CN107784303A - Licence plate recognition method and device - Google Patents
Licence plate recognition method and device Download PDFInfo
- Publication number
- CN107784303A CN107784303A CN201611158854.8A CN201611158854A CN107784303A CN 107784303 A CN107784303 A CN 107784303A CN 201611158854 A CN201611158854 A CN 201611158854A CN 107784303 A CN107784303 A CN 107784303A
- Authority
- CN
- China
- Prior art keywords
- license plate
- recognition
- classification value
- plate image
- value sequence
- 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
- 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)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention is used in field of image recognition, there is provided a kind of licence plate recognition method and device, including：Obtain license plate image；Multiple characteristic vectors in license plate image are obtained by convolutional neural networks model；Multiple characteristic vectors are sequentially input into Recognition with Recurrent Neural Network model, to obtain classification value sequence, classification value sequence includes multiple classification values, and each classification value corresponds to a character in car plate；The character according to corresponding to classification value sequence determines license plate image.The present invention directly extracts the characteristic vector in license plate image, and is classified, and can be avoided carrying out cutting and separated identification to the character of car plate, improved recognition speed with the whole car plate of Direct Recognition.
Description
Technical field
The invention belongs to field of image recognition, more particularly to a kind of licence plate recognition method and device.
Background technology
In the prior art, Car license recognition is always one of study hotspot and key technology of Intelligent traffic management systems.Car
The application of board identification is quite varied, such as highway automatic charging, parking lot fee collection management, criminal investigation detection etc..If make
These car plates are identified mode manually, certainly will waste many manpower and materials.
Traditional license plate character recognition method, it is necessary first to the car plate navigated to is subjected to Character segmentation, then to single
Character extracts character feature, finally carries out pattern-recognition.This method execution step is more, and recognition speed is slower.
The slow problem of car plate is identified for prior art, industry does not have preferable settling mode at present.
The content of the invention
Present invention aims at provide a kind of licence plate recognition method and device, it is intended to solves prior art identification car plate speed
The problem of slower.
The invention provides a kind of licence plate recognition method, this method includes：
Obtain license plate image；
Multiple characteristic vectors in the license plate image are obtained by convolutional neural networks model；
The multiple characteristic vector is sequentially input into Recognition with Recurrent Neural Network model, to obtain classification value sequence, the classification
Value sequence includes multiple classification values, and each classification value corresponds to a character in car plate；
The character according to corresponding to the classification value sequence determines the license plate image.
Present invention also offers a kind of license plate recognition device, the device includes：
First acquisition module, for obtaining license plate image；
Second acquisition module, for obtained by convolutional neural networks model multiple features in the license plate image to
Amount；
3rd acquisition module, for the multiple characteristic vector to be sequentially input into Recognition with Recurrent Neural Network model, to be divided
Class value sequence, the classification value sequence include multiple classification values, and each classification value corresponds to a character in car plate；
Determining module, for character corresponding to determining the license plate image according to the sort-type.
The present invention directly extracts the characteristic vector in license plate image, and is classified, and can be kept away with the whole car plate of Direct Recognition
Exempt to carry out cutting and separated identification to the character of car plate, improve recognition speed.
Brief description of the drawings
Fig. 1 is the flow chart of licence plate recognition method provided in an embodiment of the present invention；
Fig. 2 is the flow chart of the method provided in an embodiment of the present invention for obtaining multiple characteristic vectors；
Fig. 3 is the flow chart of the method provided in an embodiment of the present invention for obtaining classification value sequence；
Fig. 4 is the structured flowchart of license plate recognition device provided in an embodiment of the present invention.
Embodiment
In order that technical problems, technical solutions and advantageous effects to be solved by the present invention are more clearly understood, below in conjunction with
Drawings and Examples, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used
To explain the present invention, it is not intended to limit the present invention.
The embodiments of the invention provide a kind of licence plate recognition method, Fig. 1 is Car license recognition side provided in an embodiment of the present invention
The flow chart of method, as shown in figure 1, the method comprising the steps of S101 to step S104.
