CN104809443B - Detection method of license plate and system based on convolutional neural networks - Google Patents
Detection method of license plate and system based on convolutional neural networks Download PDFInfo
- Publication number
- CN104809443B CN104809443B CN201510225485.9A CN201510225485A CN104809443B CN 104809443 B CN104809443 B CN 104809443B CN 201510225485 A CN201510225485 A CN 201510225485A CN 104809443 B CN104809443 B CN 104809443B
- Authority
- CN
- China
- Prior art keywords
- picture
- license plate
- training
- cnn
- neural networks
- 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
Landscapes
- Image Analysis (AREA)
Abstract
A kind of detection method of license plate and system based on convolutional neural networks in image processing and artificial intelligence field, convolutional neural networks are trained as sample set by constructing the picture library with label, and the convolutional neural networks after training are handled into picture to be measured, license plate picture is judged whether it is according to the output vector of convolutional neural networks and most match license plate.The present invention can be improved identification and detection accuracy, have preferable feasibility and robustness.
Description
Technical field
It is specifically a kind of to be based on convolutional Neural the present invention relates to a kind of technology in image processing and artificial intelligence field
The detection method of license plate and system of network.
Background technique
Intelligent transportation prison violating the regulations takes the photograph management system (being commonly called as electronic eyes) and makes a dash across the red light, drives in the wrong direction to motor vehicle, exceeding the speed limit, getting over line row
It sails, the acts of violating regulations such as stop that break rules are monitored management.The system is taken the photograph to obtain the picture of violation vehicle by prison, then from picture
Information of vehicles is further extracted, such as: license plate, vehicle, logo.The present invention is based on this application backgrounds, propose a kind of based on volume
The detection method of license plate of product neural network, this method is a kind of method based on machine learning.
Machine learning is an important subject of artificial intelligence field.The machine learning of early stage is mainly shallow-layer study, with
Scientific technological advance, deep learning is formal in 2006 to be proposed.Deep learning originates from multi-layer artificial neural network, at present
It is successfully applied to the fields such as pattern classification, machine vision, data mining and aid decision.Current existing deep learning network master
It to include convolutional neural networks, depth confidence net and stacking automatic coding machine.Convolutional neural networks are due to the connection of its interlayer and sky
The close relation of domain information, is adapted to image procossing.
Car plate detection process in image generally comprises the positive and negative sample set created for training two classifiers, characteristics of image
It extracts, two classifiers of training carry out target detection using trained classifier.In terms of image characteristics extraction, conventional method
The feature extracting methods such as usually used histograms of oriented gradients (HOG), local binary patterns (LBP) and Haar.However, different
Feature extracting method has its scope of application, if being detached from its scope of application, will lead to not good enough classification results, has limitation
Property.And convolutional neural networks, instead of features above extracting method, deep structure can automatically extract the depth of image
Robust information, conducive to the training and final target detection of classifier.
After searching and discovering the prior art, Chinese patent literature CN104298976A discloses (bulletin) day
2015.01.21 disclosing a kind of detection method of license plate based on convolutional neural networks；It specifically includes by based on Haar feature
Adaboost car plate detection device to license plate image to be detected carry out detection obtain license plate roughing region, pass through convolutional neural networks
Complete license plate identification model carries out the final candidate region of identification acquisition license plate to license plate roughing region, passes through multi-threshold segmentation algorithm
The final candidate region of license plate is split and obtains car plate Chinese character, letter and number, by Chinese character, letter and number convolutional Neural
Network Recognition model is identified to obtain license plate recognition result to car plate Chinese character, letter and number.The technology, which utilizes, is based on Haar
Under the conditions of the complete license plate identification model of the Adaboost car plate detection device and convolutional neural networks of feature may be implemented to not having to
License plate image accurately identifies, at the same using multi-threshold segmentation algorithm to character be split can it is easier to character picture into
Row segmentation has good result in engineer application.But the acquisition final candidate region of license plate places one's entire reliance upon in the technology, and it passes through
The license plate roughing region that Adaboost car plate detection device based on Haar feature obtains, if the license plate area that roughing comes out does not wrap
Containing license plate area, then the final acquisition of license plate area necessarily fails；Meanwhile the detection method of this detour, need two steps
The rapid detection to complete license plate area, increases operation time.
