 91.32%, 95.67%,
93.36%, 93.12%, 95.42%, 96.27%, 93.83%, 92.9%.It can be seen that numerical character 0,1,6,7,9 and word
Alphabetic character A, C, E, F, H, K, L, N, P, T, W, X discrimination are higher, i.e., the present invention when identifying characters on license plate with higher
Discrimination.
Claims (6)
1. a kind of license plate character recognition method based on depth convolution production confrontation network, it is characterised in that including following step
Suddenly：
(1) car plate picture to be identified is extracted：
Using car plate picture extracting method, in the picture from traffic intersection obtained by high definition photographing device, extract to be identified
Car plate picture；
(2) build and train depth convolution production confrontation network DCGAN：
(2a) structure is containing the depth convolutional neural networks being of five storeys as generation model；
(2b) structure is containing the convolutional neural networks being of five storeys as discrimination model；
(2c) is trained generation model and discrimination model, by the generation mould trained and is differentiated mould using individually alternating training method
The depth convolution production confrontation network DCGAN of type composition generation car plate picture；
(3) car plate picture is generated：
Car plate photo resolution to be identified is normalized to 256 × 64 pixel sizes by (3a), the car plate picture after being handled；
Car plate picture after processing is input to the depth convolution production confrontation network DCGAN of generation car plate picture by (3b), defeated
The car plate picture being born；
(4) sample set of character recognition network is built：
In the picture of (4a) from traffic intersection obtained by high definition photographing device, extract car plate picture to be identified and gone respectively
Make an uproar, binaryzation and Character segmentation, obtain 7 character pictures, remove chinese character, retain letter and number character, composing training
Set of data samples A；
(4b) obtains 700 figures from the car plate picture of generation, and denoising, binaryzation and character point are carried out respectively to 700 figures
Cut, retain letter and number character composing training sample set B；
(4c) chooses 5% sample after set of data samples A and set of data samples B is mixed, as test sample collection, residue 95%
Sample as training sample set C；
(5) build and train Recognition of License Plate Characters network C NN：
(5a) structure contains 7 layers of convolutional neural networks CNN；
Training sample set C is input in convolutional Neural net CNN by (5b), training convolutional nerve net CNN, until the damage of its output layer
Lose functional value and be less than or equal to 0.0001, the Recognition of License Plate Characters network C NN trained；
(6) Recognition of License Plate Characters：
Test sample collection is input in the Recognition of License Plate Characters network C NN trained, exports the letter and number word identified
Symbol.
2. the license plate character recognition method according to claim 1 based on depth convolution production confrontation network, its feature
It is：Car plate picture extracting method described in step (1) refers to, by the picture obtained by high definition photographing device in traffic intersection
Using sectional drawing instrument, 256 × 64 area sizes are intercepted, include the car to be identified of high-visible car plate numeral and alphabetic character
Board picture.
3. the license plate character recognition method according to claim 1 based on depth convolution production confrontation network, its feature
It is：The setting of 5 layers of depth convolutional neural networks described in step (2a) is from left to right to be followed successively by full articulamentum, micro-stepping
Width convolutional layer, convolutional layer, micro-stepping width convolutional layer, convolutional layer.
4. the license plate character recognition method according to claim 1 based on depth convolution production confrontation network, its feature
It is：The setting of 5 layers of convolutional neural networks described in step (2b) is from left to right to be followed successively by convolutional layer, stride convolutional layer,
Convolutional layer, stride convolutional layer, full articulamentum.