Step S101, obtain license plate image.
License plate image is to include clearly number-plate number image.In order to quickly identify car plate, the car plate in the present embodiment
Image can be the gray scale pictures converted by Color License Plate photo.Further, based on experience and experiment effect, this implementation
The height of license plate image in example can be 32 pixels.
Step S102, multiple characteristic vectors in license plate image are obtained by convolutional neural networks model.
Convolutional neural networks model is a kind of deep learning framework, can be with efficient identification image, and the framework is avoided to figure
The pretreatment complicated early stage of picture, can directly input license plate image, obtain characteristic vector.
Fig. 2 is the flow chart of the method provided in an embodiment of the present invention for obtaining multiple characteristic vectors, including step S201 and
Step S202.Above-mentioned 2 steps are the detailed descriptions to step S102.
Step S201, by the convolutional layer of convolutional neural networks model, multiple characteristic patterns corresponding to generation license plate image.
Convolutional neural networks model includes convolutional layer and pond layer (pooling), and general convolutional layer is used for the extraction of feature,
Pond layer is used to reduce computational complexity.The specifically used multiple-level stack of convolutional layer and Max-pooling layers of the present embodiment, it is right
The car plate picture extraction feature representation of input.
Step S202, extracts N number of characteristic vector from multiple characteristic patterns, and ith feature vector is i-th in multiple characteristic patterns
The amalgamation result of column vector, wherein, N and i are positive integer, and N is more than or equal to i.
After convolutional layer, license plate image carries out convolutional calculation with multiple convolution kernels in convolutional layer, more so as to generate
Characteristic pattern is opened, multiple characteristic vectors can be extracted from this multiple characteristic pattern, each characteristic vector is by row by characteristic pattern
Generate from left to right.That is, ith feature vector is exactly the merging that all characteristic patterns i-th arrange.Also, each row pair
Answering the perception domain of a rectangle of license plate image, (artwork passes through multilayer convolution, and the size of obtained characteristic pattern will than artwork
It is small), equivalent to some rectangular area of artwork, after multiple convolution, become a column vector, therefore characteristic vector can be by
Regard the avatars to some region as.
Table 1 is the configuration of convolutional neural networks model.Wherein, k represents the size of core, behalf step-length, and p is represented and filled up big
It is small.Design parameter is as shown in table 1：
Table 1
Wherein, core, step-length and to fill up size be all the basic conception in convolutional neural networks.Core is the convolution of convolutional layer
Core；When step-length represents convolution operation, just a convolutional calculation is carried out with convolution kernel at interval of how many individual pixels；Fill up size table
Show and add how many individual pixels on the border of artwork.
Step S103, multiple characteristic vectors are sequentially input into Recognition with Recurrent Neural Network model, to obtain classification value sequence, classification
Value sequence includes multiple classification values, and each classification value corresponds to a character in car plate.
Recognition with Recurrent Neural Network model can be used for handling the object being sequentially generated, such as voice, picture or word.Pass through
The model, the classification to feature can be obtained.
Fig. 3 is the flow chart of the method provided in an embodiment of the present invention for obtaining classification value sequence, including step S301 and step
Rapid S303.Above-mentioned 3 steps are the detailed descriptions to step S103.
Step S301, calculate and layer state is hidden in Cyclic Operation Network, to obtain the pass of the connection between multiple characteristic vectors
System.
Specifically, the present embodiment can be remembered (Long Short Term Memory, referred to as LSTM) by shot and long term
Unit of the unit as Recognition with Recurrent Neural Network model, to calculate hiding layer state.Each LSTM units by a location mode and
Three thresholding compositions, it is to forget thresholding, input threshold and output thresholding respectively.This is a Rotating fields of Recognition with Recurrent Neural Network, is used
Pass through the information obtained after calculating to record input each time.
