CN108960175A - A kind of licence plate recognition method based on deep learning - Google Patents
A kind of licence plate recognition method based on deep learning Download PDFInfo
- Publication number
- CN108960175A CN108960175A CN201810766184.0A CN201810766184A CN108960175A CN 108960175 A CN108960175 A CN 108960175A CN 201810766184 A CN201810766184 A CN 201810766184A CN 108960175 A CN108960175 A CN 108960175A
- Authority
- CN
- China
- Prior art keywords
- license plate
- layers
- layer
- plate
- priorbox
- 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
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000013135 deep learning Methods 0.000 title claims abstract description 13
- 230000011218 segmentation Effects 0.000 claims abstract description 19
- 238000001514 detection method Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 5
- 238000007689 inspection Methods 0.000 claims description 4
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000003475 lamination Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 6
- 238000012804 iterative process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004438 eyesight Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- 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)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
License plate recognition technology is using very extensive.The present invention proposes a kind of licence plate recognition method based on deep learning, licence plate recognition method includes that detection license plate whether there is with two stages of Recognition of License Plate Characters, and the model used includes License Plate Segmentation model, car plate Chinese character identification model and license plate letter and digital identification model.License Plate Segmentation model includes 4 layers of convolutional layer, 3 Relu layers, 3 Pool layers, priorbox layers of the first to be connected with third layer convolutional layer, first position predicting unit and the first confidence level predicting unit, priorbox layers of the 2nd to be connected with the 4th layer of convolutional layer, second position predicting unit and the second confidence level predicting unit, with the first priorbox layers and the 2nd priorbox layers of mbox_priorbox layer being connected.Car plate Chinese character identification model includes four convolution layer units being sequentially connected, flatte unit, dropout unit and Softmax layers.License plate letter and digital identification model include two convolution layer units being sequentially connected, flatte unit, dropout unit and Softmax layers.
Description
Technical field
The invention belongs to Car license recognition field, especially a kind of licence plate recognition method based on deep learning.
Background technique
Vehicle License Plate Recognition System refers to the vehicle for being able to detect that monitored road surface and automatically extracts vehicle license information (containing the Chinese
Word character, English alphabet, Arabic numerals and number plate color) technology that is handled.Car license recognition is modern intelligent transportation system
One of important component in system, application are very extensive.It is with skills such as Digital Image Processing, pattern-recognition, computer visions
Based on art, the vehicle image or video sequence of shot by camera are analyzed, obtain each unique vehicle of automobile
Trade mark code, to complete identification process.Parking lot fee collection management, magnitude of traffic flow control may be implemented by some subsequent processing means
Index measurement processed, vehicle location, automobile burglar, high way super speed automate supervision, electronic eye used for catching red light runner, toll station
Etc. function.For maintenance traffic safety and urban public security, traffic jam is prevented, realizes that traffic automation management has reality
Meaning.
Summary of the invention
Based on this, the present invention proposes a kind of licence plate recognition method based on deep learning, the technical solution adopted is as follows:
A kind of licence plate recognition method based on deep learning, which is characterized in that the licence plate recognition method includes detection vehicle
Board whether there is and two stages of Recognition of License Plate Characters, and the model used includes License Plate Segmentation model, car plate Chinese character identification model
With license plate letter and digital identification model.
Further, License Plate Segmentation model includes 4 layers of convolutional layer, 3 Relu layers, 3 Pool layers, with third layer convolutional layer
Connected first priorbox layers, first position predicting unit and the first confidence level predicting unit, are connected with the 4th layer of convolutional layer
The 2nd priorbox layers, second position predicting unit and the second confidence level predicting unit, with the first priorbox layers and second
Priorbox layers of connected mbox_priorbox layer.In 4 layers of convolutional layer, one layer of Relu is equipped between every two layers of convolutional layer
Layer and one layer Pool layers.
Further, first position predicting unit, second position predicting unit, the first confidence level predicting unit and second are set
The structure of reliability predicting unit is identical, including level 1 volume lamination, and 1 layer Permute layers, 1 layer Flatten layers.
Further, first position predicting unit, second position predicting unit, the first confidence level predicting unit and second are set
The feature that reliability predicting unit is extracted is different.
Further, the car plate Chinese character identification model include four convolution layer units being sequentially connected, flatte unit,
Dropout unit and Softmax layers.
Further, the license plate letter and digital identification model include: two convolution layer units being sequentially connected,
Flatte unit, dropout unit and Softmax layers.
Further, the convolution layer unit includes one layer of convolutional layer, one layer relu layers and one layer Pool layers.
Further, flatte unit includes one layer Flatte layers and one layer InnerProduct layers.
Further, dropout unit includes one layer Dropout layers, one layer Relu layers and one layer InnerProduct layers.
Further, when training License Plate Segmentation model, car plate Chinese character identification model and license plate letter are with digital identification model,
The sample used is generated by license plate generator, by adjusting tilt angle in license plate generator and corrosion strength the two parameters,
Data enhancement operations are carried out, and then generate the required sample of training.
Further, complete Car license recognition process includes:
Step 1. obtains the picture of camera acquisition；
Step 2. detection license plate whether there is；
Step 3. is split license plate when detecting license plate；
Step 4. identifies Chinese character, number and letter in license plate respectively；
Step 5. exports 7 license plate numbers and confidence score, completes Car license recognition.
Further, car plate detection and the step of segmentation, include:
Step 1. utilizes SSD model inspection license plate, calculates license plate confidence level；
If step 2. license plate confidence level is more than or equal to 0.5, the License Plate Segmentation model treatment image is utilized, calculates license plate
In all characters image coordinate.
Compared with prior art, the beneficial effects of the present invention are: License Plate Segmentation model complexity is low, passes through 4 convolution
Layer unit carries out the prediction of position and confidence level in conjunction with two position prediction units and two confidence level predicting units, entire to divide
The precision for cutting process is high, and speed is fast.Feature complexity based on Chinese character, letter and number designs car plate Chinese character identification model
Guarantee Chinese character and letter and number separately identification that can improve identification while precision with license plate letter and digital identification model
Speed.
Detailed description of the invention
Fig. 1 is complete Car license recognition flow chart of the invention；
Fig. 2 is the flow chart that present invention detection license plate whether there is；
Fig. 3 is License Plate Segmentation model structure proposed by the present invention；
Fig. 4 is Chinese Character Recognition model structure proposed by the present invention；
Fig. 5 is letter and number identification model mechanism map proposed by the present invention.
Description of symbols:
First priorbox layers -1, the the 2nd priorbox layers -2, mbox_priorbox layers -3, first position predicting unit -
4, second position predicting unit -5, the first confidence level predicting unit -6, the second confidence level predicting unit -7.
Specific embodiment
Complete Car license recognition process includes: as shown in Figure 1
Step 1. obtains the picture of camera acquisition；
Step 2. detection license plate whether there is；
Step 3. is split license plate when detecting license plate；
Step 4. identifies Chinese character, number and letter in license plate respectively；
Step 5. exports 7 license plate numbers and confidence score, completes Car license recognition.
As shown in Fig. 2, the step of car plate detection and segmentation, includes:
Step 1. utilizes SSD model inspection license plate, calculates license plate confidence level；
If step 2. license plate confidence level is more than or equal to 0.5, License Plate Segmentation model treatment image proposed by the present invention is utilized,
Calculate the image coordinate of all characters in license plate.
In the present embodiment, caffe frame has been used using the foundation of model and training during Car license recognition.Wherein examine
The model that measuring car board uses when whether there is is basic SSD model, establishes and includes: the step of training SSD model
Step 1. generates sample: collecting 40,000 vehicles, (various angles, light, position, vehicle etc. embody the multiplicity of sample
Property) picture, generate training set and test set at random in proportion (ratio of training set and test set is 3:1 in the present embodiment).People
To outline license plate rectangle frame using marking tool .xml file is automatically generated, the available .lmbd format sample of training pattern is regenerated
This document.
Step 2. training pattern: write train_test.prototxt training network model file and
Deploy.prototxt test network model file exports the confidence score and license plate of license plate with sample training SSD model
Position (position on 4 vertex).
Step 3. test model: trained SSD model inspection effect is tested with vehicle pictures, is adjusted according to training effect
Solver.prototxt hyper parameter file, the parameter of adjustment include learning rate, maximum number of iterations and gradient weight, are retained most
The .caffemodel file generated afterwards.
Model for license plate number identification includes License Plate Segmentation model, car plate Chinese character identification model, license plate letter and number
Word identification model.
As shown in figure 3, License Plate Segmentation model includes 4 layers of convolutional layer, 3 Relu layers, 3 Pool layers, the first priorbox
Layer 1, the 2nd priorbox layer 2, mbox_priorbox layer 3, first position predicting unit 4, second position predicting unit 5, first
Confidence level predicting unit 6, the second confidence level predicting unit 7.First position predicting unit 4, second position predicting unit 5, first
Confidence level predicting unit 6 is identical with the structure of the second confidence level predicting unit 7, including level 1 volume lamination, and 1 layer Permute layers, 1 layer
Flatten layers, first position predicting unit 4, second position predicting unit 5, the first confidence level predicting unit 6 and the second confidence level
The feature that predicting unit 7 is extracted is different.Relu layers non-linear for increasing network, Pool layers big for reducing next layer of input
It is small, reduce calculation amount and number of parameters.Permute layers are played the role of exchanging dimension order, and Flatten is played the defeated of multidimensional
Enter to be converted to one-dimensional effect.
It establishes and includes: the step of training License Plate Segmentation model
Step 1. establishes model: according to License Plate Segmentation model write train_test.prototxt training network file and
Deploy.prototxt file.Specifically, the parameter of 4 convolutional layers be num_output:16, kernel_size:3,
Stride:1, pad:0；Pool layer parameter is pool:MAX, kernel_size:2, stride:2, pad:0.Position prediction unit
Parameter with convolutional layer in confidence level predicting unit is num_output:8, kernel_size:3, stride:1, pad:1；
Permute layer parameter is order:0,2,3,1；Flatten layer parameter is axis:1.
Step 2. test model: sample training model, the result and confidence score of output 7 characters on license plate of positioning are used.
Wherein the process of iteration includes: each time
Two position prediction units receive the output of the 3rd layer of convolutional layer and the 4th layer of convolutional layer respectively, by two position predictions
Flatten layers of output, which is sent into mbox_loc, in unit carries out channel merging, parameter axis:1；By the output of the 3rd layer of convolution,
Initial data is sent to the first PriorBox layers, the output of the 4th layer of convolution is sent into the 2nd PriorBox layers, parameter aspect_
Ratio:2,3, variance:0,1,0,1,0,2,0,2.After the first, second PriorBox layers, data are sent into mbox_
Priorbox layers, realize that channel merges, parameter axis:1.
Two confidence level predicting units receive the output of the 3rd layer of convolutional layer and the 4th layer of convolutional layer respectively, by two confidence levels
Flatten layers of output, which is sent into mbox_conf, in predicting unit carries out channel merging, parameter axis:1.
The output of mbox_loc, mbox_conf, mbox_priorbox are all sent into multiboxloss layers, parameter num_
Output:7 exports positioning result and confidence score.
Step 3. test model: trained model character locating effect is tested with license plate picture, according to test case tune
Whole solver.prototxt hyper parameter file retains the .caffemodel file ultimately produced, carries out characters on license plate positioning.
The generation method of sample in the present embodiment are as follows: generate 4000 license plates, title such as capital A first with license plate generator
11111.jpg etc., then by using filtering, scaling, the methods of perspective transform, cooperation constantly adjustment tilt angle and corrosion strength
The two parameters carry out data enhancement operations, generate the sample of 80000 titles such as capital A 11111_1.jpg.Since license plate is known
There are character inclination, " 0 " and " D " identifications to be easy the presence of the problems such as obscuring during not, can be with during generating sample
The appropriate sample size for increasing " 0 ", " D ".
It establishes and includes: the step of training car plate Chinese character identification model
Step 1. generates sample: online disclosed individual Chinese character character sample is utilized in the present embodiment.
Step 2. training pattern: train_test.prototxt training network text is write according to car plate Chinese character identification model
Part and deploy.prototxt file, with sample training model, the result and confidence level of output identification individual Chinese character character are obtained
Point.
As shown in figure 4, car plate Chinese character identification model include four convolution layer units, flatte unit, dropout unit and
Softmax layers.Each convolution layer unit includes one layer of convolutional layer, one layer Relu layers and one layer Pool layers.Flatte unit includes
One layer Flatte layers and one layer InnerProduct layers, dropout unit includes one layer Dropout layers, one layer relu layers and one
InnerProduct layers of layer.
Iterative process includes: each time
Data successively pass through convolution layer unit, carry out feature extraction, and convolution layer parameter is num_output:16, kernel_
Size:3, stride:1, pad:0；Pool layer parameter is MAX, kernel_size:2, stride:2, pad:0.Then pass through
Flatten layers, parameter axis；1；By InnerProduct layers, it is therefore an objective to input data is handled in the form of vectors, it will
The feature learnt is mapped to sample classification space, parameter num_output:256 again；By Dropout layers, parameter
Dropout_ratio:0.5, the layer are can to lose certain connections at random, prevent network over-fitting；By Relu layers；By
InnerProduct layers, the classification num_output:31 of parametric classification.
Softmax layers are finally sent into, the probability likelihood value of each classification is calculated, recognition result is exported and confidence level obtains
Point.
Step 3. test model: testing trained model character recognition effect with individual Chinese character picture, according to test feelings
Condition adjusts solver.prototxt hyper parameter file, retains the .caffemodel file ultimately produced, carries out car plate Chinese character word
Symbol identification.
Establish and training license plate letter with number identification model the step of include:
Step 1. generates sample: online disclosed single letter and digital character sample are utilized in the present embodiment.
Step 2. training pattern: train_test.prototxt training is write according to license plate letter and digital identification model
Network file and deploy.prototxt file, with sample training model, the result of output identification single letter or numerical character
And confidence score.
As shown in figure 5, license plate letter and digital identification model include: two convolution layer units, flatte unit,
Dropout unit and Softmax layers.Each convolution layer unit includes one layer of convolutional layer, one layer Relu layers and one layer Pool layers.
Flatte unit includes one layer Flatte layer and one layer InnerProduct layers, dropout unit including one layer Dropout layers,
One layer relu layers and one layer InnerProduct layers.
Iterative process includes: each time
Data successively pass through two convolution layer units, and the parameter of convolutional layer is num_output:16, kernel_size:3,
Stride:1, pad:0；Pool layer parameter is MAX, kernel_size:2, stride:2, pad:0；Then pass through Flatten
Layer；By InnerProduct layers, parameter num_output is set as 256；By Dropout layers, parameter dropout_ratio is set
It is 0.5；By Relu layers；By InnerProduct layers, the classification num_output of classification is set as 34, adds 10 for 24 letters
A number,
Finally it is sent into the Softmax layers of probability likelihood value for calculating each classification, the maximum conduct of output probability likelihood value
Recognition result and confidence score.
Step 3. test model: testing trained model character recognition effect with single letter or digital picture, according to
Test case adjusts solver.prototxt hyper parameter file, retains ultimogenitary .caffemodel file, carries out license plate word
Female or Number character recognition.
The foregoing is merely the preferred embodiments of the invention, are not intended to limit the invention creation, all at this
Within the spirit and principle of innovation and creation, any modification, equivalent replacement, improvement and so on should be included in the invention
Protection scope within.
Claims (7)
1. a kind of licence plate recognition method based on deep learning, which is characterized in that the licence plate recognition method includes detection license plate
With the presence or absence of with two stages of Recognition of License Plate Characters, the model used include License Plate Segmentation model, car plate Chinese character identification model and
License plate letter and digital identification model.
2. a kind of licence plate recognition method based on deep learning as described in claim 1, which is characterized in that License Plate Segmentation model packet
Include sequentially connected 4 layers of convolutional layer, priorbox layers of the first to be connected with third layer convolutional layer, first position predicting unit and
One confidence level predicting unit, priorbox layers of the 2nd to be connected with the 4th layer of convolutional layer, second position predicting unit and second are set
Reliability predicting unit, with the first priorbox layers and the 2nd priorbox layers of mbox_priorbox layer being connected, 4 layers of volume
In lamination, one layer Relu layers and one layer Pool layers are equipped between every two layers of convolutional layer.
3. a kind of licence plate recognition method based on deep learning as described in claim 1, which is characterized in that the car plate Chinese character is known
Other model includes four convolution layer units being sequentially connected, flatte unit, dropout unit and Softmax layers.
4. a kind of licence plate recognition method based on deep learning as described in claim 1, which is characterized in that the license plate letter with
Digital identification model includes two convolution layer units being sequentially connected, flatte unit, dropout unit and Softmax layers.
5. a kind of licence plate recognition method based on deep learning as described in claim 1, which is characterized in that training License Plate Segmentation mould
When type, car plate Chinese character identification model and license plate letter and digital identification model, the sample used is generated by license plate generator, is passed through
Tilt angle and corrosion strength the two parameters in license plate generator are adjusted, data enhancement operations are carried out, and then generate training institute
The sample needed.
6. a kind of licence plate recognition method based on deep learning as described in claim 1, which is characterized in that complete Car license recognition
Process includes:
Step 1. obtains the picture of camera acquisition；
Step 2. detection license plate whether there is；
Step 3. is split license plate when detecting license plate；
Step 4. identifies Chinese character, number and letter in license plate respectively；
Step 5. exports 7 license plate numbers and confidence score, completes Car license recognition.
7. a kind of licence plate recognition method based on deep learning as claimed in claim 6, which is characterized in that whether detection license plate is deposited
Include: in the step of with being split to license plate
Step 1. utilizes SSD model inspection license plate, calculates license plate confidence level；
If step 2. license plate confidence level is more than or equal to 0.5, the License Plate Segmentation model treatment image is utilized, calculates institute in license plate
There is the image coordinate of character.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810766184.0A CN108960175A (en) | 2018-07-12 | 2018-07-12 | A kind of licence plate recognition method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810766184.0A CN108960175A (en) | 2018-07-12 | 2018-07-12 | A kind of licence plate recognition method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108960175A true CN108960175A (en) | 2018-12-07 |
Family
ID=64483226
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810766184.0A Pending CN108960175A (en) | 2018-07-12 | 2018-07-12 | A kind of licence plate recognition method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108960175A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070072A (en) * | 2019-05-05 | 2019-07-30 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110097044A (en) * | 2019-05-13 | 2019-08-06 | 苏州大学 | Stage car plate detection recognition methods based on deep learning |
| CN110674802A (en) * | 2019-09-09 | 2020-01-10 | 电子科技大学 | Improved text detection method for parallelogram candidate box |
| CN111582263A (en) * | 2020-05-12 | 2020-08-25 | 上海眼控科技股份有限公司 | License plate recognition method and device, electronic equipment and storage medium |
| CN113685770A (en) * | 2021-09-06 | 2021-11-23 | 盐城香农智能科技有限公司 | Street lamp for environment monitoring and monitoring method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975968A (en) * | 2016-05-06 | 2016-09-28 | 西安理工大学 | Caffe architecture based deep learning license plate character recognition method |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
| CN107220638A (en) * | 2017-07-03 | 2017-09-29 | 深圳市唯特视科技有限公司 | A kind of car plate detection recognition methods based on deep learning convolutional neural networks |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN107491752A (en) * | 2017-08-14 | 2017-12-19 | 中国石油大学(华东) | Ship board character recognition method, device in a kind of natural scene based on deep learning |
-
2018
- 2018-07-12 CN CN201810766184.0A patent/CN108960175A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975968A (en) * | 2016-05-06 | 2016-09-28 | 西安理工大学 | Caffe architecture based deep learning license plate character recognition method |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
| CN107220638A (en) * | 2017-07-03 | 2017-09-29 | 深圳市唯特视科技有限公司 | A kind of car plate detection recognition methods based on deep learning convolutional neural networks |
| CN107423760A (en) * | 2017-07-21 | 2017-12-01 | 西安电子科技大学 | Based on pre-segmentation and the deep learning object detection method returned |
| CN107491752A (en) * | 2017-08-14 | 2017-12-19 | 中国石油大学(华东) | Ship board character recognition method, device in a kind of natural scene based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 赵振兴: "基于深度学习的车牌识别技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 郭叶军 等: "SSD算法推理过程的探析", 《现代计算机》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070072A (en) * | 2019-05-05 | 2019-07-30 | 厦门美图之家科技有限公司 | A method of generating object detection model |
| CN110097044A (en) * | 2019-05-13 | 2019-08-06 | 苏州大学 | Stage car plate detection recognition methods based on deep learning |
| CN110674802A (en) * | 2019-09-09 | 2020-01-10 | 电子科技大学 | Improved text detection method for parallelogram candidate box |
| CN111582263A (en) * | 2020-05-12 | 2020-08-25 | 上海眼控科技股份有限公司 | License plate recognition method and device, electronic equipment and storage medium |
| CN113685770A (en) * | 2021-09-06 | 2021-11-23 | 盐城香农智能科技有限公司 | Street lamp for environment monitoring and monitoring method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960175A (en) | A kind of licence plate recognition method based on deep learning | |
| CN111274970B (en) | Traffic sign detection method based on improved YOLO v3 algorithm | |
| CN113688652B (en) | Abnormal driving behavior processing method and device | |
| CN109948416A (en) | A kind of illegal occupancy bus zone automatic auditing method based on deep learning | |
| CN109508715A (en) | A kind of License Plate and recognition methods based on deep learning | |
| CN110175613A (en) | Street view image semantic segmentation method based on Analysis On Multi-scale Features and codec models | |
| CN106156766A (en) | The generation method and device of line of text grader | |
| CN107506763A (en) | A kind of multiple dimensioned car plate precise positioning method based on convolutional neural networks | |
| CN105354568A (en) | Convolutional neural network based vehicle logo identification method | |
| CN105868700A (en) | Vehicle type recognition and tracking method and system based on monitoring video | |
| CN107085696A (en) | A kind of vehicle location and type identifier method based on bayonet socket image | |
| CN103279738B (en) | Automatic identification method and system for vehicle logo | |
| CN109886147A (en) | A kind of more attribute detection methods of vehicle based on the study of single network multiple-task | |
| CN110009648A (en) | Trackside image Method of Vehicle Segmentation based on depth Fusion Features convolutional neural networks | |
| CN104978567A (en) | Vehicle detection method based on scenario classification | |
| Azad et al. | New method for optimization of license plate recognition system with use of edge detection and connected component | |
| CN105574489A (en) | Layered stack based violent group behavior detection method | |
| CN110287879A (en) | A kind of video behavior recognition methods based on attention mechanism | |
| CN110009058A (en) | A kind of parking lot Vehicle License Plate Recognition System and method | |
| CN111178282A (en) | Road traffic speed limit sign positioning and identifying method and device | |
| Zhang et al. | DetReco: object-text detection and recognition based on deep neural network | |
| CN103679209B (en) | Character identifying method based on sparse theory | |
| CN113673527A (en) | License plate recognition method and system | |
| CN105335758A (en) | Model identification method based on video Fisher vector descriptors | |
| CN111832463A (en) | Deep learning-based traffic sign detection method |
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: 20181207 |