5. the license plate character recognition method according to claim 1 based on depth convolution production confrontation network, its feature
It is：Independent alternately training method comprises the following steps that described in step (2c)：
1st step, fixed discrimination model, trains generation model, and 1 100 dimension random Gaussian is input into generation model connects entirely
Layer is connect, obtains the full articulamentum output characteristic figure of generation model of 512 16 × 4 pixel sizes；
2nd step, the full articulamentum output characteristic figure of generation model is input to generation model micro-stepping width convolutional layer, it is carried out successively
Standardization and linear R elu conversion, deconvolution operation are criticized, the generation model micro-stepping width for obtaining 256 32 × 8 pixel sizes is rolled up
Lamination output characteristic figure；
3rd step, generation model micro-stepping width convolutional layer output characteristic figure is input to generation model convolutional layer, it is criticized successively
Standardization and linear R elu conversion, deconvolution operation, obtain the generation model convolutional layer output of 128 64 × 16 pixel sizes
Characteristic pattern；
4th step, generation model convolutional layer output characteristic figure is input to generation model micro-stepping width convolutional layer, it is criticized successively
Standardization and linear R elu conversion, deconvolution operation, obtain the generation model micro-stepping width convolution of 64 128 × 32 pixel sizes
Layer output characteristic figure；
5th step, generation model micro-stepping width convolutional layer output characteristic figure is input to generation model convolutional layer, it is criticized successively
Standardization and linear R elu conversion, deconvolution operation, obtain the generation car plate picture of 256 × 64 pixel sizes；
6th step, fixed generation model, discrimination model is trained, by the generation car plate picture of 256 × 64 pixel sizes and to be identified
Car plate picture is input to discrimination model convolutional layer jointly, and convolution operation is carried out to it, exports as the differentiation of 128 × 32 pixel sizes
Model convolutional layer output characteristic figure；
7th step, discrimination model convolutional layer output characteristic figure is input to discrimination model stride convolutional layer, carries out batch mark successively to it
Standardization, convolution and linear R elu conversion, export the discrimination model stride convolutional layer output characteristic figure of 64 × 16 pixel sizes；
8th step, discrimination model stride convolutional layer output characteristic figure is input to discrimination model convolutional layer, carries out batch mark successively to it
Standardization, convolution and linear R elu conversion, export the discrimination model convolutional layer output characteristic figure of 32 × 8 pixel sizes；
9th step, discrimination model convolutional layer output characteristic figure is input to discrimination model stride convolutional layer, carries out batch mark successively to it
Standardization, convolution and linear R elu conversion, export the discrimination model stride convolutional layer output characteristic figure of 16 × 4 pixel sizes；
10th step, discrimination model stride convolutional layer output characteristic figure is input to the full articulamentum of discrimination model, it carried out non-thread
Property conversion, obtain the full articulamentum output probability of discrimination model；
11st step, the full articulamentum output probability of discrimination model is delivered to generation model, using optimizer optimize discrimination model and
Generation model, obtain loss function value；
12nd step, judge loss function value whether and meanwhile meet generation model output layer loss function value be less than or equal to 10.000
It is less than or equal to 0.0001 with the loss function value of discrimination model output layer, if so, then performing the 13rd step, otherwise, performs the 1st step；
13rd step, obtain the depth convolution production confrontation network DCGAN trained.
6. the license plate character recognition method according to claim 1 based on depth convolution production confrontation network, its feature