It is as follows to calculate the step of hiding layer state：
Step 1, select which information abandoned from the location mode of a upper time：
ft=σ (Wt·[ht-1,xt]+bf)
Step 2, determine which information will be stored in the location mode of current time：
it=σ (Wt·[ht-1,xt]+bi)
Step 3, the location mode of current time is updated：
Step 4, it is determined that last output：
ot=σ (Wo[ht-1,xt]+bo)
ht=ot*tanh(Ct)
Preferably, the present embodiment can use two-way LSTM units.Following steps S401 to step S403 is step S301
A kind of preferred implementation.
Step S401, to LSTM units and backward LSTM units before obtaining.
Such as in the present embodiment, including from left to right and from right to left, image or the corresponding sequential sequence of word,
The information of both direction is all useful.
Step S402, it is combined and stacks to LSTM units and backward LSTM units to preceding, it is mono- to form two-way LSTM
Member.
Step S403, calculated by two-way LSTM and hide layer state.
By two-way LSTM units, more rich information can be extracted, so as to more accurately obtain multiple characteristic vectors
Between annexation.
Step S302, according to annexation, multiple characteristic vectors are merged into sort-type.
By the annexation between each characteristic vector, excessively individual characteristic vector is merged into sort-type.
Step S303, value sequence of classifying corresponding to sort-type is determined using sequential sorting algorithm.
The present embodiment can pass through sequential sorting algorithm (Connectionist Temporal Classification, letter
Referred to as CTC) treatment classification formula, and by constructing Softmax layers displaying classification value sequence：By removing the label and ' non-repeated
Character ' label, the classification value sequence finally obtained is set of number, for example, last result be " 2,33,57,58,59,60,
61”。
Step S104, the character according to corresponding to classification value sequence determines license plate image.
The character that car plate may relate to has tens, including Chinese character, letter and number, wherein, Chinese character includes the letter of each provincial capital
The Chinese character that various abbreviations, public security or embassy that title, army are related to etc. may use, letter include A to Z, and numeral includes 0 to 9.
The present embodiment has preset the corresponding relation of classification value and character, by preset corresponding relation, 2 corresponding characters " Guangdong ", 33 pairs
Answer " B ", 57 to 61 respectively correspond to " 1 " to " 5 ", thus identify that car plate be " Guangdong B12345 ".
By the present embodiment, can avoid carrying out cutting and separated identification to the character of car plate with the whole car plate of Direct Recognition,
Improve recognition speed.
The embodiment of the present invention additionally provides a kind of license plate recognition device, and Fig. 4 is Car license recognition provided in an embodiment of the present invention
The structured flowchart of device, as shown in figure 4, the device includes the first acquisition module 41, the second acquisition module 42, the 3rd acquisition module
43 and determining module 44.
First acquisition module 41 is used to obtain license plate image.
Second acquisition module 42 be used for by convolutional neural networks model obtain multiple features in the license plate image to
Amount.
3rd acquisition module 43 is used to the multiple characteristic vector sequentially inputting Recognition with Recurrent Neural Network model, to be divided
Class value sequence, the classification value sequence include multiple classification values, and each classification value corresponds to a character in car plate.
Determining module 44 is used for the character according to corresponding to the sort-type determines the license plate image.
Alternatively, second acquisition module 42 includes：Submodule is generated, for passing through the convolutional neural networks model
Convolutional layer, generate multiple characteristic patterns corresponding to the license plate image.Extracting sub-module, for being carried from multiple described characteristic patterns
N number of characteristic vector is taken, i-th of characteristic vector is the amalgamation result of the i-th column vector in multiple described characteristic patterns, its
In, N and i are positive integer, and N is more than or equal to i.
Alternatively, the 3rd acquisition module 43 includes：Calculating sub module, it is hidden in the Cyclic Operation Network for calculating
Layer state is hidden, to obtain the annexation between multiple characteristic vectors；Merge submodule, for according to the annexation, inciting somebody to action
Multiple characteristic vectors are merged into sort-type.The classification value sequence according to corresponding to sequential sorting algorithm determines the sort-type.
Alternatively, the calculating sub module includes：Computing unit, described in being used as by shot and long term memory LSTM units
The unit of Recognition with Recurrent Neural Network model, to calculate the hiding layer state.
