CN108509954A - A kind of more car plate dynamic identifying methods of real-time traffic scene - Google Patents
A kind of more car plate dynamic identifying methods of real-time traffic scene Download PDFInfo
- Publication number
- CN108509954A CN108509954A CN201810368034.4A CN201810368034A CN108509954A CN 108509954 A CN108509954 A CN 108509954A CN 201810368034 A CN201810368034 A CN 201810368034A CN 108509954 A CN108509954 A CN 108509954A
- Authority
- CN
- China
- Prior art keywords
- convolutional layer
- network
- bounding box
- confidence
- convolutional
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/63—Scene text, e.g. street names
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of more car plate dynamic identifying methods of real-time traffic scene, are broadly divided into several key steps such as image preprocessing, License Plate network training, post processing of image, identification network training.Utilize License Plate network and character recognition network, it can be in the case where not dividing character information, multiple license board informations in single picture are efficiently identified in real time, there is innovation advantage compared to traditional Character segmentation, recognition methods, and can widely apply in the monitoring of the acts of violating regulations such as parking fee collective system, highway.
Description
Technical field
The present invention relates to machine vision and technical field of image processing, more car plates of specifically a kind of real-time traffic scene are dynamic
State recognition methods.
Background technology
In recent years, with the fast development of intelligent transportation field, the research of license auto-recognition system algorithm causes increasingly
Extensive concern.Automation, quick, accurate and powerful Vehicle License Plate Recognition System have become traffic control and traffic law law enforcement
Demand.
Traditional licence plate recognition method is broadly divided into three license plate area positioning, License Plate Character Segmentation, character recognition steps,
And the single car plate for being all based on single picture is identified.Prior art CN107220638A discloses a kind of based on depth
Learn the car plate detection recognition methods of convolutional neural networks, main contents include：Data acquisition module, detection recognition training mould
Block, character locating test module, process are to contain car plate in real world using construction automatic storage system to sort out first
Image, acquired in different illumination, visible angle, scene sufficient amount of car plate with cutting character picture, then use one
Serial deep neural network carries out the training of car plate detection and identification, and obtained model is individually examined by the character of well cutting again
It surveys and identification, final merging becomes result.The technological merit is the vehicle intact with cutting of the car plate under the conditions of different scenes
Board character input trained network, wherein Character segmentation and identification division are the cutting operations for devising multi-step, so
Character is classified and detected afterwards.But although the technology can identify car plate, due to being illuminated by the light intensity, car plate shape
Shape, stain such as block at a series of influence of reasons, can not perfectly be cut to character, this was just identified to subsequent
Journey brings prodigious uncertainty.
Have again prior art CN107590774A provide it is a kind of based on generate confrontation network car plate clarification method and
Device, including：Target license plate image is input to generation network, network is generated and trains to obtain by sample license plate image, being used for will
The license plate image of low resolution carries out sharpening processing, and generates the high-resolution license plate image for meeting car plate standard, wherein
Sample license plate image is the license plate image for meeting car plate standard；Based on network is generated, target license plate image is carried out at sharpening
Reason；Output meets the clear license plate image of car plate standard.The technological merit be by license plate image carry out sharpening processing, according to
Clear image analysis and identification car plate, improves recognition accuracy.But the processing of the sharpening of the technology is in low resolution figure
Image is extracted in piece, low resolution extraction process is likely to result in car plate photo and changes, and eventually leads to Car license recognition mistake.
Invention content
The purpose of the present invention is to provide a kind of more car plate dynamic identifying methods of real-time traffic scene, to solve the above-mentioned back of the body
The Car license recognition proposed in scape technology is easily influenced by outside environmental elements, easily occurs license plate retrieving mistake during car plate sharpening
Mislead the problem of causing Car license recognition error.
To achieve the above object, the present invention provides the following technical solutions：
A kind of more car plate dynamic identifying methods of real-time traffic scene, it is characterised in that use a variety of convolutional neural networks pair
Car plate carries out whole identification, includes the following steps：
S1, convolutional neural networks are established comprising License Plate network and Car license recognition；
S2, image preprocessing：The image collected is adjusted to 224*224 resolution ratio；
S3, positioning network training：Processed image in S2 is input in License Plate network and carries out feature learning；
S4, post processing of image：License plate image after being positioned in S3 is stretched to 224*224 resolution ratio；
S5, identification network training：Processed license plate image in S4 is input in Car license recognition network and is learnt；
S6, neural network test：Test pictures are input in License Plate network and Car license recognition network and carry out entirety
Test.
Preferably, the feature learning that image is inputted to License Plate network in the S3 includes the following steps：
S31, the License Plate network have 16 convolutional layers, 4 pond layers, 2 full articulamentums, adopt between layers
With linear activation primitive, the full articulamentum of last layer uses modified linear activation primitive, and convolutional layer is for extracting target spy
Sign, full articulamentum is for predicting coordinate and class probability；
S32, input picture is divided into S*S grid, is traversed, when detection target's center appears in grid,
The lattice detect the target and provide the target indicates that the grid possesses the value of the confidence of the confidence value of target, the value of the confidence model in this lattice
It is 0~1 to enclose, and the value of the confidence is defined as：Wherein IOU is to indicate bounding box and realistic objective frame
The friendship of the ratio of intersection and union and than function, is defined as bounding box, each grid is pre- by the target area detected
It measures B bounding box, bounding box and predicts five values：x、y、h、w、confidence；
The each grid divided in S33, S32 predicts that a classification information for being denoted as C classes, prediction C kinds assume the item of classification
Part probability is Pr (Classi|Object)；
S34, tensor is calculated, tensor computation formula is：S*S(B*5+C)；
S35, it is multiplied with the value of the confidence by the class probability for obtaining S32 and S33, obtains the confidence score of the category, used
In the probability for indicating that the category occurs in bounding box, the formula is as follows：
Wherein, equation left side first item represents the classification information of each grid forecasting；
S36, coordinates computed predict that loss function, the formula are as follows：
Wherein, λcoordRepresent bounding box prediction weight, λnoobjRepresent the confidence level prediction power not comprising object boundary frame
Weight, whereinIndicate whether target appears in grid cell i,Indicate j-th of bounding box prediction in grid cell i
Device is responsible for prediction；
Bounding box the value of the confidence prediction loss function containing target is calculated as：
Bounding box the value of the confidence prediction loss function not comprising target is calculated as：
Class prediction loss function is calculated as：
S37, setting recognition threshold be used as standard, will obtain the particular category of each bounding box confidence score and
Given threshold value compares, and filters the lower bounding box of score, and non-maxima suppression is carried out to the bounding box of reservation
Processing, obtains final testing result.
More preferably, there are the License Plate network 16 convolutional layers, 2 full articulamentums to be based on figure to realize
As full figure information exports end to end, network structure is followed successively by：Convolutional layer conv1, pond layer max pool, convolutional layer
Conv2, pond layer max pool, convolutional layer conv3, convolutional layer conv4, convolutional layer conv5, convolutional layer conv6, pond layer
Max pool, convolutional layer conv7, convolutional layer conv8, convolutional layer conv9, convolutional layer conv10, pond layer max pool, convolution
It is layer conv11, convolutional layer conv12, convolutional layer conv13, convolutional layer conv14, convolutional layer conv15, convolutional layer conv16, complete
Articulamentum FC1 and full articulamentum FC2.
Preferably, the Car license recognition network further includes character recognition network algorithm, and it includes 5 convolutional layers, 3 connect entirely
Layer is connect, network structure is followed successively by：Convolutional layer conv1, pond layer max pool, convolutional layer conv2, pond layer max pool,
Convolutional layer conv3, convolutional layer conv4, convolutional layer conv5, pond layer max pool, full articulamentum FC1, full articulamentum FC2 and
Full articulamentum FC3, wherein full articulamentum FC3 dimensions are divided into 7 parts, for follow-up softmax classification characters.
More preferably, the full articulamentum FC3 dimensions of the character recognition network structure are divided into 7 parts and include：Using
7 softmax graders of exportable 7 label respectively classify to character, and each softmax graders are responsible for output
The probability of the character of car plate corresponding position, each grader output character is：
And the output character probability of each softmax graders adds up to 1, i.e.,：
Compared with prior art, the beneficial effects of the invention are as follows：
1. the present invention obtains multiple license board informations by carrying out neural network analysis to single picture, the height of resource is realized
Effect utilizes；
2. the present invention realizes system by merging a variety of License Plates based on deep learning and car plate global recognition method
Robustness promotion；
3. the present invention is identified by the global information to car plate, overcomes in conventional method and band is split to character
The uncertainty come, realizes the accuracy of Car license recognition；
4. the present invention is changed to 7 for character by the improvement to classic network, by last layer of character recognition network
The softmax layers of classification, parallel output character string realize the raising of character recognition efficiency, have wound compared to traditional network
New meaning.
Description of the drawings
Fig. 1 is a kind of more car plate dynamic identifying method Car license recognition flow charts of real-time traffic scene of the present invention；
Fig. 2 is a kind of more car plate dynamic identifying method License Plate convolutional neural networks of real-time traffic scene of the present invention
Figure；
Fig. 3 is a kind of more car plate dynamic identifying method character recognition convolutional neural networks of real-time traffic scene of the present invention
Figure.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Please refer to Fig.1-3
As shown in Figure 1, a kind of more car plate dynamic identifying methods of real-time traffic scene of the present invention, by being used for car plate
The convolutional neural networks of positioning and for Recognition of License Plate Characters convolutional neural networks form, can be with the positioning and knowledge of real-time high-efficiency
Multiple car plates in other single picture.
Sample set is constructed first and it is pre-processed.Collect the candid photograph figure above a large amount of crossing cameras and road
Picture, and be unified 224*224 resolution sizes by Image Adjusting.We use and reserve the collection that data set is divided into two mutual exclusions by method
It closes, one of set is used as training set, another is as test set.After training model on training set, with test set come
Assess test error.
Ready-portioned test set, is input to License Plate network as shown in Figure 2 by the training for carrying out License Plate network
It is middle to carry out feature training and the self study of parameter.Algorithm the specific steps are：
A. piece image is divided into S*S grid.If the central point of objective body is fallen in this grid, the object
The prediction work of body just transfers to this grid to be responsible for；
B. each grid will predict B bounding box and a classification information C class, each bounding box predictions
Five values, i.e., respectively abscissa x of the target's center relative to grid, ordinate y of the target's center relative to grid,
The high h and the value of the confidence confidence, the value of the confidence of the wide w of bounding box, bounding box are defined as：Predict that C kinds assume that the conditional probability of classification is Pr (Classi|Object)；
C. finally obtained tensor is calculated using S*S (B*5+C), the grid for being divided into 7*7 predicts target, each
Grid predicts that 2 bounding box, classification C take 20.So the tensor that finally obtained output is 7*7*30；
D. when test, the confidence letters of the classification information and bounding box predictions of each grid forecasting
Manner of breathing multiplies, it will be able to obtain the confidence score of the particular category of each bounding box；
E. the loss function that we use in training process includes four parts, wherein λcoordRepresent bounding box prediction power
Weight, λnoobjIt represents the confidence level not comprising object boundary frame and predicts weight, whereinIndicate whether target appears in grid list
In first i,Indicate that j-th of bounding box fallout predictor in grid cell i is responsible for prediction.
Coordinate prediction loss function is calculated as：
Bounding box the value of the confidence prediction loss function containing target is calculated as：
Bounding box the value of the confidence prediction loss function not comprising target is calculated as：
Class prediction loss function is calculated as：
F. after the confidence score for obtaining each bounding box particular categories, threshold value is set, it is lower to filter score
Bounding box carry out non-maxima suppression processing to the bounding box of reservation, just obtain final testing result.
Further the picture exported in positioning network is post-processed, is uniformly adjusted to 224*224 sizes.So
It is input in character recognition network as shown in Figure 3 and is learnt afterwards.The full articulamentum of the first two in character recognition network, we
Prevent model from serious over-fitting occur using dropout technologies.Last layer in character recognition network connects entirely
Layer we using 7 softmax graders to classifying respectively to character.Each softmax graders are responsible for exporting car plate
The character of corresponding position.
Finally, test set is input in network and carries out integrated testability, further Optimal Parameters are to reach better
Effect.
As shown in Fig. 2, License Plate network structure is followed successively by：The pond layer max of convolutional layer conv1,2*2 of 7*7*64
Pond layer max pool of convolutional layer conv2,2*2 of pool, 3*3*192, convolutional layer conv3,3*3*256 of 1*1*128
Pond layer max pool, the 1* of convolutional layer conv6,2*2 of convolutional layer conv5,3*3*512 of convolutional layer conv4,1*1*256
The convolution of convolutional layer conv9,3*3*1024 of convolutional layer conv8,1*1*512 of convolutional layer conv7,3*3*512 of 1*256
The convolutional layer conv12 of convolutional layer conv11,3*3*1024 of layer pond layer max pool of conv10,2*2,1*1*512,
Convolutional layer conv15,3*3* of convolutional layer conv14,3*3*1024 of convolutional layer conv13,3*3*1024 of 3*3*1024
1024 convolutional layer conv16, full articulamentum FC1 and full articulamentum FC2.
As shown in figure 3, character recognition network structure is followed successively by：Convolutional layer conv1 (11*11), pond layer max pool,
Convolutional layer conv2 (5*5), pond layer max pool, convolutional layer conv3 (3*3), convolutional layer conv4 (3*3), convolutional layer conv5
(3*3), pond layer max pool, full articulamentum FC1, full articulamentum FC2 and full articulamentum FC3.
It although an embodiment of the present invention has been shown and described, for the ordinary skill in the art, can be with
Understanding without departing from the principles and spirit of the present invention can carry out these embodiments a variety of variations, modification, replace
And modification, the scope of the present invention is defined by the appended.
Claims (5)
1. a kind of more car plate dynamic identifying methods of real-time traffic scene, it is characterised in that with a variety of convolutional neural networks to vehicle
Board carries out whole identification, includes the following steps：
S1, convolutional neural networks are established comprising License Plate network and Car license recognition；
S2, image preprocessing：The image collected is adjusted to 224*224 resolution ratio；
S3, positioning network training：Processed image in S2 is input in License Plate network and carries out feature learning；
S4, post processing of image：License plate image after being positioned in S3 is stretched to 224*224 resolution ratio；
S5, identification network training：Processed license plate image in S4 is input in Car license recognition network and is learnt；
S6, neural network test：Test pictures are input in License Plate network and Car license recognition network and carry out integrated testability.
2. a kind of more car plate dynamic identifying methods of real-time traffic scene according to claim 1, which is characterized in that described
The feature learning that image is inputted to License Plate network in S3 includes the following steps：
S31, the License Plate network have 16 convolutional layers, 4 pond layers, 2 full articulamentums, use line between layers
Property activation primitive, the full articulamentum of last layer use modified linear activation primitive, and convolutional layer is for extracting target signature, entirely
Articulamentum is for predicting coordinate and class probability；
S32, input picture is divided into S*S grid, is traversed, when detection target's center appears in grid, the lattice
It detects the target and provides the target and indicate that the grid possesses the value of the confidence of the confidence value of target in this lattice, the value of the confidence is ranging from
0~1, the value of the confidence is defined as：Wherein IOU is to indicate bounding box and realistic objective frame intersection
Friendship with the ratio of union and than function, is defined as bounding box, each grid predicts B by the target area detected
A bounding box, bounding box predict five values：x、y、h、w、confidence；
The each grid divided in S33, S32 predicts that a classification information for being denoted as C classes, prediction C kinds assume that the condition of classification is general
Rate is Pr (Classi|Object)；
S34, tensor is calculated, tensor computation formula is：S*S(B*5+C)；
S35, it is multiplied with the value of the confidence by the class probability for obtaining S32 and S33, obtains the confidence score of the category, be used for table
Show the probability that the category occurs in bounding box, the formula is as follows：
Wherein, equation left side first item represents the classification information of each grid forecasting；
S36, coordinates computed predict that loss function, the formula are as follows：
Wherein, λcoordRepresent bounding box prediction weight, λnoobjIt represents the confidence level not comprising object boundary frame and predicts weight, whereinIndicate whether target appears in grid cell i,It is pre- to indicate that j-th of bounding box fallout predictor in grid cell i is responsible for
It surveys；
Bounding box the value of the confidence prediction loss function containing target is calculated as：
Bounding box the value of the confidence prediction loss function not comprising target is calculated as：
Class prediction loss function is calculated as：
S37, setting recognition threshold will obtain the confidence score of the particular category of each bounding box and give as standard
Threshold value comparison filters the lower bounding box of score, and non-maxima suppression processing is carried out to the bounding box of reservation,
Obtain final testing result.
3. a kind of more car plate dynamic identifying methods of real-time traffic scene according to claim 2, which is characterized in that described
There are License Plate network 16 convolutional layers, 2 full articulamentums to be exported end to end based on image full figure information to realize,
Its network structure is followed successively by：Convolutional layer conv1, pond layer max pool, convolutional layer conv2, pond layer max pool, convolutional layer
Conv3, convolutional layer conv4, convolutional layer conv5, convolutional layer conv6, pond layer max pool, convolutional layer conv7, convolutional layer
Conv8, convolutional layer conv9, convolutional layer conv10, pond layer max pool, convolutional layer conv11, convolutional layer conv12, convolution
Layer conv13, convolutional layer conv14, convolutional layer conv15, convolutional layer conv16, full articulamentum FC1 and full articulamentum FC2.
4. a kind of more car plate dynamic identifying methods of real-time traffic scene according to claim 1, which is characterized in that described
Car license recognition network further includes character recognition network algorithm, and it includes 5 convolutional layers, 3 full articulamentums, and network structure is successively
For：Convolutional layer conv1, pond layer max pool, convolutional layer conv2, pond layer max pool, convolutional layer conv3, convolutional layer
Conv4, convolutional layer conv5, pond layer max pool, full articulamentum FC1, full articulamentum FC2 and full articulamentum FC3, wherein complete
Articulamentum FC3 dimensions are divided into 7 parts, for follow-up softmax classification characters.
5. a kind of more car plate dynamic identifying methods of real-time traffic scene according to claim 4, which is characterized in that described
The full articulamentum FC3 dimensions of character recognition network structure are divided into 7 parts：Using 7 of exportable 7 label
Softmax graders respectively classify to character, and each softmax graders are responsible for exporting the character of car plate corresponding position,
Each the probability of grader output character is：
And the output character probability of each softmax graders adds up to 1, i.e.,：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810368034.4A CN108509954A (en) | 2018-04-23 | 2018-04-23 | A kind of more car plate dynamic identifying methods of real-time traffic scene |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810368034.4A CN108509954A (en) | 2018-04-23 | 2018-04-23 | A kind of more car plate dynamic identifying methods of real-time traffic scene |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108509954A true CN108509954A (en) | 2018-09-07 |
Family
ID=63383040
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810368034.4A Pending CN108509954A (en) | 2018-04-23 | 2018-04-23 | A kind of more car plate dynamic identifying methods of real-time traffic scene |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108509954A (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109447171A (en) * | 2018-11-05 | 2019-03-08 | 电子科技大学 | A kind of vehicle attitude classification method based on deep learning |
| CN109460720A (en) * | 2018-10-25 | 2019-03-12 | 中科院成都信息技术股份有限公司 | Ballot paper recognition methods based on convolutional neural networks |
| CN109726739A (en) * | 2018-12-04 | 2019-05-07 | 深圳大学 | A kind of object detection method and system |
| CN109993138A (en) * | 2019-04-08 | 2019-07-09 | 北京易华录信息技术股份有限公司 | A kind of car plate detection and recognition methods and device |
| CN110097044A (en) * | 2019-05-13 | 2019-08-06 | 苏州大学 | Stage car plate detection recognition methods based on deep learning |
| CN110163153A (en) * | 2019-05-23 | 2019-08-23 | 百度在线网络技术（北京）有限公司 | The method and device on traffic mark board boundary for identification |
| CN110245664A (en) * | 2019-06-26 | 2019-09-17 | 深兰科技（上海）有限公司 | Licence plate recognition method |
| CN110322509A (en) * | 2019-06-26 | 2019-10-11 | 重庆邮电大学 | Object localization method, system and computer equipment based on level Class Activation figure |
| CN110321853A (en) * | 2019-07-05 | 2019-10-11 | 杭州巨骐信息科技股份有限公司 | Distribution cable external force damage prevention system based on video intelligent detection |
| CN110348446A (en) * | 2019-06-24 | 2019-10-18 | 西安艾润物联网技术服务有限责任公司 | Image processing method and device |
| CN110532904A (en) * | 2019-08-13 | 2019-12-03 | 桂林电子科技大学 | A kind of vehicle identification method |
| CN110659632A (en) * | 2019-09-29 | 2020-01-07 | 公安部交通管理科学研究所 | System and method for testing motor vehicle number plate identification performance of traffic technology monitoring equipment based on image block assignment |
| CN110929726A (en) * | 2020-02-11 | 2020-03-27 | 南京智莲森信息技术有限公司 | Railway contact network support number plate identification method and system |
| CN111027526A (en) * | 2019-10-25 | 2020-04-17 | 深圳羚羊极速科技有限公司 | Method for improving vehicle target detection, identification and detection efficiency |
| CN111414855A (en) * | 2020-03-19 | 2020-07-14 | 国网陕西省电力公司电力科学研究院 | Telegraph pole sign target detection and identification method based on end-to-end regression model |
| US20200410282A1 (en) * | 2019-06-27 | 2020-12-31 | Robert Bosch Gmbh | Method for determining a confidence value of an object of a class |
| CN112329768A (en) * | 2020-10-23 | 2021-02-05 | 上善智城(苏州)信息科技有限公司 | Improved YOLO-based method for identifying fuel-discharging stop sign of gas station |
| CN112560578A (en) * | 2020-11-18 | 2021-03-26 | 北京理工大学 | Imaging-free license plate content identification method and system |
| CN112949754A (en) * | 2021-03-29 | 2021-06-11 | 中国科学院合肥物质科学研究院 | Text recognition data synthesis method based on image fusion |
| CN113205107A (en) * | 2020-11-02 | 2021-08-03 | 哈尔滨理工大学 | Vehicle type recognition method based on improved high-efficiency network |
| CN113221887A (en) * | 2021-05-18 | 2021-08-06 | 上海新纪元机器人有限公司 | License plate recognition method and device |
| CN113344037A (en) * | 2021-05-18 | 2021-09-03 | 国网江西省电力有限公司电力科学研究院 | Cable insulation layer and sheath parameter measuring method and measuring device |
| CN113792739A (en) * | 2021-08-25 | 2021-12-14 | 电子科技大学 | Universal license plate text recognition method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000251016A (en) * | 1999-03-02 | 2000-09-14 | Hitachi Ltd | Number plate recognizing device |
| CN104217214A (en) * | 2014-08-21 | 2014-12-17 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Configurable convolutional neural network based red green blue-distance (RGB-D) figure behavior identification method |
| CN105335743A (en) * | 2015-10-28 | 2016-02-17 | 重庆邮电大学 | Vehicle license plate recognition method |
| CN106971185A (en) * | 2017-03-27 | 2017-07-21 | 深圳市丰巨泰科电子有限公司 | A kind of license plate locating method and device based on full convolutional network |
| CN107038442A (en) * | 2017-03-27 | 2017-08-11 | 新智认知数据服务有限公司 | A kind of car plate detection and global recognition method based on deep learning |
| CN107220603A (en) * | 2017-05-18 | 2017-09-29 | 惠龙易通国际物流股份有限公司 | Vehicle checking method and device based on deep learning |
-
2018
- 2018-04-23 CN CN201810368034.4A patent/CN108509954A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000251016A (en) * | 1999-03-02 | 2000-09-14 | Hitachi Ltd | Number plate recognizing device |
| CN104217214A (en) * | 2014-08-21 | 2014-12-17 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | Configurable convolutional neural network based red green blue-distance (RGB-D) figure behavior identification method |
| CN105335743A (en) * | 2015-10-28 | 2016-02-17 | 重庆邮电大学 | Vehicle license plate recognition method |
| CN106971185A (en) * | 2017-03-27 | 2017-07-21 | 深圳市丰巨泰科电子有限公司 | A kind of license plate locating method and device based on full convolutional network |
| CN107038442A (en) * | 2017-03-27 | 2017-08-11 | 新智认知数据服务有限公司 | A kind of car plate detection and global recognition method based on deep learning |
| CN107220603A (en) * | 2017-05-18 | 2017-09-29 | 惠龙易通国际物流股份有限公司 | Vehicle checking method and device based on deep learning |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109460720A (en) * | 2018-10-25 | 2019-03-12 | 中科院成都信息技术股份有限公司 | Ballot paper recognition methods based on convolutional neural networks |
| CN109447171A (en) * | 2018-11-05 | 2019-03-08 | 电子科技大学 | A kind of vehicle attitude classification method based on deep learning |
| CN109726739A (en) * | 2018-12-04 | 2019-05-07 | 深圳大学 | A kind of object detection method and system |
| CN109993138A (en) * | 2019-04-08 | 2019-07-09 | 北京易华录信息技术股份有限公司 | A kind of car plate detection and recognition methods and device |
| CN110097044A (en) * | 2019-05-13 | 2019-08-06 | 苏州大学 | Stage car plate detection recognition methods based on deep learning |
| CN110163153A (en) * | 2019-05-23 | 2019-08-23 | 百度在线网络技术（北京）有限公司 | The method and device on traffic mark board boundary for identification |
| US11003926B2 (en) | 2019-05-23 | 2021-05-11 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method and apparatus for recognizing boundary of traffic sign |
| CN110163153B (en) * | 2019-05-23 | 2022-01-04 | 阿波罗智能技术(北京)有限公司 | Method and device for recognizing traffic sign board boundary |
| CN110348446A (en) * | 2019-06-24 | 2019-10-18 | 西安艾润物联网技术服务有限责任公司 | Image processing method and device |
| CN110245664A (en) * | 2019-06-26 | 2019-09-17 | 深兰科技（上海）有限公司 | Licence plate recognition method |
| CN110322509A (en) * | 2019-06-26 | 2019-10-11 | 重庆邮电大学 | Object localization method, system and computer equipment based on level Class Activation figure |
| CN110322509B (en) * | 2019-06-26 | 2021-11-12 | 重庆邮电大学 | Target positioning method, system and computer equipment based on hierarchical class activation graph |
| US20200410282A1 (en) * | 2019-06-27 | 2020-12-31 | Robert Bosch Gmbh | Method for determining a confidence value of an object of a class |
| US11531832B2 (en) * | 2019-06-27 | 2022-12-20 | Robert Bosch Gmbh | Method for determining a confidence value of an object of a class |
| CN110321853B (en) * | 2019-07-05 | 2021-05-11 | 杭州巨骐信息科技股份有限公司 | Distributed cable external-damage-prevention system based on video intelligent detection |
| CN110321853A (en) * | 2019-07-05 | 2019-10-11 | 杭州巨骐信息科技股份有限公司 | Distribution cable external force damage prevention system based on video intelligent detection |
| CN110532904A (en) * | 2019-08-13 | 2019-12-03 | 桂林电子科技大学 | A kind of vehicle identification method |
| CN110659632A (en) * | 2019-09-29 | 2020-01-07 | 公安部交通管理科学研究所 | System and method for testing motor vehicle number plate identification performance of traffic technology monitoring equipment based on image block assignment |
| CN111027526A (en) * | 2019-10-25 | 2020-04-17 | 深圳羚羊极速科技有限公司 | Method for improving vehicle target detection, identification and detection efficiency |
| CN110929726A (en) * | 2020-02-11 | 2020-03-27 | 南京智莲森信息技术有限公司 | Railway contact network support number plate identification method and system |
| CN111414855A (en) * | 2020-03-19 | 2020-07-14 | 国网陕西省电力公司电力科学研究院 | Telegraph pole sign target detection and identification method based on end-to-end regression model |
| CN111414855B (en) * | 2020-03-19 | 2023-03-24 | 国网陕西省电力公司电力科学研究院 | Telegraph pole sign target detection and identification method based on end-to-end regression model |
| CN112329768A (en) * | 2020-10-23 | 2021-02-05 | 上善智城(苏州)信息科技有限公司 | Improved YOLO-based method for identifying fuel-discharging stop sign of gas station |
| CN113205107A (en) * | 2020-11-02 | 2021-08-03 | 哈尔滨理工大学 | Vehicle type recognition method based on improved high-efficiency network |
| CN112560578A (en) * | 2020-11-18 | 2021-03-26 | 北京理工大学 | Imaging-free license plate content identification method and system |
| CN112560578B (en) * | 2020-11-18 | 2022-12-20 | 北京理工大学 | Imaging-free license plate content identification method and system |
| CN112949754A (en) * | 2021-03-29 | 2021-06-11 | 中国科学院合肥物质科学研究院 | Text recognition data synthesis method based on image fusion |
| CN112949754B (en) * | 2021-03-29 | 2022-10-14 | 中国科学院合肥物质科学研究院 | Text recognition data synthesis method based on image fusion |
| CN113221887A (en) * | 2021-05-18 | 2021-08-06 | 上海新纪元机器人有限公司 | License plate recognition method and device |
| CN113344037A (en) * | 2021-05-18 | 2021-09-03 | 国网江西省电力有限公司电力科学研究院 | Cable insulation layer and sheath parameter measuring method and measuring device |
| CN113792739A (en) * | 2021-08-25 | 2021-12-14 | 电子科技大学 | Universal license plate text recognition method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108509954A (en) | A kind of more car plate dynamic identifying methods of real-time traffic scene | |
| CN108830188B (en) | Vehicle detection method based on deep learning | |
| CN109977812B (en) | Vehicle-mounted video target detection method based on deep learning | |
| Wang et al. | RENet: Rectangular convolution pyramid and edge enhancement network for salient object detection of pavement cracks | |
| CN111784685A (en) | Power transmission line defect image identification method based on cloud edge cooperative detection | |
| CN109800736A (en) | A kind of method for extracting roads based on remote sensing image and deep learning | |
| CN103258213B (en) | A kind of for the dynamic vehicle model recognizing method in intelligent transportation system | |
| CN111079640B (en) | Vehicle type identification method and system based on automatic amplification sample | |
| CN108491797A (en) | A kind of vehicle image precise search method based on big data | |
| CN107085696A (en) | A kind of vehicle location and type identifier method based on bayonet socket image | |
| CN109816024A (en) | A kind of real-time automobile logo detection method based on multi-scale feature fusion and DCNN | |
| CN105574550A (en) | Vehicle identification method and device | |
| CN106408015A (en) | Road fork identification and depth estimation method based on convolutional neural network | |
| CN105809121A (en) | Multi-characteristic synergic traffic sign detection and identification method | |
| CN110532961B (en) | Semantic traffic light detection method based on multi-scale attention mechanism network model | |
| CN109034024B (en) | Logistics vehicle type classification and identification method based on image target detection | |
| CN109993138A (en) | A kind of car plate detection and recognition methods and device | |
| CN109886147A (en) | A kind of more attribute detection methods of vehicle based on the study of single network multiple-task | |
| CN109815945A (en) | A kind of respiratory tract inspection result interpreting system and method based on image recognition | |
| CN108268865A (en) | Licence plate recognition method and system under a kind of natural scene based on concatenated convolutional network | |
| CN103679214A (en) | Vehicle detection method based on online area estimation and multi-feature decision fusion | |
| CN113807347A (en) | Kitchen waste impurity identification method based on target detection technology | |
| CN112785610B (en) | Lane line semantic segmentation method integrating low-level features | |
| CN106548195A (en) | A kind of object detection method based on modified model HOG ULBP feature operators | |
| CN104331708B (en) | A kind of zebra crossing automatic detection analysis method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180907 |