It is：7 layers of convolutional neural networks CNN setting described in step (5a) is from left to right to be followed successively by convolutional layer Conv1, pond
Change layer Pool2, convolutional layer Conv3, pond layer Pool4, full articulamentum Fc5, full articulamentum Fc6, classification layer Softmax7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710781905.0A CN107563385B (en) | 2017-09-02 | 2017-09-02 | License plate character recognition method based on depth convolution production confrontation network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710781905.0A CN107563385B (en) | 2017-09-02 | 2017-09-02 | License plate character recognition method based on depth convolution production confrontation network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107563385A true CN107563385A (en) | 2018-01-09 |
| CN107563385B CN107563385B (en) | 2019-10-25 |
Family
ID=60978829
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710781905.0A Active CN107563385B (en) | 2017-09-02 | 2017-09-02 | License plate character recognition method based on depth convolution production confrontation network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107563385B (en) |
Cited By (45)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229434A (en) * | 2018-02-01 | 2018-06-29 | 福州大学 | A kind of vehicle identification and the method for careful reconstruct |
| CN108229130A (en) * | 2018-01-30 | 2018-06-29 | 中国银联股份有限公司 | A kind of verification method and device |
| 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 |
| CN108446700A (en) * | 2018-03-07 | 2018-08-24 | 浙江工业大学 | A kind of car plate attack generation method based on to attack resistance |
| CN108491837A (en) * | 2018-03-07 | 2018-09-04 | 浙江工业大学 | A kind of confrontation attack method improving car plate attack robust |
| CN108647683A (en) * | 2018-04-28 | 2018-10-12 | 浙江大学 | It is a kind of that the character made an uproar is added to fight verification code generation method and system based on frequency domain |
| CN108710892A (en) * | 2018-04-04 | 2018-10-26 | 浙江工业大学 | Synergetic immunity defence method towards a variety of confrontation picture attacks |
| CN108802041A (en) * | 2018-03-16 | 2018-11-13 | 浙江大学 | A kind of method that the small sample set of screen detection is quickly remodeled |
| CN109117863A (en) * | 2018-07-04 | 2019-01-01 | 全球能源互联网研究院有限公司 | The insulation subsample extending method and device of confrontation network are generated based on depth convolution |
| CN109143353A (en) * | 2018-08-20 | 2019-01-04 | 电子科技大学 | A kind of pre-stack seismic waveform classification generating confrontation network based on depth convolution |
| CN109190648A (en) * | 2018-06-26 | 2019-01-11 | Oppo（重庆）智能科技有限公司 | Simulated environment generation method, device, mobile terminal and computer-readable storage medium |
| CN109190504A (en) * | 2018-08-10 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | Processing method, device and the readable storage medium storing program for executing of automobile image data |
| CN109447064A (en) * | 2018-10-09 | 2019-03-08 | 温州大学 | A kind of duplicate rows License Plate Segmentation method and system based on CNN |
| CN109447081A (en) * | 2018-11-14 | 2019-03-08 | 广东小天才科技有限公司 | Method and system for acquiring optical character recognition data set |
| CN109508715A (en) * | 2018-10-30 | 2019-03-22 | 南昌大学 | A kind of License Plate and recognition methods based on deep learning |
| CN109583493A (en) * | 2018-11-27 | 2019-04-05 | 上海交通大学 | A kind of credit card detection and digit recognition method based on deep learning |
| CN109766805A (en) * | 2018-12-28 | 2019-05-17 | 安徽清新互联信息科技有限公司 | A kind of double-deck license plate character recognition method based on deep learning |
| CN109801375A (en) * | 2019-01-24 | 2019-05-24 | 电子科技大学 | Porous material three-dimensional reconstruction method based on depth convolution confrontation neural network |
| CN109815943A (en) * | 2019-03-18 | 2019-05-28 | 北京石油化工学院 | A kind of harmful influence storage stacking picture sample generation method and system |
| CN109815956A (en) * | 2018-12-28 | 2019-05-28 | 安徽清新互联信息科技有限公司 | A kind of license plate character recognition method based on adaptive location segmentation |
| CN109858488A (en) * | 2018-12-28 | 2019-06-07 | 众安信息技术服务有限公司 | A kind of handwriting samples recognition methods and system based on sample enhancing |