Alternatively, the computing unit includes：Subelement is obtained, for mono- to LSTM units and backward LSTM before obtaining
Member；Merge subelement, it is double to be formed for the forward direction LSTM units and the backward LSTM units to be combined and stacked
To LSTM units；Computation subunit, for calculating the hiding layer state by the two-way LSTM.
It is apparent to those skilled in the art that for convenience of description and succinctly, only with above-mentioned each work(
Can unit division progress for example, in practical application, can be as needed and by above-mentioned function distribution by different functions
Unit is completed, i.e., the internal structure of device is divided into different functional units or module, with complete it is described above whole or
Person's partial function.Each functional unit in embodiment can be integrated in a processing unit or unit is independent
Be physically present, can also two or more units it is integrated in a unit, above-mentioned integrated unit can both use hard
The form of part is realized, can also be realized in the form of SFU software functional unit.In addition, the specific name of each functional unit is also
For the ease of mutually distinguishing, the protection domain of the application is not limited to.The specific work process of unit in said apparatus, can
With reference to the corresponding process in aforementioned means embodiment, will not be repeated here.
Those of ordinary skill in the art are it is to be appreciated that the list of each example described with reference to the embodiments described herein
Member and algorithm steps, it can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually
Performed with hardware or software mode, application-specific and design constraint depending on technical scheme.Professional and technical personnel
Described function can be realized using different device to each specific application, but this realization is it is not considered that exceed
The scope of the present invention.
In embodiment provided by the present invention, it should be understood that disclosed device and device, others can be passed through
Mode is realized.For example, device embodiment described above is only schematical, for example, the division of module or unit, only
For a kind of division of logic function, there can be other dividing mode when actually realizing, such as multiple units or component can combine
Or another system is desirably integrated into, or some features can be ignored, or do not perform.Another, shown or discussed phase
Coupling or direct-coupling or communication connection between mutually can be by some interfaces, the INDIRECT COUPLING or communication of device or unit
Connection, can be electrical, mechanical or other forms.
The unit illustrated as separating component can be or may not be physically separate, be shown as unit
Part can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple networks
On unit.Some or all of unit therein can be selected to realize the purpose of this embodiment scheme according to the actual needs.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also
That unit is individually physically present, can also two or more units it is integrated in a unit.Above-mentioned integrated list
Member can both be realized in the form of hardware, can also be realized in the form of SFU software functional unit.
If integrated unit is realized in the form of SFU software functional unit and is used as independent production marketing or in use, can
To be stored in a computer read/write memory medium.Based on such understanding, the technical scheme essence of the embodiment of the present invention
On all or part of the part that is contributed in other words to prior art or the technical scheme can be with the shape of software product
Formula is embodied, and the computer software product is stored in a storage medium, including some instructions to cause one calculating
It is real that machine equipment (can be personal computer, server, or network equipment etc.) or processor (processor) perform the present invention
Apply all or part of step of each embodiment device of example.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage
Device (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or light
Disk etc. is various can be with the medium of store program codes.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although with reference to the foregoing embodiments
The present invention is described in detail, it will be understood by those within the art that：It still can be to foregoing each implementation
Technical scheme described in example is modified, or carries out equivalent substitution to which part technical characteristic；And these modification or
Replace, the essence of appropriate technical solution is departed from the spirit and scope of each embodiment technical scheme of the embodiment of the present invention.
These are only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
All any modification, equivalent and improvement made within principle etc., should be included in the scope of the protection.
Claims (10)
- A kind of 1. licence plate recognition method, it is characterised in that including：Obtain license plate image；Multiple characteristic vectors in the license plate image are obtained by convolutional neural networks model；The multiple characteristic vector is sequentially input into Recognition with Recurrent Neural Network model, to obtain classification value sequence, the classification value sequence Row include multiple classification values, and each classification value corresponds to a character in car plate；The character according to corresponding to the classification value sequence determines the license plate image.