Summary of the invention
The present invention In view of the above shortcomings of the prior art, proposes a kind of car plate detection side based on convolutional neural networks
Method and system, the present invention directly choose detection of the softmax regressand value the best part of CNN the last layer as license plate area
As a result, detection settles at one go, not only reduce operation time, but also the step of independent of may cause detection failure.Energy of the present invention
It is enough to improve identification and detection accuracy, there is preferable feasibility and robustness.
The present invention is achieved by the following technical solutions:
The present invention relates to a kind of detection method of license plate based on convolutional neural networks, and the picture library of label is had by constructing
Convolutional neural networks are trained as sample set, and the convolutional neural networks after training are handled into picture to be measured, according to volume
The output vector of product neural network judges whether it is license plate picture and most matches license plate.
The method specifically includes following steps:
Step A, the picture in sample set is pre-processed as gray level image block: license plate and non-is sorted out from existing picture library
License plate picture, and labelled (license plate or non-license plate), are stored in sample set, obtain identical quantity positive sample and negative sample
This, converts gray level image for the color image in sample set, obtained gray level image size is normalized to the gray scale of 32*32
Image block；
Step B is constructed seven layers of convolutional neural networks (CNN, Convolutional Neural Network): this 7 layers volume
Product neural network includes: that three convolutional layers, two sample levels, a full articulamentum and a softmax return layer, in which:
Input is the gray level image block of 32*32, and convolutional layer C1 has 6 characteristic patterns, and secondary sample level S2 has 6 characteristic patterns, convolutional layer C3 by
S2 layers of 6 characteristic patterns combine after convolution and obtain 16 characteristic patterns, and secondary sample level S4 has 16 characteristic patterns, and convolutional layer C5 has
100 nodes, full articulamentum F6 have 50 nodes, and there are two nodes for output layer；
Step C, training CNN: the positive and negative samples in sample set are input in CNN, using cross entropy loss function, knot
It closes backpropagation BP algorithm and adjusts CNN parameter, returned using softmax as sorting algorithm, complete the training of CNN, it is specific to walk
It is rapid as follows:
Step C1: CNN is initialized: is initialized in networks with some different small random numbers to training parameter；
Step C2: into the CNN after initialization input 10000 training samples come train CNN and obtain reality output to
Amount.
The training sample includes: input vector and ideal output vector, after in input vector input CNN by by
Layer transformation, is transmitted to output layer, obtains reality output vector.
Step C3: using cross entropy loss function, adjusts CNN parameter in conjunction with backpropagation BP algorithm, is returned using softmax
Return the training that CNN is completed as sorting algorithm.
Step D detects license plate: by the CNN after picture to be measured input training, detect whether there is license plate in the picture, and
Obtain testing result, the specific steps are as follows:
Picture to be detected: being converted into grayscale image by step D1, and 1.5 times for being amplified to original image are used as the pyramidal tower of picture
Bottom；
Step D2: with the continuous 7 diminutions tower bottom picture of 0.9 multiplying power, pyramidal 7 layers above of picture are obtained；
Step D3: 8 layers of pyramidal every picture are scanned successively with the scan box of fixed size, the figure in each scan box
CNN after piece input training, obtains two-dimensional output vector (u1, u2), works as u1 > u2, then testing result is license plate；Otherwise, it examines
Survey result is non-license plate.
Step D4: it is all to be detected as in the result of license plate, the corresponding picture of the maximum output vector of u1 value is selected, as
Final car plate detection result.