| CN109993172A (en) * | 2019-03-25 | 2019-07-09 | 西安电子科技大学 | The licence plate recognition method of road wagon flow |
| CN110147709A (en) * | 2018-11-02 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Training method, device, terminal and the storage medium of vehicle attribute model |
| CN110245613A (en) * | 2019-06-17 | 2019-09-17 | 珠海华园信息技术有限公司 | Ship board recognition methods based on deep learning Characteristic Contrast |
| CN110309889A (en) * | 2019-07-04 | 2019-10-08 | 西南大学 | A kind of Old-Yi character symbol restorative procedure of double arbiter GAN |
| CN110318731A (en) * | 2019-07-04 | 2019-10-11 | 东北大学 | A kind of oil well fault diagnostic method based on GAN |
| CN110348448A (en) * | 2019-07-05 | 2019-10-18 | 黑龙江科技大学 | A kind of license plate character recognition method based on convolutional neural networks |
| CN110659649A (en) * | 2019-09-29 | 2020-01-07 | 哈尔滨工业大学 | Image processing and character recognition algorithm based on near infrared light imaging |
| CN110689001A (en) * | 2018-07-05 | 2020-01-14 | 山东华软金盾软件股份有限公司 | Method for generating license plate training sample in complex environment |
| CN110728626A (en) * | 2018-07-16 | 2020-01-24 | 宁波舜宇光电信息有限公司 | Image deblurring method and apparatus and training thereof |
| CN110969165A (en) * | 2019-11-28 | 2020-04-07 | 中国科学院半导体研究所 | Handwritten character recognition method and device, electronic equipment and storage medium |
| CN111046891A (en) * | 2018-10-11 | 2020-04-21 | 杭州海康威视数字技术股份有限公司 | Training method of license plate recognition model, and license plate recognition method and device |
| CN111046755A (en) * | 2019-11-27 | 2020-04-21 | 上海眼控科技股份有限公司 | Character recognition method, character recognition device, computer equipment and computer-readable storage medium |
| CN111104826A (en) * | 2018-10-26 | 2020-05-05 | 浙江宇视科技有限公司 | License plate character recognition method and device and electronic equipment |
| CN111160338A (en) * | 2019-12-24 | 2020-05-15 | 苏州奥易克斯汽车电子有限公司 | Fuzzy license plate recognition method |
| CN111310791A (en) * | 2020-01-17 | 2020-06-19 | 电子科技大学 | Dynamic progressive automatic target identification method based on small sample number set |
| CN111401360A (en) * | 2020-03-02 | 2020-07-10 | 杭州雄迈集成电路技术股份有限公司 | Method and system for optimizing license plate detection model and license plate detection method and system |
| CN111612081A (en) * | 2020-05-25 | 2020-09-01 | 深圳前海微众银行股份有限公司 | Recognition model training method, device, equipment and storage medium |
| CN111612759A (en) * | 2020-05-19 | 2020-09-01 | 佛山科学技术学院 | Printed matter defect identification method based on deep convolution generation type countermeasure network |
| CN111639542A (en) * | 2020-05-06 | 2020-09-08 | 中移雄安信息通信科技有限公司 | License plate recognition method, device, equipment and medium |
| CN111881914A (en) * | 2020-06-23 | 2020-11-03 | 安徽清新互联信息科技有限公司 | License plate character segmentation method and system based on self-learning threshold |
| CN111932645A (en) * | 2020-06-12 | 2020-11-13 | 重庆大学 | Method for automatically generating ink and wash painting based on generation countermeasure network GAN |
| CN113435446A (en) * | 2021-07-07 | 2021-09-24 | 南京云创大数据科技股份有限公司 | Inclined license plate correction method based on deep learning |
| CN113553885A (en) * | 2020-04-26 | 2021-10-26 | 复旦大学 | Natural scene text recognition method based on generation countermeasure network |
| CN114581903A (en) * | 2022-03-24 | 2022-06-03 | 南通大学 | License plate character recognition method based on convolutional neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170177965A1 (en) * | 2015-12-17 | 2017-06-22 | Xerox Corporation | Coarse-to-fine cascade adaptations for license plate recognition with convolutional neural networks |
| CN106951867A (en) * | 2017-03-22 | 2017-07-14 | 成都擎天树科技有限公司 | Face identification method, device, system and equipment based on convolutional neural networks |
-
2017
- 2017-09-02 CN CN201710781905.0A patent/CN107563385B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170177965A1 (en) * | 2015-12-17 | 2017-06-22 | Xerox Corporation | Coarse-to-fine cascade adaptations for license plate recognition with convolutional neural networks |
| CN106951867A (en) * | 2017-03-22 | 2017-07-14 | 成都擎天树科技有限公司 | Face identification method, device, system and equipment based on convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 王晶: "基于深度学习神经网络的车牌字符识别技术的研究", 《工业控制计算机》 * |