- 2. the method as described in claim 1, it is characterised in that obtained by convolutional neural networks model in the license plate image Multiple characteristic vectors include：By the convolutional layer of the convolutional neural networks model, multiple characteristic patterns corresponding to the license plate image are generated；N number of characteristic vector is extracted from multiple described characteristic patterns, i-th of characteristic vector is in multiple described characteristic patterns The amalgamation result of i-th column vector, wherein, N and i are positive integer, and N is more than or equal to i.
- 3. the method as described in claim 1, it is characterised in that the multiple characteristic vector is sequentially input into Recognition with Recurrent Neural Network Model, included with obtaining classification value sequence：Calculate and layer state is hidden in the Cyclic Operation Network, to obtain the annexation between multiple characteristic vectors；According to the annexation, multiple characteristic vectors are merged into sort-type；The classification value sequence according to corresponding to sequential sorting algorithm determines the sort-type.
- 4. method as claimed in claim 3, it is characterised in that calculate the state bag of hidden layer in the Cyclic Operation Network Include：It is used as the unit of the Recognition with Recurrent Neural Network model by shot and long term memory LSTM units, to calculate the hiding stratiform State.
- 5. method as claimed in claim 4, it is characterised in that LSTM units are remembered by shot and long term and are used as the circulation nerve The unit of network model, included with calculating the hiding layer state：To LSTM units and backward LSTM units before obtaining；The forward direction LSTM units and the backward LSTM units are combined and stacked, to form two-way LSTM units；The hiding layer state is calculated by the two-way LSTM.
- A kind of 6. license plate recognition device, it is characterised in that including：First acquisition module, for obtaining license plate image；Second acquisition module, for obtaining multiple characteristic vectors in the license plate image by convolutional neural networks model；3rd acquisition module, for the multiple characteristic vector to be sequentially input into Recognition with Recurrent Neural Network model, to obtain classification value Sequence, the classification value sequence include multiple classification values, and each classification value corresponds to a character in car plate；Determining module, for character corresponding to determining the license plate image according to the sort-type.
- 7. device as claimed in claim 6, it is characterised in that second acquisition module includes：Submodule is generated, for the convolutional layer by the convolutional neural networks model, is generated more corresponding to the license plate image Open characteristic pattern；Extracting sub-module, for extracting N number of characteristic vector from multiple described characteristic patterns, i-th of characteristic vector is The amalgamation result of i-th column vector in multiple described characteristic patterns, wherein, N and i are positive integer, and N is more than or equal to i.
- 8. device as claimed in claim 6, it is characterised in that the 3rd acquisition module includes：Calculating sub module, layer state is hidden in the Cyclic Operation Network for calculating, to obtain between multiple characteristic vectors Annexation；Merge submodule, for according to the annexation, multiple characteristic vectors to be merged into sort-typeThe classification value sequence according to corresponding to sequential sorting algorithm determines the sort-type.
- 9. device as claimed in claim 8, it is characterised in that the calculating sub module includes：Computing unit, for being used as the unit of the Recognition with Recurrent Neural Network model by shot and long term memory LSTM units, to calculate The hiding layer state.
- 10. device as claimed in claim 9, it is characterised in that the computing unit includes：Subelement is obtained, it is preceding to LSTM units and backward LSTM units for obtaining；Merge subelement, for the forward direction LSTM units and the backward LSTM units to be combined and stacked, to be formed Two-way LSTM units；Computation subunit, for calculating the hiding layer state by the two-way LSTM.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611158854.8A CN107784303A (en) | 2016-12-15 | 2016-12-15 | Licence plate recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611158854.8A CN107784303A (en) | 2016-12-15 | 2016-12-15 | Licence plate recognition method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107784303A true CN107784303A (en) | 2018-03-09 |
Family
ID=61437551
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611158854.8A Pending CN107784303A (en) | 2016-12-15 | 2016-12-15 | Licence plate recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107784303A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830213A (en) * | 2018-06-12 | 2018-11-16 | 北京理工大学 | Car plate detection and recognition methods and device based on deep learning |
| CN109893492A (en) * | 2019-04-09 | 2019-06-18 | 广州昀东生物科技有限公司 | One Plant Extracts moisturizing effect cosmetic |
| CN109961057A (en) * | 2019-04-03 | 2019-07-02 | 罗克佳华科技集团股份有限公司 | A kind of vehicle location preparation method and device |
| CN109977824A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | Article picks and places recognition methods, device and equipment |
| CN110070085A (en) * | 2019-04-30 | 2019-07-30 | 北京百度网讯科技有限公司 | Licence plate recognition method and device |
| WO2019174405A1 (en) * | 2018-03-14 | 2019-09-19 | 台达电子工业股份有限公司 | License plate identification method and system thereof |
| CN110689002A (en) * | 2018-07-05 | 2020-01-14 | 山东华软金盾软件股份有限公司 | License plate recognition method and device based on recurrent neural network model |
| CN110688880A (en) * | 2018-07-06 | 2020-01-14 | 山东华软金盾软件股份有限公司 | License plate identification method based on simplified ResNet residual error network |
| CN110728167A (en) * | 2018-07-16 | 2020-01-24 | 杭州海康威视数字技术股份有限公司 | Text detection method and device and computer readable storage medium |
| CN111832568A (en) * | 2020-06-12 | 2020-10-27 | 北京百度网讯科技有限公司 | License plate recognition method, and training method and device of license plate recognition model |
| CN113449760A (en) * | 2020-03-27 | 2021-09-28 | 北京沃东天骏信息技术有限公司 | Character recognition method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572892A (en) * | 2014-12-24 | 2015-04-29 | 中国科学院自动化研究所 | Text classification method based on cyclic convolution network |
| CN105654127A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | End-to-end-based picture character sequence continuous recognition method |
| CN105678292A (en) * | 2015-12-30 | 2016-06-15 | 成都数联铭品科技有限公司 | Complex optical text sequence identification system based on convolution and recurrent neural network |
| CN105938485A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Image description method based on convolution cyclic hybrid model |
| CN106126492A (en) * | 2016-06-07 | 2016-11-16 | 北京高地信息技术有限公司 | Statement recognition methods based on two-way LSTM neutral net and device |
-
2016
- 2016-12-15 CN CN201611158854.8A patent/CN107784303A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104572892A (en) * | 2014-12-24 | 2015-04-29 | 中国科学院自动化研究所 | Text classification method based on cyclic convolution network |
| CN105654127A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | End-to-end-based picture character sequence continuous recognition method |
| CN105678292A (en) * | 2015-12-30 | 2016-06-15 | 成都数联铭品科技有限公司 | Complex optical text sequence identification system based on convolution and recurrent neural network |
| CN105938485A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Image description method based on convolution cyclic hybrid model |
| CN106126492A (en) * | 2016-06-07 | 2016-11-16 | 北京高地信息技术有限公司 | Statement recognition methods based on two-way LSTM neutral net and device |
Non-Patent Citations (4)
| Title |
|---|
| CHUNTING ZHOU 等: "A C-LSTM Neural Network for Text Classification", 《ARXIV:1511.08630V2》 * |
| GUO-BING ZHOU 等: "Minimal Gated Unit for Recurrent Neural Networks", 《ARXIV:1603.09420V1》 * |
| SOROUSH VOSOUGHI 等: "Tweet2Vec: Learning Tweet Embeddings Using Character-level CNN-LSTM Encoder-Decoder", 《ARXIV:1607.07514V1》 * |
| 陈雯柏编: "《人工神经网络原理与实践》", 31 January 2016 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11443535B2 (en) | 2018-03-14 | 2022-09-13 | Delta Electronics, Inc. | License plate identification method and system thereof |
| WO2019174405A1 (en) * | 2018-03-14 | 2019-09-19 | 台达电子工业股份有限公司 | License plate identification method and system thereof |
| CN108830213A (en) * | 2018-06-12 | 2018-11-16 | 北京理工大学 | Car plate detection and recognition methods and device based on deep learning |
| CN110689002A (en) * | 2018-07-05 | 2020-01-14 | 山东华软金盾软件股份有限公司 | License plate recognition method and device based on recurrent neural network model |
| CN110688880A (en) * | 2018-07-06 | 2020-01-14 | 山东华软金盾软件股份有限公司 | License plate identification method based on simplified ResNet residual error network |
| CN110728167A (en) * | 2018-07-16 | 2020-01-24 | 杭州海康威视数字技术股份有限公司 | Text detection method and device and computer readable storage medium |
| CN109977824A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | Article picks and places recognition methods, device and equipment |
| CN109961057A (en) * | 2019-04-03 | 2019-07-02 | 罗克佳华科技集团股份有限公司 | A kind of vehicle location preparation method and device |
| CN109961057B (en) * | 2019-04-03 | 2021-09-03 | 罗克佳华科技集团股份有限公司 | Vehicle position obtaining method and device |
| CN109893492A (en) * | 2019-04-09 | 2019-06-18 | 广州昀东生物科技有限公司 | One Plant Extracts moisturizing effect cosmetic |
| CN110070085A (en) * | 2019-04-30 | 2019-07-30 | 北京百度网讯科技有限公司 | Licence plate recognition method and device |
| CN113449760A (en) * | 2020-03-27 | 2021-09-28 | 北京沃东天骏信息技术有限公司 | Character recognition method and device |
| CN111832568A (en) * | 2020-06-12 | 2020-10-27 | 北京百度网讯科技有限公司 | License plate recognition method, and training method and device of license plate recognition model |
| CN111832568B (en) * | 2020-06-12 | 2024-01-12 | 北京百度网讯科技有限公司 | License plate recognition method, training method and device of license plate recognition model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107784303A (en) | Licence plate recognition method and device | |
| US9811718B2 (en) | Method and a system for face verification | |
| KR102554724B1 (en) | Method for identifying an object in an image and mobile device for practicing the method | |
| Qi et al. | Image-based action recognition using hint-enhanced deep neural networks | |
| CN108197532A (en) | The method, apparatus and computer installation of recognition of face | |
| Türkyılmaz et al. | License plate recognition system using artificial neural networks | |
| Lee et al. | Recognizing pedestrian’s unsafe behaviors in far-infrared imagery at night | |
| CN109117879A (en) | Image classification method, apparatus and system | |
| Ye et al. | Detecting USM image sharpening by using CNN | |
| CN110163111A (en) | Method, apparatus of calling out the numbers, electronic equipment and storage medium based on recognition of face | |
| Wang et al. | Multiscale deep alternative neural network for large-scale video classification | |
| CN109508625A (en) | Emotional data analysis method and device | |
| Saddique et al. | Classification of authentic and tampered video using motion residual and parasitic layers | |
| CN111832573A (en) | Image emotion classification method based on class activation mapping and visual saliency | |
| CN110378203A (en) | Image processing method, device, terminal and storage medium | |
| Yuan et al. | A lightweight network for smoke semantic segmentation | |
| CN114783021A (en) | Intelligent detection method, device, equipment and medium for wearing of mask | |
| Hassan et al. | An empirical analysis of deep learning architectures for vehicle make and model recognition | |
| Lu et al. | An efficient fine-grained vehicle recognition method based on part-level feature optimization | |
| Pan et al. | C3DBed: Facial micro-expression recognition with three-dimensional convolutional neural network embedding in transformer model | |
| CN112749576B (en) | Image recognition method and device, computing equipment and computer storage medium | |
| Rakowski et al. | Hand shape recognition using very deep convolutional neural networks | |
| Mi et al. | Dual-branch network with a subtle motion detector for microaction recognition in videos | |
| CN112419249B (en) | Special clothing picture conversion method, terminal device and storage medium | |
| Ko et al. | View-independent object detection using shared local features |
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 |
Application publication date: 20180309 |
|
| RJ01 | Rejection of invention patent application after publication |