The present invention relates to a kind of systems for realizing the above method, comprising: training sample obtains module, convolutional neural networks mould
Block, license plate area detection module, in which: the positive negative sample that training sample obtains module inputs convolutional neural networks module, volume
Product neural metwork training can identify two classifiers of license plate and non-license plate picture at one.License plate area detection module creation figure
Piece pyramid and the license plate area that picture in pyramid is detected using trained convolutional neural networks module.
Technical effect
Compared with prior art, the present invention detects in 100 pictures, detects that the picture 98 of license plate is opened, wherein 96
Detection is correct.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart.
Fig. 2 is convolutional neural networks schematic diagram of the present invention.
Fig. 3 is embodiment photo schematic diagram to be processed.
Fig. 4 is that embodiment handles photo schematic diagram.
Specific embodiment
It elaborates below to the embodiment of the present invention, the present embodiment carries out under the premise of the technical scheme of the present invention
Implement, the detailed implementation method and specific operation process are given, but protection scope of the present invention is not limited to following implementation
Example.
Embodiment 1
As shown in Figure 1, the present embodiment the following steps are included:
Step 1, picture pre-processes:
Step 1.1: vehicle photo (including background) is shot under different weather situation, different scenes, from shooting photo
It is partitioned into license plate and non-license plate picture, and labelled (license plate or non-license plate), obtains 5000, license plate picture, non-license plate picture
5000；
Step 1.2: above 10000 color images are converted into gray level image；
Step 1.3: the gray level image size in step 1.2 being normalized to 32*32, every picture obtains two pictures of 32*3
Element value；
Step 1.4: pixel, spatially position is constant, is arranged in the matrix of 32*32, is stored in training sample concentration,
One picture of matrix representative of each 32*32, each sample include picture matrix and corresponding label；
Step 2, convolutional neural networks CNN is constructed:
The CNN used in the present embodiment is the neural network of a multilayer, is made of every time multiple two-dimensional surfaces, Mei Geping
Face is made of multiple independent neurons, and CNN is the topological structure for aiming at two dimensional image and designing, and feature extraction and pattern classification
It carries out simultaneously, the pattern classification better than shallow-layer machine learning algorithm needs additional extractions characteristics of image.In addition, the weight of CNN is shared
Reduce the training parameter of network, adds its multiple feature extraction, make it have robustness.
Step 2.1: construct 7 layers of convolutional neural networks as shown in Figure 2, including three convolutional layers (feature extraction layer), two
A secondary sample level (Feature Mapping layer), a full articulamentum, softmax return layer, after preceding 2 convolutional layers (C layer) all tightly
And then one is used to ask (S layers) of local weighted average secondary sample level to be used as Further Feature Extraction, this distinctive feature twice
The ability that the structure combined makes network have certain tolerance noise to input picture in pattern classification is extracted, that is, is shown as
The robustness of network.
Step 2.2: the input of specified convolutional neural networks is the grayscale image of 32*32, and convolutional layer C1 has 6 characteristic patterns, secondary
Sample level S2 has 6 characteristic patterns, and convolutional layer C3 is combined after convolution by S2 layers of 6 characteristic patterns and obtained 16 characteristic patterns, group
Conjunction mode is shown in Table 1, and secondary sample level S4 has 16 characteristic patterns, and 100 nodes are arranged in convolutional layer C5, and 50 sections are arranged in full articulamentum F6
Two nodes are arranged in point, output layer；
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | |
| 1 | X | X | X | X | X | X | X | X | X | X | ||||||
| 2 | X | X | X | X | X | X | X | X | X | X | ||||||
| 3 | X | X | X | X | X | X | X | X | X | X | ||||||
| 4 | X | X | X | X | X | X | X | X | X | X | ||||||
| 5 | X | X | X | X | X | X | X | X | X | X | ||||||
| 6 | X | X | X | X | X | X | X | X | X | X |
1C3 layers of characteristic pattern combination of table
2.3:C1 layers of step after the convolution mask convolution of 5*5, the size of 6 characteristic patterns is 28*28, every in characteristic pattern
A neuron is connected with the convolution mask of 5*5 in input, each filter 5*5 totally 25 member (unit) parameters and biasing
(bias) parameter, totally 6 filters, altogether (5*5+1) * 6=156 can training parameter, total 156* (28*28)=122304 company
It connects；S2 layers obtain the characteristic pattern of 6 14*14 after time sampling, the 2*2 of each unit in characteristic pattern and character pair figure in C1
Field connection, 4 inputs addition of S2 layer each unit, multiplied by one can training parameter, biasing can be trained along with one.Knot
Fruit is calculated by sigmoid function f (x)=1/ (1+e^ (- x)), and secondary sampling is equivalent to blurred picture, uses 2*2 sampling mould
Be not overlapped when plate, therefore the size of each characteristic pattern is 14*14 in S2, S2 layers have 1 two can training parameter.(2*2+1) * altogether
(14*14)=5880 connection；C3 layers shared 5*5*60+16=1516 can training parameter, 1516* (10*10)=151600
A connection；S4 layers of total 2*1,6=3 two can training parameter, (2*2+1) * 16* (5*5)=2000 connection；C5 layers of (5*5*16+
1) * 100=40100 is a can training parameter；F6 layers (100+1) * 50=5050 can training parameter.
Step 3, training convolutional neural networks CNN:
Step 3.1: being initialized in networks with some different small random numbers to training parameter.
Step 3.2: CNN into Fig. 1 inputs 10000 training samples to train CNN, every sample include (input to
Amount, ideal output vector), input vector is transmitted to output layer, obtains reality output vector by successively transformation.
Step 3.3: using cross entropy loss function, adjust CNN parameter in conjunction with backpropagation BP algorithm, utilize softmax
It returns and is used as sorting algorithm, complete the training of CNN, specifically using document Y Lecun etc. in " Convolutional Networks
for Images,Speech,and Time‐Series”(《Handbook of Brain Theory&Neural Networks》
1995) mode recorded in is realized.
Step 4, car plate detection:
Step 4.1: picture to be detected as shown in figure 3, be converted into grayscale image, and be amplified to by picture to be detected in this experiment
1.5 times of original image are used as the pyramidal tower bottom of picture；
Step 4.2: with the continuous 7 diminutions tower bottom picture of 0.9 multiplying power, obtaining pyramidal 7 layers above of picture；
Step 4.3: according to license plate general proportions shared in picture, the scan box of 160*51 size being selected to move every time
The step-length of 5 pixels, successively scans 8 layers of pyramidal every picture, and the dimension of picture in each scan box resets to 32*
Trained CNN network in 32, then input step 3, obtains two-dimensional output vector (u1, u2), if u1 > u2, testing result
For license plate；Otherwise, testing result is non-license plate.
Step 4.4: it is all to be detected as in the result of license plate, the corresponding picture of the maximum output vector of u1 value is selected, is made
For final car plate detection result.Testing result is as shown in figure 4, be the license plate area detected in white box.After tested, our
Method detects accuracy are as follows: and 98.0%, detection recall rate: 96.0%, it may be assumed that in 100 pictures of detection, detect the picture of license plate
98, wherein 96 detections are correct.
Claims (1)
1. a kind of car plate detection system based on convolutional neural networks characterized by comprising training sample obtains module, volume
Product neural network module and license plate area detection module, in which: training sample obtains the positive negative sample input convolution mind of module
Through network module, convolutional neural networks are trained to two classifiers that can identify license plate and non-license plate picture, license plate area
Domain detection module creation picture pyramid and the vehicle that picture in pyramid is detected using trained convolutional neural networks module
Board region；Training sample obtains module and is instructed as sample set to convolutional neural networks by constructing the picture library with label
Practice, and the convolutional neural networks after training are handled into picture to be measured, is judged whether it is according to the output vector of convolutional neural networks
License plate picture and most matching license plate；
The convolutional neural networks are seven layers of convolutional neural networks, comprising: three convolutional layers, two time sample levels, one it is complete
Articulamentum and a softmax return layer, in which: input is the gray level image block of 32*32, and convolutional layer C1 has 6 characteristic patterns, secondary
Sample level S2 has 6 characteristic patterns, and convolutional layer C3 is combined after convolution by S2 layers of 6 characteristic patterns and obtained 16 characteristic patterns, secondary to adopt
Sample layer S4 has 16 characteristic patterns, and convolutional layer C5 has 100 nodes, and full articulamentum F6 has 50 nodes, and there are two nodes for output layer；
The sample set refers to: sorting out license plate and non-license plate picture from picture library, and labelled, obtains identical quantity
The color image in sample set is converted gray level image by positive sample and negative sample, and obtained gray level image size is normalized
For the gray level image block of 32*32；
The training refers to: the positive and negative samples in sample set being input in CNN, using cross entropy loss function, in conjunction with anti-
CNN parameter is adjusted to BP algorithm is propagated, is returned using softmax as sorting algorithm, completes the training of CNN；
The training includes:
1: CNN being initialized: being initialized in networks with some different small random numbers to training parameter；
2: inputting 10000 training samples into the CNN after initialization to train CNN and obtain reality output vector；
3: use cross entropy loss function, in conjunction with backpropagation BP algorithm adjust CNN parameter, using softmax return as divide
Class algorithm completes the training of CNN；
The training sample includes: input vector and ideal output vector, by successively becoming after in input vector input CNN
It changes, is transmitted to output layer, obtain reality output vector；
The judgement refers to: by the CNN after picture to be measured input training, detecting whether there is license plate in the picture, and obtain
Testing result；
The judgement includes:
1: picture to be detected is converted into grayscale image, and is amplified to original image 1.5 times are used as the pyramidal tower bottom of picture；
2: with the continuous 7 diminutions tower bottom picture of 0.9 multiplying power, obtaining pyramidal 7 layers above of picture；
3: 8 layers of pyramidal every picture being scanned successively with the scan box of fixed size, the picture in each scan box inputs instruction
CNN after white silk obtains two-dimensional output vector (u1, u2), works as u1 > u2, then testing result is license plate；It otherwise is non-license plate；
4: it is all to be detected as in the result of license plate, the corresponding picture of the maximum output vector of u1 value is selected, as final vehicle
Board testing result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510225485.9A CN104809443B (en) | 2015-05-05 | 2015-05-05 | Detection method of license plate and system based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510225485.9A CN104809443B (en) | 2015-05-05 | 2015-05-05 | Detection method of license plate and system based on convolutional neural networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104809443A CN104809443A (en) | 2015-07-29 |
| CN104809443B true CN104809443B (en) | 2018-12-28 |
Family
ID=53694254
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510225485.9A Expired - Fee Related CN104809443B (en) | 2015-05-05 | 2015-05-05 | Detection method of license plate and system based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104809443B (en) |
Families Citing this family (56)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105069472B (en) * | 2015-08-03 | 2018-07-27 | 电子科技大学 | A kind of vehicle checking method adaptive based on convolutional neural networks |
| CN105335754A (en) * | 2015-10-29 | 2016-02-17 | 小米科技有限责任公司 | Character recognition method and device |
| CN106056595B (en) * | 2015-11-30 | 2019-09-17 | 浙江德尚韵兴医疗科技有限公司 | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules |
| CN105469100B (en) * | 2015-11-30 | 2018-10-12 | 广东工业大学 | Skin biopsy image pathological characteristics recognition methods based on deep learning |
| CN105354572B (en) * | 2015-12-10 | 2018-10-12 | 苏州大学 | A kind of automatic license plate identification system based on simplified convolutional neural networks |
| CN105654135A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | Image character sequence recognition system based on recurrent neural network |
| CN105590102A (en) * | 2015-12-30 | 2016-05-18 | 中通服公众信息产业股份有限公司 | Front car face identification method based on deep learning |
| CN105574550B (en) * | 2016-02-02 | 2019-04-12 | 北京格灵深瞳信息技术有限公司 | A kind of vehicle identification method and device |
| CN108475331B (en) * | 2016-02-17 | 2022-04-05 | 英特尔公司 | Method, apparatus, system and computer readable medium for object detection |
| CN105787510A (en) * | 2016-02-26 | 2016-07-20 | 华东理工大学 | System and method for realizing subway scene classification based on deep learning |
| CN107153810A (en) * | 2016-03-04 | 2017-09-12 | 中国矿业大学 | A kind of Handwritten Numeral Recognition Method and system based on deep learning |
| CN105868774A (en) * | 2016-03-24 | 2016-08-17 | 西安电子科技大学 | Selective search and convolutional neural network based vehicle logo recognition method |
| CN105956524A (en) * | 2016-04-22 | 2016-09-21 | 北京智芯原动科技有限公司 | Method and device for identifying traffic signs |
| CN105975915B (en) * | 2016-04-28 | 2019-05-21 | 大连理工大学 | A kind of front vehicles parameter identification method based on multitask convolutional neural networks |
| CN105956597A (en) * | 2016-05-04 | 2016-09-21 | 浙江大学 | Binocular stereo matching method based on convolution neural network |
| CN105975968B (en) * | 2016-05-06 | 2019-03-26 | 西安理工大学 | A kind of deep learning license plate character recognition method based on Caffe frame |
| CN105956626A (en) * | 2016-05-12 | 2016-09-21 | 成都新舟锐视科技有限公司 | Deep learning based vehicle license plate position insensitive vehicle license plate recognition method |
| CN105975941B (en) * | 2016-05-31 | 2019-04-12 | 电子科技大学 | A kind of multi-direction vehicle detection identifying system based on deep learning |
| CN106295646B (en) * | 2016-08-10 | 2019-08-23 | 东方网力科技股份有限公司 | A kind of registration number character dividing method and device based on deep learning |
| CN106469299B (en) * | 2016-08-31 | 2019-07-19 | 北京邮电大学 | A kind of vehicle search method and device |
| CN106570516A (en) * | 2016-09-06 | 2017-04-19 | 国网重庆市电力公司电力科学研究院 | Obstacle recognition method using convolution neural network |
| CN107844794B (en) * | 2016-09-21 | 2022-02-22 | 北京旷视科技有限公司 | Image recognition method and device |
| CN106599994B (en) * | 2016-11-23 | 2019-02-15 | 电子科技大学 | A kind of gaze estimation method based on depth Recurrent networks |
| CN106557768B (en) * | 2016-11-25 | 2021-07-06 | 北京小米移动软件有限公司 | Method and device for recognizing characters in picture |
| CN106778590B (en) * | 2016-12-09 | 2020-07-17 | 厦门大学 | Violence and terrorism video detection method based on convolutional neural network model |
| CN106599892A (en) * | 2016-12-14 | 2017-04-26 | 四川长虹电器股份有限公司 | Television station logo identification system based on deep learning |
| CN106650660A (en) * | 2016-12-19 | 2017-05-10 | 深圳市华尊科技股份有限公司 | Vehicle type recognition method and terminal |
| US10984289B2 (en) | 2016-12-23 | 2021-04-20 | Shenzhen Institute Of Advanced Technology | License plate recognition method, device thereof, and user equipment |
| CN106778745A (en) * | 2016-12-23 | 2017-05-31 | 深圳先进技术研究院 | A kind of licence plate recognition method and device, user equipment |
| CN108268885B (en) * | 2017-01-03 | 2020-06-30 | 京东方科技集团股份有限公司 | Feature point detection method, device, and computer-readable storage medium |
| CN106874863B (en) * | 2017-01-24 | 2020-02-07 | 南京大学 | Vehicle illegal parking and reverse running detection method based on deep convolutional neural network |
| CN106971185B (en) * | 2017-03-27 | 2020-03-13 | 张东森 | License plate positioning method and device based on full convolution network |
| CN107092661A (en) * | 2017-03-28 | 2017-08-25 | 桂林明辉信息科技有限公司 | A kind of image search method based on depth convolutional neural networks |
| CN107134144B (en) * | 2017-04-27 | 2019-07-12 | 武汉理工大学 | A kind of vehicle checking method for traffic monitoring |
| CN107203754B (en) * | 2017-05-26 | 2019-04-12 | 北京邮电大学 | A kind of license plate locating method and device based on deep learning |
| CN107273832B (en) * | 2017-06-06 | 2020-09-22 | 青海省交通科学研究院 | License plate recognition method and system based on integral channel characteristics and convolutional neural network |
| CN107239778B (en) * | 2017-06-09 | 2020-01-03 | 中国科学技术大学 | Efficient and accurate license plate recognition method |
| CN107273894A (en) * | 2017-06-15 | 2017-10-20 | 珠海习悦信息技术有限公司 | Recognition methods, device, storage medium and the processor of car plate |
| CN107341480A (en) * | 2017-07-12 | 2017-11-10 | 中国电子科技集团公司第二十八研究所 | A kind of crowd massing detection method of modified PCCNN neural network models |
| CN107560849B (en) * | 2017-08-04 | 2020-02-18 | 华北电力大学 | Wind turbine generator bearing fault diagnosis method of multichannel deep convolutional neural network |
| CN107590498B (en) * | 2017-09-27 | 2020-09-01 | 哈尔滨工业大学 | Self-adaptive automobile instrument detection method based on character segmentation cascade two classifiers |
| CN107909021A (en) * | 2017-11-07 | 2018-04-13 | 浙江师范大学 | A kind of guideboard detection method based on single deep layer convolutional neural networks |
| CN108090423B (en) * | 2017-12-01 | 2021-06-29 | 上海工程技术大学 | Depth license plate detection method based on thermodynamic diagram and key point regression |
| CN108229474B (en) * | 2017-12-29 | 2019-10-01 | 北京旷视科技有限公司 | Licence plate recognition method, device and electronic equipment |
| CN110659631B (en) * | 2018-06-30 | 2023-12-15 | 华为技术有限公司 | License plate recognition method and terminal equipment |
| TWI677826B (en) * | 2018-09-19 | 2019-11-21 | 國家中山科學研究院 | License plate recognition system and method |
| CN111522228B (en) * | 2019-02-02 | 2023-09-22 | 北京京东乾石科技有限公司 | Method and device for detecting aircraft |
| CN111738032B (en) * | 2019-03-24 | 2022-06-24 | 魔门塔(苏州)科技有限公司 | Vehicle driving information determination method and device and vehicle-mounted terminal |
| CN110837837B (en) * | 2019-11-05 | 2023-10-17 | 安徽工业大学 | Vehicle violation detection method based on convolutional neural network |
| CN112926610A (en) * | 2019-12-06 | 2021-06-08 | 顺丰科技有限公司 | Construction method of license plate image screening model and license plate image screening method |
| CN111444911B (en) * | 2019-12-13 | 2021-02-26 | 珠海大横琴科技发展有限公司 | Training method and device of license plate recognition model and license plate recognition method and device |
| CN111489560B (en) * | 2020-04-13 | 2022-10-14 | 深圳市海圳汽车技术有限公司 | Detection method and control method for detecting front vehicle starting based on shallow convolutional neural network |
| CN111508269B (en) * | 2020-04-23 | 2021-05-18 | 深圳智优停科技有限公司 | Open type parking space vehicle distinguishing method and device based on image recognition |
| CN111591715A (en) * | 2020-05-28 | 2020-08-28 | 华中科技大学 | Belt longitudinal tearing detection method and device |
| CN112906699B (en) * | 2020-12-23 | 2024-06-14 | 深圳市信义科技有限公司 | Detection and identification method for license plate amplified number |
| CN113569844A (en) * | 2021-06-29 | 2021-10-29 | 深圳市捷顺科技实业股份有限公司 | License plate detection method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036323A (en) * | 2014-06-26 | 2014-09-10 | 叶茂 | Vehicle detection method based on convolutional neural network |
| CN104166861A (en) * | 2014-08-11 | 2014-11-26 | 叶茂 | Pedestrian detection method |
| CN104298976A (en) * | 2014-10-16 | 2015-01-21 | 电子科技大学 | License plate detection method based on convolutional neural network |
| CN104408435A (en) * | 2014-12-05 | 2015-03-11 | 浙江大学 | Face identification method based on random pooling convolutional neural network |
-
2015
- 2015-05-05 CN CN201510225485.9A patent/CN104809443B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036323A (en) * | 2014-06-26 | 2014-09-10 | 叶茂 | Vehicle detection method based on convolutional neural network |
| CN104166861A (en) * | 2014-08-11 | 2014-11-26 | 叶茂 | Pedestrian detection method |
| CN104298976A (en) * | 2014-10-16 | 2015-01-21 | 电子科技大学 | License plate detection method based on convolutional neural network |
| CN104408435A (en) * | 2014-12-05 | 2015-03-11 | 浙江大学 | Face identification method based on random pooling convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于卷积神经网络LeNet-5的车牌字符识别研究;赵志宏 等;《系统仿真学报》;20100331;第22卷(第3期);第638-641页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104809443A (en) | 2015-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104809443B (en) | Detection method of license plate and system based on convolutional neural networks | |
| CN110956094B (en) | RGB-D multi-mode fusion personnel detection method based on asymmetric double-flow network | |
| Chakma et al. | Image-based air quality analysis using deep convolutional neural network | |
| CN109584248B (en) | Infrared target instance segmentation method based on feature fusion and dense connection network | |
| Anagnostopoulos et al. | A license plate-recognition algorithm for intelligent transportation system applications | |
| CN111723748A (en) | Infrared remote sensing image ship detection method | |
| Kim et al. | Multi-task convolutional neural network system for license plate recognition | |
| CN109410184B (en) | Live broadcast pornographic image detection method based on dense confrontation network semi-supervised learning | |
| KR20160143494A (en) | Saliency information acquisition apparatus and saliency information acquisition method | |
| CN106845487A (en) | A kind of licence plate recognition method end to end | |
| CN112395951B (en) | Complex scene-oriented domain-adaptive traffic target detection and identification method | |
| CN104504395A (en) | Method and system for achieving classification of pedestrians and vehicles based on neural network | |
| CN106529578A (en) | Vehicle brand model fine identification method and system based on depth learning | |
| CN103049763A (en) | Context-constraint-based target identification method | |
| CN104299006A (en) | Vehicle license plate recognition method based on deep neural network | |
| CN110766020A (en) | System and method for detecting and identifying multi-language natural scene text | |
| CN114841972A (en) | Power transmission line defect identification method based on saliency map and semantic embedded feature pyramid | |
| CN110929593A (en) | Real-time significance pedestrian detection method based on detail distinguishing and distinguishing | |
| CN108154160A (en) | Color recognizing for vehicle id method and system | |
| CN101364263A (en) | Method and system for detecting skin texture to image | |
| CN110659550A (en) | Traffic sign recognition method, traffic sign recognition device, computer equipment and storage medium | |
| CN114067444A (en) | Face spoofing detection method and system based on meta-pseudo label and illumination invariant feature | |
| Sulehria et al. | Vehicle number plate recognition using mathematical morphology and neural networks | |
| CN108229434A (en) | A kind of vehicle identification and the method for careful reconstruct | |
| Fleyeh | Traffic and road sign recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181228 Termination date: 20210505 |