Cited By (68)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229130B (en) * | 2018-01-30 | 2021-04-16 | 中国银联股份有限公司 | Verification method and device |
| CN108229130A (en) * | 2018-01-30 | 2018-06-29 | 中国银联股份有限公司 | A kind of verification method and device |
| 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 |
| CN108229434A (en) * | 2018-02-01 | 2018-06-29 | 福州大学 | A kind of vehicle identification and the method for careful reconstruct |
| CN108446700A (en) * | 2018-03-07 | 2018-08-24 | 浙江工业大学 | A kind of car plate attack generation method based on to attack resistance |
| CN108491837A (en) * | 2018-03-07 | 2018-09-04 | 浙江工业大学 | A kind of confrontation attack method improving car plate attack robust |
| CN108446700B (en) * | 2018-03-07 | 2021-10-29 | 浙江工业大学 | License plate attack generation method based on anti-attack |
| CN108802041A (en) * | 2018-03-16 | 2018-11-13 | 浙江大学 | A kind of method that the small sample set of screen detection is quickly remodeled |
| CN108710892A (en) * | 2018-04-04 | 2018-10-26 | 浙江工业大学 | Synergetic immunity defence method towards a variety of confrontation picture attacks |
| CN108710892B (en) * | 2018-04-04 | 2020-09-01 | 浙江工业大学 | Cooperative immune defense method for multiple anti-picture attacks |
| CN108647683A (en) * | 2018-04-28 | 2018-10-12 | 浙江大学 | It is a kind of that the character made an uproar is added to fight verification code generation method and system based on frequency domain |
| CN109190648A (en) * | 2018-06-26 | 2019-01-11 | Oppo（重庆）智能科技有限公司 | Simulated environment generation method, device, mobile terminal and computer-readable storage medium |
| CN109190648B (en) * | 2018-06-26 | 2020-12-29 | Oppo（重庆）智能科技有限公司 | Simulation environment generation method and device, mobile terminal and computer readable storage medium |
| CN109117863B (en) * | 2018-07-04 | 2021-08-13 | 全球能源互联网研究院有限公司 | Insulator sample expansion method and device based on deep convolution generation countermeasure network |
| CN109117863A (en) * | 2018-07-04 | 2019-01-01 | 全球能源互联网研究院有限公司 | The insulation subsample extending method and device of confrontation network are generated based on depth convolution |
| CN110689001A (en) * | 2018-07-05 | 2020-01-14 | 山东华软金盾软件股份有限公司 | Method for generating license plate training sample in complex environment |
| CN110689001B (en) * | 2018-07-05 | 2023-06-20 | 山东华软金盾软件股份有限公司 | Method for generating license plate training sample in complex environment |
| CN110728626A (en) * | 2018-07-16 | 2020-01-24 | 宁波舜宇光电信息有限公司 | Image deblurring method and apparatus and training thereof |
| CN109190504A (en) * | 2018-08-10 | 2019-01-11 | 百度在线网络技术（北京）有限公司 | Processing method, device and the readable storage medium storing program for executing of automobile image data |
| US11449707B2 (en) | 2018-08-10 | 2022-09-20 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method for processing automobile image data, apparatus, and readable storage medium |
| CN109143353A (en) * | 2018-08-20 | 2019-01-04 | 电子科技大学 | A kind of pre-stack seismic waveform classification generating confrontation network based on depth convolution |
| CN109143353B (en) * | 2018-08-20 | 2019-10-01 | 电子科技大学 | A kind of pre-stack seismic waveform classification generating confrontation network based on depth convolution |
| CN109447064B (en) * | 2018-10-09 | 2019-07-30 | 温州大学 | A kind of duplicate rows License Plate Segmentation method and system based on CNN |
| CN109447064A (en) * | 2018-10-09 | 2019-03-08 | 温州大学 | A kind of duplicate rows License Plate Segmentation method and system based on CNN |
| CN111046891A (en) * | 2018-10-11 | 2020-04-21 | 杭州海康威视数字技术股份有限公司 | Training method of license plate recognition model, and license plate recognition method and device |
| CN111104826B (en) * | 2018-10-26 | 2024-04-26 | 浙江宇视科技有限公司 | License plate character recognition method and device and electronic equipment |
| CN111104826A (en) * | 2018-10-26 | 2020-05-05 | 浙江宇视科技有限公司 | License plate character recognition method and device and electronic equipment |
| CN109508715A (en) * | 2018-10-30 | 2019-03-22 | 南昌大学 | A kind of License Plate and recognition methods based on deep learning |
| CN110147709A (en) * | 2018-11-02 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Training method, device, terminal and the storage medium of vehicle attribute model |
