CN108446696A - A kind of end-to-end licence plate recognition method based on deep learning - Google Patents
A kind of end-to-end licence plate recognition method based on deep learning Download PDFInfo
- Publication number
- CN108446696A CN108446696A CN201810132848.8A CN201810132848A CN108446696A CN 108446696 A CN108446696 A CN 108446696A CN 201810132848 A CN201810132848 A CN 201810132848A CN 108446696 A CN108446696 A CN 108446696A
- Authority
- CN
- China
- Prior art keywords
- car plate
- image
- plate
- car
- layers
- 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
- G06V20/635—Overlay text, e.g. embedded captions in a TV program
-
- 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
- G07—CHECKING-DEVICES
- G07B—TICKET-ISSUING APPARATUS; FARE-REGISTERING APPARATUS; FRANKING APPARATUS
- G07B15/00—Arrangements or apparatus for collecting fares, tolls or entrance fees at one or more control points
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Finance (AREA)
- Image Analysis (AREA)
Abstract
The present invention proposes a kind of end-to-end licence plate recognition method based on deep learning, and the image of car plate to be detected is exported after gauss hybrid models algorithm process；Convolutional neural networks extract sharing feature module, by the image input of car plate to be detected, trained pulleying product neural network algorithm extracts feature, output regression location information, recurrence angle character mapping graph and the corresponding Feature Mapping figure for discriminating whether car plate after being handled using license plate image preliminary detection module；Car plate preliminary detection module obtains confidence score by the corresponding Feature Mapping map analysis for discriminating whether car plate, obtains a collection of candidate license plate image, carries out license plate image fusion using non-maxima suppression algorithm, finally obtains the image of true car plate position；The image of true car plate position is input to car plate content recognition feedback module and identifies car plate content.The present invention is applied in managing system of car parking, promotes parking fee collective system efficiency, improves Car license recognition efficiency, saved cost.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of end-to-end Car license recognition side based on deep learning
Method.
Background technology
Existing detection method of license plate is to judge car plate using the feature of license plate area, by license plate area from whole picture vehicle figure
It is split as in.There is car plate itself many inherent features, these features to be different in different countries.China's car plate
With the following feature that can be used for detecting:(1) car plate background color generally has larger difference with body color, character color；(2) car plate
There are one the continuous or discontinuous frames due to abrasion；(3) in car plate character have it is multiple, substantially be in it is horizontally arranged, in licence plate
Rectangular area in there are abundant edge, the textural characteristics of rule are presented；(4) interval in car plate between character is more uniform,
Character and licence plate background color there are larger saltus step, have relatively uniform ash on gray value inside character itself and licence plate bottom
Degree；(5) specific size of licence plate, position are uncertain in different images, but its length-width ratio is in certain variation range.It is above several
Kind feature is all conceptual, and individually apparently all non-license plate image institute is exclusive for various features, but combining them can be only
One ground determines car plate.In these features, color, shape, position feature are the most intuitive, are easy to extract.Textural characteristics compare pumping
As, it is necessary to corresponding available characteristic index can be obtained by passing through certain processing or being converted to other features.
The method of car plate detection is varied at present, and being summed up mainly has method based on analysis of texture, is based on
The method of edge detection is detected based on mathematical morphology, based on wavelet analysis detection etc..
1. the case where method based on edge detection seriously fades to car plate, due to can't detect the edge meeting of character stroke
Cause positioning fail, have external interference and it is license plate sloped when, the region after detection is more slightly larger than car plate；
2. the Detection of License currently with wavelet analysis is mutually tied with other a variety of methods using wavelet transformation mostly
Close realize it is more acurrate, quickly detect car plate, the disadvantages of this method is that speed is slower, and in the larger mistiming localization machine of noise
Rate also increases therewith；
3. traditional analysis of texture detection algorithm is mostly based on gray level image to analyze, therefore algorithm needs pair
Image is pre-processed, and coloured image is converted to gray level image, then into rank scanning, to determine license plate candidate area.It should
Licence plate is tilted or is deformed for class algorithm and uneven illumination, on the weak side or have good effect by force partially, but to noise-sensitive, for
The image of background complexity may not be able to detect license plate area well.
4. cannot accurately determine the position of car plate right boundary based on the license plate area localization method of mathematical morphology.The party
Method must be likely to carry out accurate license plate area inspection in conjunction with other localization methods.
Invention content
In view of the deficiencies of the prior art, the present invention provides a kind of end-to-end Car license recognition side being based on deep learning
Method, solves that traditional there are the training time is too long and weight adjusted adaptation etc. and asks based on Adaboost Detection of License
Topic.
In order to solve the above technical problem, the present invention provides a kind of end-to-end licence plate recognition method based on deep learning,
Including convolutional neural networks extraction sharing feature module, car plate preliminary detection module, car plate content recognition feedback module, further include
Following steps：
S1:The image of car plate to be detected is exported after gauss hybrid models algorithm process；
S2:Convolutional neural networks extract sharing feature module, and by the image input of car plate to be detected, trained pulleying is accumulated
Neural network algorithm extracts feature, output regression location information, recurrence angle after being handled using license plate image preliminary detection module
Spend Feature Mapping figure and the corresponding Feature Mapping figure for discriminating whether car plate；
S3:Car plate preliminary detection module obtains confidence level by the corresponding Feature Mapping map analysis for discriminating whether car plate and obtains
Point, a collection of candidate license plate image is obtained, license plate image fusion is carried out using non-maxima suppression algorithm, finally obtains true vehicle
The image that memorial tablet is set；
S4:The image of true car plate position is input to car plate content recognition feedback module and identifies car plate content.
As a kind of preferred method, the convolutional neural networks structure of the S2 steps extraction sharing feature includes 6 layers, respectively
For 1 input layer, 3 convolutional layers, 2 pond layers.
As a kind of preferred method, the core number of 3 convolutional layers of the convolutional neural networks is respectively 24,48,64, first
The size of the convolution kernel of layer is 5*5, other two layers convolution kernel sizes are all 3*3, and the span of all convolutional layers is all 1*1.
As a kind of preferred method, the S2 steps license plate image preliminary detection module includes a convolutional neural networks,
Using extract sharing feature convolutional neural networks output characteristic pattern as input layer, 5 convolutional layers, 2 pond layers.
As a kind of preferred method, the core number of 5 convolutional layers of the convolutional neural networks is respectively 128,256, first
The core and span of a pond layer are all 2*1, and the core of second pond layer is 3*3, span 2*2；First convolution kernel
Size is 3*3, and second convolution kernel size is 2*2, and the span of all convolutional layers is all 1*1.
As a kind of preferred method, the car plate content recognition feedback module includes a convolutional neural networks, network knot
Structure is 6 layers, 1 input layer, including the input that the sharing feature that extracts of S1 step convolutional neural networks is the module, 1
Anti- pond layer for alleviating fuzzy characteristics, 1 convolutional layer, 1 pond layer, one two-way LSTM layers and 1 full articulamentum.
As a kind of preferred method, the core size of the anti-pond layer of the convolutional neural networks is 2*1, span 2*1；Volume
The core size of lamination is 3*3, span 1*1；The core number of convolutional layer is respectively 64；The core size of pond layer is 1*2, and span is
1*2；Two-way LSTM cores number is that 96, CTC sequential is 26, finally connects a full articulamentum again, for identification car plate content.
Present invention advantageous effect compared with prior art：The present invention is applied in managing system of car parking, promotes parking lot
Charge efficiency meets people's conveniently demand.The present invention improves Car license recognition efficiency, has saved manpower and materials, simultaneously
Ensure entirely to manage system security reliability.
Description of the drawings
Fig. 1 is the method for the present invention flow chart.
Fig. 2 is the shared convolution neural network structure figure of the present invention.
Fig. 3 is car plate preliminary detection module convolutional neural networks structure chart of the present invention.
Fig. 4 is car plate content recognition feedback module convolutional neural networks structure chart of the present invention.
Specific implementation mode
The embodiment of the present invention is further described below.Following embodiment only carries out furtherly the application
It is bright, it should not be construed as the limitation to the application.
As shown in Figure 1, the present invention provides a kind of end-to-end licence plate recognition method embodiment based on deep learning, including volume
Product neural network extraction sharing feature module, car plate preliminary detection module, car plate content recognition feedback module, further include following step
Suddenly：
S1:The image of car plate to be detected is exported after gauss hybrid models algorithm process；
S2:Convolutional neural networks extract sharing feature module, and by the image input of car plate to be detected, trained pulleying is refreshing
Feature is extracted through network algorithm, output regression location information, recurrence angle are special after the processing of license plate image preliminary detection module
Levy mapping graph and the corresponding Feature Mapping figure for discriminating whether car plate；
S3:Car plate preliminary detection module obtains confidence level by the corresponding Feature Mapping map analysis for discriminating whether car plate and obtains
Point, a collection of candidate license plate image is obtained, license plate image fusion is carried out using non-maxima suppression algorithm, finally obtains true vehicle
The image that memorial tablet is set；
S4:The image of true car plate position is input to car plate content recognition feedback module and identifies car plate content.
The present embodiment the specific implementation process is as follows：
One, sample preparation processes：
1, outfield car plate image data is demarcated using car plate calibration tool；
2, positive negative sample and third class sample are obtained in the way of sliding window on picture by calibration information：IOU>0.7,
It is determined as positive sample；IOU<0.1, it is determined as negative sample；Remaining is third class sample, neither positive sample is nor negative sample；
Wherein, IOU is that the system prediction frame come out and the frame that is marked in original picture overlap degree.
3, training image is 72 (width) * 24 (height) * 1 (gray level image).
Two, training network frame：
Step 1：It is special that the sample acquired in sample preparation processes is inputted into volume neural network extraction feature device extraction
Sign, volume neural network extraction are characterized in what car plate preliminary detection module and car plate content recognition feedback module were shared.Wherein,
Car plate preliminary detection module is mainly used for car plate recurrence and whether car plate is classified, and car plate content recognition feedback module helps to assist
And the index of car plate detection is promoted, the volume neural network extraction specific process of feature is as follows：
As shown in Figure 2, wherein conv indicates that convolutional layer, 5*5,3*3 indicate convolution kernel size, and 24,48,64 indicate convolution
Core number, relu indicate that active coating, pool indicate that pond layer, max indicate maximum pond mode, the network of shared volume neural network
Structure is 6 layers of structure：Including 1 input layer, 3 convolutional layers, 2 pond layers, the core number of convolutional layer be respectively 24,48,
64 }, the size of the convolution kernel of first layer is 5*5, other two layers convolution kernel sizes are all 3*3, and the span of all convolutional layers is all
For 1*1.Relu layers of an activation primitive is connect after each convolutional layer, the core and span of first pond layer are all 2*1, the
The core and span of two pond layers are all 2*2.The sample mode of pond layer is all by the way of being maximized.
Step 2：After the volume neural network sharing feature extracted through step 1, due to needing to detect the position of car plate
Confidence ceases and needs discriminate whether it is car plate, wherein and location information includes the location information and angle information returned, therefore I
Propose car plate preliminary detection module, detailed process is as follows：
As shown in Figure 3, wherein conv expression convolutional layers, 1*1,2*2,3*3 expression convolution kernel size, 128,256,8,2,1
Indicate that convolution kernel number, relu indicate that active coating, pool indicate that pond layer, max indicate that maximum pond mode, AVE indicate average
Pond mode.After the volume neural network sharing feature extracted through step 1, the car plate preliminary detection module of our designs
Convolutional neural networks structure is 5 layers：Including the input that the sharing feature that first step volume neural network extracts is the module；Two
A pond layer and two convolutional layers.The core and span of first pond layer are all 2*1, and the core of second pond layer is 3*3,
Span is 2*2.The core number of convolutional layer is respectively { 128,256 }, and the size of first and second convolution kernel is 3*3,
The convolution kernel size of third is 2*2, and the span of all convolutional layers is all 1*1.An active coating Relu is all met behind convolutional layer.
Car plate Preliminary detection finally returns out one group of location information box_fc, angle information theta_fc, whether it is car plate judges
Score_fc finally carries out loss calculating.
Step 3：In order to which Car license recognition is we have proposed car plate content identifier module, specific process is as follows：
As shown in Figure 4, wherein conv indicates that convolutional layer, 3*3 indicate convolution kernel size, and 64 indicate convolution kernel number, relu
Indicate that active coating, pool indicate that pond layer, AVE indicate that average pond mode, unpool indicate that anti-pond layer, blstm indicate double
Indicate that two-way LSTM cores number, fc indicate that full articulamentum, ctc indicate that CTC algorithms are predicted and calculated to train to Lstm, 96
The loss of journey.After the volume neural network sharing feature extracted through step 1, the convolution for the Car license recognition module that we design
Neural network structure is 6 layers, 1 input layer, including the sharing feature that first step volume neural network extracts is the module
Input, 1 anti-pond layer for alleviating fuzzy characteristics, 1 convolutional layer, 1 pond layer, one two-way LSTM layers and 1
Full articulamentum.The core size of wherein anti-pond layer is 2*1, span 2*1；The core size of convolutional layer is 3*3, span 1*1；Volume
The core number of lamination is respectively { 64 }；The core size of pond layer is 1*2, span 1*2.BLSTM refers to two-way LSTM, core
It is 26 that number, which is 96, CTC sequential, finally connects a full articulamentum again, for identification car plate content.Assuming that 70 points in total of car plate content
Class, class categories are 70 classes, including 10 numbers, 35 province Chinese, 24 letters and 1 space.Finally utilize CTC algorithms
It carries out prediction and loss is calculated, while the accuracy of feedback optimized car plate detection, anti-pond layer can solve to a certain extent
Certainly car plate fuzzy problem.
Step 4：The calculating process of training process loss：For positive sample, it would be desirable to calculate four loss：Step 2
In calculate smoothLN Loss using the recurrence position that the location information and prediction of calibration obtain when finally calculate loss,
Using the angle information of calibration and it is revert to angle calculation Euclidean distance Loss when finally calculating loss in step 2, walked
Sentenced using the positive and negative sample information of calibration and the positive and negative sample class predicted calculating when finally calculating loss in rapid two
In the softmax Loss and step 3 of not positive negative sample calculate loss when using calibration car plate content and predict come
Car plate content calculate car plate content recognition CTC Loss；For negative sample：We, which calculate, returns not positive and negative sample in step 2
The CTC Loss of car plate content recognition in this softmax Loss and step 3；For third class sample, we only calculate
The CTC Loss of car plate content recognition, wherein smoothLN calculation formula are as follows in step 3：
smoothLn(x)=(| d |+1) ln (| d |+1)-| d |
CTC is decoded, and T indicates that sequential, t indicate sometime, and y indicates that the output of sometime t, L indicate set, class of classifying
Not Wei 70 classes, including 10 numbers, 35 province Chinese, 24 letters and 1 space.Wherein CTC Loss calculation formula are such as
Under：
h(x)≈B(π*)
It the above is only the preferred embodiment of the present invention, protection scope of the present invention is not limited to above-described embodiment, all categories
Technical solution under thinking of the present invention belongs to the scope of the present invention.It should be pointed out that for the common skill of the art
For art personnel, several improvements and modifications without departing from the principles of the present invention, these improvements and modifications also should be regarded as this
The protection domain of invention.
Claims (7)
1. a kind of end-to-end licence plate recognition method based on deep learning, which is characterized in that altogether including convolutional neural networks extraction
Characteristic module, car plate preliminary detection module, car plate content recognition feedback module are enjoyed, it is further comprising the steps of：
S1:The image of car plate to be detected is exported after gauss hybrid models algorithm process；
S2:Convolutional neural networks extract sharing feature module, and by the image input of car plate to be detected, nerve is accumulated in trained pulleying
Network algorithm extracts feature, and output regression location information, recurrence angle are special after being handled using license plate image preliminary detection module
Levy mapping graph and the corresponding Feature Mapping figure for discriminating whether car plate；
S3:Car plate preliminary detection module obtains confidence score by the corresponding Feature Mapping map analysis for discriminating whether car plate,
A collection of candidate license plate image is obtained, license plate image fusion is carried out using non-maxima suppression algorithm, finally obtains true car plate
The image of position；
S4:The image of true car plate position is input to car plate content recognition feedback module and identifies car plate content.
2. a kind of end-to-end licence plate recognition method based on deep learning as described in claim 1, which is characterized in that the S2
The convolutional neural networks structure that step extracts sharing feature includes 6 layers, respectively 1 input layer, 3 convolutional layers, 2 ponds
Layer.
3. a kind of end-to-end licence plate recognition method based on deep learning as claimed in claim 2, which is characterized in that the volume
The core number of product 3 convolutional layers of neural network is respectively 24,48,64, and the size of the convolution kernel of first layer is 5*5, other two layers
Convolution kernel size be all 3*3, the span of all convolutional layers is all 1*1.
4. a kind of end-to-end licence plate recognition method based on deep learning as described in claim 1, which is characterized in that the S2
Step license plate image preliminary detection module include a convolutional neural networks, with extract sharing feature convolutional neural networks it is defeated
It is input layer, 5 convolutional layers, 2 pond layers to go out characteristic pattern.
5. a kind of end-to-end licence plate recognition method based on deep learning as claimed in claim 4, which is characterized in that the volume
The core number of product neural network 5 convolutional layers is respectively 128,256, and the core and span of first pond layer are all 2*1, and second
The core of a pond layer is 3*3, span 2*2；The size of first convolution kernel is 3*3, and second convolution kernel size is 2*
2, the span of all convolutional layers is all 1*1.
6. a kind of end-to-end licence plate recognition method based on deep learning as described in claim 1, which is characterized in that the vehicle
Board content recognition feedback module includes a convolutional neural networks, and network structure is 6 layers, 1 input layer, including S1 step convolution
The sharing feature that neural network extracts is the input of the module, 1 anti-pond layer for alleviating fuzzy characteristics, 1 volume
Lamination, 1 pond layer, one two-way LSTM layers and 1 full articulamentum.
7. a kind of end-to-end licence plate recognition method based on deep learning as claimed in claim 6, which is characterized in that the volume
The core size of the anti-pond layer of product neural network is 2*1, span 2*1；The core size of convolutional layer is 3*3, span 1*1；Volume
The core number of lamination is respectively 64；The core size of pond layer is 1*2, span 1*2；Two-way LSTM cores number is 96, CTC sequential
It is 26, finally connects a full articulamentum again, for identification car plate content.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810132848.8A CN108446696A (en) | 2018-02-09 | 2018-02-09 | A kind of end-to-end licence plate recognition method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810132848.8A CN108446696A (en) | 2018-02-09 | 2018-02-09 | A kind of end-to-end licence plate recognition method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108446696A true CN108446696A (en) | 2018-08-24 |
Family
ID=63192104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810132848.8A Pending CN108446696A (en) | 2018-02-09 | 2018-02-09 | A kind of end-to-end licence plate recognition method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108446696A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109493336A (en) * | 2018-11-14 | 2019-03-19 | 上海艾策通讯科技股份有限公司 | Video mosaic based on artificial intelligence identifies the system and method learnt automatically |
| CN109784340A (en) * | 2018-12-14 | 2019-05-21 | 北京市首都公路发展集团有限公司 | A kind of licence plate recognition method and device |
| CN110414451A (en) * | 2019-07-31 | 2019-11-05 | 深圳市捷顺科技实业股份有限公司 | It is a kind of based on end-to-end licence plate recognition method, device, equipment and storage medium |
| CN110874556A (en) * | 2018-09-04 | 2020-03-10 | 上海集光安防科技股份有限公司 | License plate detecting system in natural scene based on deep learning |
| CN111027534A (en) * | 2018-10-09 | 2020-04-17 | 杭州海康威视数字技术股份有限公司 | Compact double-license-plate detection method and device |
| CN111127440A (en) * | 2019-12-22 | 2020-05-08 | 上海眼控科技股份有限公司 | Tire specification detection method and device, electronic equipment and readable storage medium |
| CN112001386A (en) * | 2020-08-31 | 2020-11-27 | 重庆紫光华山智安科技有限公司 | License plate character recognition method, system, medium and terminal |
| CN112560856A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | License plate detection and identification method, device, equipment and storage medium |
| CN113344005A (en) * | 2021-05-12 | 2021-09-03 | 武汉大学 | Image edge detection method based on optimized small-scale features |
| 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 |
| CN117173416A (en) * | 2023-11-01 | 2023-12-05 | 山西阳光三极科技股份有限公司 | Railway freight train number image definition processing method based on image processing |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022232A (en) * | 2016-05-12 | 2016-10-12 | 成都新舟锐视科技有限公司 | License plate detection method based on deep learning |
| CN107229929A (en) * | 2017-04-12 | 2017-10-03 | 西安电子科技大学 | A kind of license plate locating method based on R CNN |
-
2018
- 2018-02-09 CN CN201810132848.8A patent/CN108446696A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022232A (en) * | 2016-05-12 | 2016-10-12 | 成都新舟锐视科技有限公司 | License plate detection method based on deep learning |
| CN107229929A (en) * | 2017-04-12 | 2017-10-03 | 西安电子科技大学 | A kind of license plate locating method based on R CNN |
Non-Patent Citations (2)
| Title |
|---|
| ALEX GRAVES等: "A Novel Connectionist System for Unconstrained Handwriting Recognition", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| HUI LI等: "Towards End-to-End Car License Plates Detection and Recognition with Deep Neural Networks", 《COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110874556A (en) * | 2018-09-04 | 2020-03-10 | 上海集光安防科技股份有限公司 | License plate detecting system in natural scene based on deep learning |
| CN110874556B (en) * | 2018-09-04 | 2024-02-09 | 上海集光安防科技股份有限公司 | License plate detection system in natural scene based on deep learning |
| CN111027534A (en) * | 2018-10-09 | 2020-04-17 | 杭州海康威视数字技术股份有限公司 | Compact double-license-plate detection method and device |
| CN111027534B (en) * | 2018-10-09 | 2023-05-26 | 杭州海康威视数字技术股份有限公司 | Compact double license plate detection method and device |
| CN109493336B (en) * | 2018-11-14 | 2022-03-04 | 上海艾策通讯科技股份有限公司 | System and method for video mosaic identification automatic learning based on artificial intelligence |
| CN109493336A (en) * | 2018-11-14 | 2019-03-19 | 上海艾策通讯科技股份有限公司 | Video mosaic based on artificial intelligence identifies the system and method learnt automatically |
| CN109784340A (en) * | 2018-12-14 | 2019-05-21 | 北京市首都公路发展集团有限公司 | A kind of licence plate recognition method and device |
| CN110414451A (en) * | 2019-07-31 | 2019-11-05 | 深圳市捷顺科技实业股份有限公司 | It is a kind of based on end-to-end licence plate recognition method, device, equipment and storage medium |
| CN110414451B (en) * | 2019-07-31 | 2023-11-10 | 深圳市捷顺科技实业股份有限公司 | License plate recognition method, device, equipment and storage medium based on end-to-end |
| CN111127440A (en) * | 2019-12-22 | 2020-05-08 | 上海眼控科技股份有限公司 | Tire specification detection method and device, electronic equipment and readable storage medium |
| CN112001386B (en) * | 2020-08-31 | 2021-06-15 | 重庆紫光华山智安科技有限公司 | License plate character recognition method, system, medium and terminal |
| CN112001386A (en) * | 2020-08-31 | 2020-11-27 | 重庆紫光华山智安科技有限公司 | License plate character recognition method, system, medium and terminal |
| CN112560856A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | License plate detection and identification method, device, equipment and storage medium |
| CN112560856B (en) * | 2020-12-18 | 2024-04-12 | 深圳赛安特技术服务有限公司 | License plate detection and identification method, device, equipment and storage medium |
| CN113344005B (en) * | 2021-05-12 | 2022-04-15 | 武汉大学 | Image edge detection method based on optimized small-scale features |
| CN113344005A (en) * | 2021-05-12 | 2021-09-03 | 武汉大学 | Image edge detection method based on optimized small-scale features |
| 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 |
| CN117173416A (en) * | 2023-11-01 | 2023-12-05 | 山西阳光三极科技股份有限公司 | Railway freight train number image definition processing method based on image processing |
| CN117173416B (en) * | 2023-11-01 | 2024-01-05 | 山西阳光三极科技股份有限公司 | Railway freight train number image definition processing method based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108446696A (en) | A kind of end-to-end licence plate recognition method based on deep learning | |
| CN107229929A (en) | A kind of license plate locating method based on R CNN | |
| Ni et al. | An improved deep network-based scene classification method for self-driving cars | |
| Frizzi et al. | Convolutional neural network for video fire and smoke detection | |
| Yin et al. | Hot region selection based on selective search and modified fuzzy C-means in remote sensing images | |
| CN103218831B (en) | A kind of video frequency motion target classifying identification method based on profile constraint | |
| CN107886080A (en) | One kind is parked position detecting method | |
| CN111079640B (en) | Vehicle type identification method and system based on automatic amplification sample | |
| CN106803301A (en) | A kind of recognition of face guard method and system based on deep learning | |
| CN105844621A (en) | Method for detecting quality of printed matter | |
| CN107220603A (en) | Vehicle checking method and device based on deep learning | |
| CN108647700A (en) | Multitask vehicle part identification model based on deep learning, method and system | |
| CN110119726A (en) | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model | |
| CN109635784A (en) | Traffic sign recognition method based on improved convolutional neural networks | |
| CN107092876A (en) | The low-light (level) model recognizing method combined based on Retinex with S SIFT features | |
| CN106407951B (en) | A kind of night front vehicles detection method based on monocular vision | |
| CN104239905A (en) | Moving target recognition method and intelligent elevator billing system having moving target recognition function | |
| CN109255326A (en) | A kind of traffic scene smog intelligent detecting method based on multidimensional information Fusion Features | |
| CN107292933A (en) | A kind of vehicle color identification method based on BP neural network | |
| CN109508659A (en) | A kind of face identification system and method for crossing | |
| CN103761747A (en) | Target tracking method based on weighted distribution field | |
| CN110473168A (en) | A kind of Lung neoplasm automatic checkout system squeezing-motivate residual error network based on 3D | |
| Zhang et al. | Road marking segmentation based on siamese attention module and maximum stable external region | |
| Tao et al. | Smoky vehicle detection based on range filtering on three orthogonal planes and motion orientation histogram | |
| Wang et al. | License plate location algorithm based on edge detection and morphology |
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: 20180824 |