| CN109447081B (en) * | 2018-11-14 | 2021-07-23 | 广东小天才科技有限公司 | Method and system for acquiring optical character recognition data set |
| CN109447081A (en) * | 2018-11-14 | 2019-03-08 | 广东小天才科技有限公司 | Method and system for acquiring optical character recognition data set |
| CN109583493A (en) * | 2018-11-27 | 2019-04-05 | 上海交通大学 | A kind of credit card detection and digit recognition method based on deep learning |
| CN109815956B (en) * | 2018-12-28 | 2022-12-09 | 安徽清新互联信息科技有限公司 | License plate character recognition method based on self-adaptive position segmentation |
| CN109766805B (en) * | 2018-12-28 | 2022-12-09 | 安徽清新互联信息科技有限公司 | Deep learning-based double-layer license plate character recognition method |
| CN109815956A (en) * | 2018-12-28 | 2019-05-28 | 安徽清新互联信息科技有限公司 | A kind of license plate character recognition method based on adaptive location segmentation |
| CN109766805A (en) * | 2018-12-28 | 2019-05-17 | 安徽清新互联信息科技有限公司 | A kind of double-deck license plate character recognition method based on deep learning |
| CN109858488A (en) * | 2018-12-28 | 2019-06-07 | 众安信息技术服务有限公司 | A kind of handwriting samples recognition methods and system based on sample enhancing |
| CN109801375A (en) * | 2019-01-24 | 2019-05-24 | 电子科技大学 | Porous material three-dimensional reconstruction method based on depth convolution confrontation neural network |
| CN109801375B (en) * | 2019-01-24 | 2021-02-09 | 电子科技大学 | Three-dimensional reconstruction method of porous material based on deep convolution anti-neural network |
| CN109815943A (en) * | 2019-03-18 | 2019-05-28 | 北京石油化工学院 | A kind of harmful influence storage stacking picture sample generation method and system |
| CN109993172A (en) * | 2019-03-25 | 2019-07-09 | 西安电子科技大学 | The licence plate recognition method of road wagon flow |
| CN109993172B (en) * | 2019-03-25 | 2022-12-06 | 西安电子科技大学 | License plate recognition method for road traffic flow |
| CN110245613A (en) * | 2019-06-17 | 2019-09-17 | 珠海华园信息技术有限公司 | Ship board recognition methods based on deep learning Characteristic Contrast |
| CN110309889A (en) * | 2019-07-04 | 2019-10-08 | 西南大学 | A kind of Old-Yi character symbol restorative procedure of double arbiter GAN |
| CN110318731A (en) * | 2019-07-04 | 2019-10-11 | 东北大学 | A kind of oil well fault diagnostic method based on GAN |
| CN110348448A (en) * | 2019-07-05 | 2019-10-18 | 黑龙江科技大学 | A kind of license plate character recognition method based on convolutional neural networks |
| CN110659649A (en) * | 2019-09-29 | 2020-01-07 | 哈尔滨工业大学 | Image processing and character recognition algorithm based on near infrared light imaging |
| CN111046755A (en) * | 2019-11-27 | 2020-04-21 | 上海眼控科技股份有限公司 | Character recognition method, character recognition device, computer equipment and computer-readable storage medium |
| CN110969165A (en) * | 2019-11-28 | 2020-04-07 | 中国科学院半导体研究所 | Handwritten character recognition method and device, electronic equipment and storage medium |
| CN110969165B (en) * | 2019-11-28 | 2024-04-09 | 中国科学院半导体研究所 | Handwritten character recognition method, handwritten character recognition device, electronic equipment and storage medium |
| CN111160338A (en) * | 2019-12-24 | 2020-05-15 | 苏州奥易克斯汽车电子有限公司 | Fuzzy license plate recognition method |
| CN111160338B (en) * | 2019-12-24 | 2023-08-01 | 北京奥易克斯科技有限公司 | Fuzzy license plate recognition method |
| CN111310791A (en) * | 2020-01-17 | 2020-06-19 | 电子科技大学 | Dynamic progressive automatic target identification method based on small sample number set |
| CN111401360A (en) * | 2020-03-02 | 2020-07-10 | 杭州雄迈集成电路技术股份有限公司 | Method and system for optimizing license plate detection model and license plate detection method and system |
| CN111401360B (en) * | 2020-03-02 | 2023-06-20 | 杭州雄迈集成电路技术股份有限公司 | Method and system for optimizing license plate detection model, license plate detection method and system |
| CN113553885A (en) * | 2020-04-26 | 2021-10-26 | 复旦大学 | Natural scene text recognition method based on generation countermeasure network |
| CN111639542A (en) * | 2020-05-06 | 2020-09-08 | 中移雄安信息通信科技有限公司 | License plate recognition method, device, equipment and medium |
| CN111612759A (en) * | 2020-05-19 | 2020-09-01 | 佛山科学技术学院 | Printed matter defect identification method based on deep convolution generation type countermeasure network |
| CN111612759B (en) * | 2020-05-19 | 2023-11-07 | 佛山科学技术学院 | Printed matter defect identification method based on deep convolution generation type countermeasure network |
| CN111612081B (en) * | 2020-05-25 | 2024-04-02 | 深圳前海微众银行股份有限公司 | Training method, device, equipment and storage medium for recognition model |
| CN111612081A (en) * | 2020-05-25 | 2020-09-01 | 深圳前海微众银行股份有限公司 | Recognition model training method, device, equipment and storage medium |
| CN111932645A (en) * | 2020-06-12 | 2020-11-13 | 重庆大学 | Method for automatically generating ink and wash painting based on generation countermeasure network GAN |
| CN111932645B (en) * | 2020-06-12 | 2022-08-26 | 重庆大学 | Method for automatically generating ink and wash painting based on generation countermeasure network GAN |
| CN111881914A (en) * | 2020-06-23 | 2020-11-03 | 安徽清新互联信息科技有限公司 | License plate character segmentation method and system based on self-learning threshold |
| CN111881914B (en) * | 2020-06-23 | 2024-02-13 | 安徽清新互联信息科技有限公司 | License plate character segmentation method and system based on self-learning threshold |
| CN113435446B (en) * | 2021-07-07 | 2023-10-31 | 南京云创大数据科技股份有限公司 | Deep learning-based inclined license plate correction method |
| CN113435446A (en) * | 2021-07-07 | 2021-09-24 | 南京云创大数据科技股份有限公司 | Inclined license plate correction method based on deep learning |
| CN114581903A (en) * | 2022-03-24 | 2022-06-03 | 南通大学 | License plate character recognition method based on convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107563385B (en) | 2019-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563385B (en) | License plate character recognition method based on depth convolution production confrontation network | |
| CN104392463B (en) | Image salient region detection method based on joint sparse multi-scale fusion | |
| CN114038037B (en) | Expression label correction and identification method based on separable residual error attention network | |
| CN105354568A (en) | Convolutional neural network based vehicle logo identification method | |
| CN110175613A (en) | Street view image semantic segmentation method based on Analysis On Multi-scale Features and codec models | |
| CN107766787A (en) | Face character recognition methods, device, terminal and storage medium | |
| CN113239954B (en) | Attention mechanism-based image semantic segmentation feature fusion method | |
| CN106156766A (en) | The generation method and device of line of text grader | |
| CN106529446A (en) | Vehicle type identification method and system based on multi-block deep convolutional neural network | |
| CN106909946A (en) | A kind of picking system of multi-modal fusion | |
| CN106845525A (en) | A kind of depth confidence network image bracket protocol based on bottom fusion feature | |
| CN110210027B (en) | Fine-grained emotion analysis method, device, equipment and medium based on ensemble learning | |
| CN115966010A (en) | Expression recognition method based on attention and multi-scale feature fusion | |
| CN106503616A (en) | A kind of Mental imagery Method of EEG signals classification of the learning machine that transfinited based on layering | |
| CN106228139A (en) | A kind of apparent age prediction algorithm based on convolutional network and system thereof | |
| CN107273824A (en) | Face identification method based on multiple dimensioned multi-direction local binary patterns | |
| CN109670559A (en) | Recognition methods, device, equipment and the storage medium of handwritten Chinese character | |
| CN115049534A (en) | Knowledge distillation-based real-time semantic segmentation method for fisheye image | |
| CN116012653A (en) | Method and system for classifying hyperspectral images of attention residual unit neural network | |
| Chaabouni et al. | Combining of off-line and on-line feature extraction approaches for writer identification | |
| CN103136546A (en) | Multi-dimension authentication method and authentication device of on-line signature | |
| CN106650573B (en) | A kind of face verification method and system across the age | |
| CN111310820A (en) | Foundation meteorological cloud chart classification method based on cross validation depth CNN feature integration | |
| CN110533066A (en) | A kind of image data set method for auto constructing based on deep neural network | |
| CN108734170A (en) | Registration number character dividing method based on machine learning and template |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |