CN109766819A - Testing vehicle register identification method and device - Google Patents
Testing vehicle register identification method and device Download PDFInfo
- Publication number
- CN109766819A CN109766819A CN201910008768.6A CN201910008768A CN109766819A CN 109766819 A CN109766819 A CN 109766819A CN 201910008768 A CN201910008768 A CN 201910008768A CN 109766819 A CN109766819 A CN 109766819A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- model
- vehicle location
- image
- training
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of testing vehicle register identification method and identification devices, which comprises obtains the original image of vehicle to be identified；According to vehicle location identification model, vehicle location is obtained from the original image；According to the vehicle location and brand recognition model, the identity information of the vehicle is obtained；The vehicle identity information includes vehicle brand, and vehicle location identification model is based on deep neural network.The vehicle location identification model of the embodiment of the present invention is based on deep neural network, the complexity of model based on deep neural network algorithm is lower, it can use GPU and carry out operation acceleration, accelerate the training speed of model, classification speed and the accuracy of model are improved simultaneously, to improve the accuracy of testing vehicle register identification.
Description
Technical field
The present invention relates to field of traffic.It is more particularly related to a kind of testing vehicle register identification method and device.
Background technique
Testing vehicle register identification is of great significance for traffic safety problem, and vehicle brand is a weight of vehicle identification
Component part is wanted, vehicle investigation, fake-licensed car are detected automatically with extremely important effect.
Vehicle brand identification may include vehicle location and brand recognition in image, and vehicle location is identified from background
Vehicle out.Existing vehicle brand is up to several hundred kinds, and difference is smaller between the brand of part, and existing vehicle brand identification method
It is easy to produce identification mistake.
Summary of the invention
The object of the present invention is to provide a kind of testing vehicle register identification methods, improve the accuracy of testing vehicle register identification.
In order to realize these purposes and other advantages according to the present invention, following technical scheme is provided:
In a first aspect, providing a kind of testing vehicle register identification method, which comprises
Obtain the original image of vehicle to be identified；
According to vehicle location identification model, vehicle location is obtained from the original image；
According to the vehicle location and brand recognition model, the identity information of the vehicle is obtained；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on depth nerve net
Network.
Optionally, described according to vehicle location identification model, before obtaining vehicle location in original image, the method
Further include:
Construct yolov2 model or yolov3 model based on deep neural network；
The yolov2 model or yolov3 model are trained, the vehicle location identification model is obtained.
It is optionally, described that yolov2 model or yolov3 model are trained, comprising:
Obtain several the first images；
From several the first images, the second image comprising automobile image is filtered out；
Several described second images are transformed to several third images of unified pre-set dimension；
According to third image described in several, expand training set, the multiple image in the training set is training image；
By convolution method, feature extraction is carried out to the training image, obtains the automobile position in the training image；
Training parameter, including initial learning rate, initial learning rate adjusting parameter, momentum coefficient, regular factor are set；
According to training parameter, yolov2 model is trained using the automobile position in training image described in several, is obtained
Obtain the vehicle location identification model.
Optionally, described according to vehicle location and brand recognition model, obtain the identity information of the vehicle, comprising:
According to the vehicle location, the general image of vehicle to be identified is obtained；
According to color, to the vehicle classification to be identified；
To the general image of sorted vehicle to be identified, logo is extracted using AlexNet network；
According to logo, the vehicle brand is identified.
Optionally, described according to third image, expand training set, comprising:
Method using random cropping, random overturning, colour switching expands training set.
Optionally, the method using random cropping expands training set, comprising:
The scale factor of random cropping is set.
It is optionally, described that training set is expanded using the method overturn at random, comprising:
The probability of flip horizontal is set.
Optionally, the method using colour switching expands training set, comprising:
The random variability of each Color Channel is set.
Second aspect, provides a kind of vehicle identification device, and described device includes:
Original image acquiring unit, for obtaining the original image of vehicle to be identified；
Vehicle location acquiring unit, for obtaining vehicle position from the original image according to vehicle location identification model
It sets；
Identity information acquiring unit, for obtaining the body of the vehicle according to the vehicle location and brand recognition model
Part information；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on depth nerve net
Network.
Optionally, the vehicle location acquiring unit includes:
Model construction subelement, for constructing yolov2 model or yolov3 model based on deep neural network；
Training subelement obtains the vehicle location for being trained to the yolov2 model or yolov3 model
Identification model.
The present invention is include at least the following beneficial effects:
The embodiment of the invention discloses a kind of testing vehicle register identification method and identification devices, which comprises obtain to
Identify the original image of vehicle；According to vehicle location identification model, vehicle location is obtained from the original image；According to described
Vehicle location and brand recognition model, obtain the identity information of the vehicle；The vehicle identity information includes vehicle brand, vehicle
Position identification model is based on deep neural network.The vehicle location identification model of the embodiment of the present invention is based on depth nerve net
The complexity of network, the model based on deep neural network algorithm is lower, can use GPU and carries out operation acceleration, accelerates model
Training speed, while classification speed and the accuracy of model are improved, to improve the accuracy of testing vehicle register identification.
Further advantage, target and feature of the invention will be partially reflected by the following instructions, and part will also be by this
The research and practice of invention and be understood by the person skilled in the art.
Detailed description of the invention
Fig. 1 is the flow chart of testing vehicle register identification method according to an embodiment of the invention；
Fig. 2 is the automobile image schematic diagram in testing vehicle register identification method according to an embodiment of the invention；
Fig. 3 show the structural schematic diagram of the vehicle identification device of the embodiment of the present invention.
Specific embodiment
Present invention will be described in further detail below with reference to the accompanying drawings, to enable those skilled in the art referring to specification text
Word can be implemented accordingly.
It should be noted that experimental method described in following embodiments is unless otherwise specified conventional method, institute
Reagent and material are stated, unless otherwise specified, is commercially obtained；In the description of the present invention, term " transverse direction ", " vertical
To ", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", the instructions such as "outside" side
Position or positional relationship are to be based on the orientation or positional relationship shown in the drawings, and are merely for convenience of description of the present invention and simplification of the description,
It is not that the device of indication or suggestion meaning or element must have a particular orientation, be constructed and operated in a specific orientation, because
This is not considered as limiting the invention.
Fig. 1 show the flow chart of the testing vehicle register identification method of the embodiment of the present invention, as shown in Figure 1, the vehicle body
Part recognition methods includes:
Step 110, the original image of vehicle to be identified is obtained；
Step 120, according to vehicle location identification model, vehicle location is obtained from the original image；
Step 130, according to the vehicle location and brand recognition model, the identity information of the vehicle is obtained；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on depth nerve net
Network.
It is before step 120, i.e., described according to vehicle location identification model in the embodiment of the present invention, it is obtained from original image
Before taking vehicle location, the method also includes:
Construct yolov2 model or yolov3 model based on deep neural network；
The yolov2 model or yolov3 model are trained, the vehicle location identification model is obtained.
It is described that yolov2 model or yolov3 model are trained in the embodiment of the present invention, comprising:
Obtain several the first images；
From several the first images, the second image comprising automobile image is filtered out；
Several described second images are transformed to several third images of unified pre-set dimension；
According to third image described in several, expand training set, the multiple image in the training set is training image；
By convolution method, feature extraction is carried out to the training image, obtains the automobile position in the training image；
Training parameter, including initial learning rate, initial learning rate adjusting parameter, momentum coefficient, regular factor are set；
According to training parameter, yolov2 model is trained using the automobile position in training image described in several, is obtained
Obtain the vehicle location identification model.
In one embodiment of the invention, in training parameter, initial learning rate can be 0.001, and training is to 30 and 40
Respectively multiplied by 0.1 after epoch；Momentum coefficient can be 00.9；Using the l2 norm of Model Weight as regular terms, canonical because
Son is 0.0005；Batch_size is set as 16.
In the embodiment of the present invention, vehicle location can actually be referred to centered on vehicle to be identified/one of center of gravity
Region, or can be a region comprising vehicle most information to be identified, or can be and believe comprising vehicle key to be identified
One region of breath.
It is described according to vehicle location and brand recognition model in step 130 in the embodiment of the present invention, obtain the vehicle
Identity information, comprising:
According to the vehicle location, the general image of vehicle to be identified is obtained；
According to color, to the vehicle classification to be identified；
To the general image of sorted vehicle to be identified, logo is extracted using AlexNet network；
According to logo, the vehicle brand is identified.
In fact, the brand recognition model of the embodiment of the present invention, can also there is other auxiliary informations, such as vehicle auxiliary is known
Not, five, seven, truck, cargo, extraordinary vehicle automobile etc. can be identified.
It is described according to third image in the embodiment of the present invention, expand training set, comprising:
Method using random cropping, random overturning, colour switching expands training set.
In the embodiment of the present invention, the method using random cropping expands training set, comprising:
The scale factor of random cropping is set.
It is described that training set is expanded using the method overturn at random in the embodiment of the present invention, comprising:
The probability of flip horizontal is set.
In the embodiment of the present invention, the method using colour switching expands training set, comprising:
The random variability of each Color Channel is set.
In the embodiment of the present invention, the road vehicle image that sample can be MSCOCO data set or manually mark, but sample
Originally it being limited, therefore in the embodiment of the present invention, the method using random cropping, random overturning, colour switching expands training set,
And it is trained.For example, the scale factor of random cropping can be 0.1；Random overturning can be the flip horizontal of 50% probability；
Color change method can be each Color Channel and change 10% at random.
The embodiment of the invention discloses a kind of testing vehicle register identification methods, which comprises obtains vehicle to be identified
Original image；According to vehicle location identification model, vehicle location is obtained from the original image；According to the vehicle location and
Brand recognition model obtains the identity information of the vehicle；The vehicle identity information includes vehicle brand, vehicle location identification
Model is based on deep neural network.The vehicle location identification model of the embodiment of the present invention is based on deep neural network, is based on depth
The complexity of the model of neural network algorithm is lower, can use GPU and carries out operation acceleration, accelerates the training speed of model, together
The classification speed of Shi Tigao model and accuracy, to improve the accuracy of testing vehicle register identification.
Fig. 2 show the schematic diagram of the testing vehicle register identification method of the embodiment of the present invention, as shown in Fig. 2, to include vehicle
One sub-picture of image, in the embodiment of the present invention, which is the original image of vehicle to be identified, identifies mould according to vehicle location
Type obtains vehicle location from the figure, such as 210 in Fig. 2.
Vehicle location 210 and brand recognition model according to fig. 2, first identify color, then identify vehicle brand
Equal identity informations.
In the testing vehicle register identification method of the embodiment of the present invention, vehicle location identification model is based on deep neural network, base
It is lower in the complexity of the model of deep neural network algorithm, it can use GPU and carry out operation acceleration, accelerate the training speed of model
Degree, while classification speed and the accuracy of model are improved, to improve the accuracy of testing vehicle register identification.
Fig. 3 show a kind of schematic diagram of vehicle identification device of the embodiment of the present invention, as shown in figure 3, the dress
It sets and includes:
Original image acquiring unit 310, for obtaining the original image of vehicle to be identified；
Vehicle location acquiring unit 320, for obtaining vehicle from the original image according to vehicle location identification model
Position；
Identity information acquiring unit 330, for obtaining the vehicle according to the vehicle location and brand recognition model
Identity information；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on depth nerve net
Network.
In the embodiment of the present invention, the vehicle location acquiring unit 320 includes:
Model construction subelement, for constructing yolov2 model or yolov3 model based on deep neural network；
Training subelement obtains the vehicle location for being trained to the yolov2 model or yolov3 model
Identification model.
In the vehicle identification device of the embodiment of the present invention, vehicle location identification model is based on deep neural network, base
It is lower in the complexity of the model of deep neural network algorithm, it can use GPU and carry out operation acceleration, accelerate the training speed of model
Degree, while classification speed and the accuracy of model are improved, to improve the accuracy of testing vehicle register identification.
Although the embodiments of the present invention have been disclosed as above, but its is not only in the description and the implementation listed
With it can be fully applied to various fields suitable for the present invention, for those skilled in the art, can be easily
Realize other modification, therefore without departing from the general concept defined in the claims and the equivalent scope, the present invention is simultaneously unlimited
In specific details and legend shown and described herein.
Claims (10)
1. a kind of testing vehicle register identification method, which is characterized in that the described method includes:
Obtain the original image of vehicle to be identified；
According to vehicle location identification model, vehicle location is obtained from the original image；
According to the vehicle location and brand recognition model, the identity information of the vehicle is obtained；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on deep neural network.
2. the method as described in claim 1, which is characterized in that it is described according to vehicle location identification model, from original image
Before obtaining vehicle location, the method also includes:
Construct yolov2 model or yolov3 model based on deep neural network；
The yolov2 model or yolov3 model are trained, the vehicle location identification model is obtained.
3. method according to claim 2, which is characterized in that it is described that yolov2 model or yolov3 model are trained,
Include:
Obtain several the first images；
From several the first images, the second image comprising automobile image is filtered out；
Several described second images are transformed to several third images of unified pre-set dimension；
According to third image described in several, expand training set, the multiple image in the training set is training image；
By convolution method, feature extraction is carried out to the training image, obtains the automobile position in the training image；
Training parameter, including initial learning rate, initial learning rate adjusting parameter, momentum coefficient, regular factor are set；
According to training parameter, yolov2 model is trained using the automobile position in training image described in several, obtains institute
State vehicle location identification model.
4. method as claimed in claim 3, which is characterized in that it is described according to vehicle location and brand recognition model, obtain institute
State the identity information of vehicle, comprising:
According to the vehicle location, the general image of vehicle to be identified is obtained；
According to color, to the vehicle classification to be identified；
To the general image of sorted vehicle to be identified, logo is extracted using AlexNet network；
According to logo, the vehicle brand is identified.
5. method as claimed in claim 3, which is characterized in that it is described according to third image, expand training set, comprising:
Method using random cropping, random overturning, colour switching expands training set.
6. method as claimed in claim 5, which is characterized in that the method using random cropping expands training set, comprising:
The scale factor of random cropping is set.
7. method as claimed in claim 5, which is characterized in that described to expand training set using the method overturn at random, comprising:
The probability of flip horizontal is set.
8. method as claimed in claim 5, which is characterized in that the method using colour switching expands training set, comprising:
The random variability of each Color Channel is set.
9. a kind of vehicle identification device, which is characterized in that described device includes:
Original image acquiring unit, for obtaining the original image of vehicle to be identified；
Vehicle location acquiring unit, for obtaining vehicle location from the original image according to vehicle location identification model；
Identity information acquiring unit, for obtaining the identity letter of the vehicle according to the vehicle location and brand recognition model
Breath；
Wherein, the vehicle identity information includes vehicle brand, and the vehicle location identification model is based on deep neural network.
10. device as claimed in claim 9, which is characterized in that the vehicle location acquiring unit includes:
Model construction subelement, for constructing yolov2 model or yolov3 model based on deep neural network；
Training subelement obtains the vehicle location identification for being trained to the yolov2 model or yolov3 model
Model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910008768.6A CN109766819A (en) | 2019-01-04 | 2019-01-04 | Testing vehicle register identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910008768.6A CN109766819A (en) | 2019-01-04 | 2019-01-04 | Testing vehicle register identification method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109766819A true CN109766819A (en) | 2019-05-17 |
Family
ID=66452594
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910008768.6A Pending CN109766819A (en) | 2019-01-04 | 2019-01-04 | Testing vehicle register identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109766819A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119726A (en) * | 2019-05-20 | 2019-08-13 | 四川九洲视讯科技有限责任公司 | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model |
| CN110321864A (en) * | 2019-07-09 | 2019-10-11 | 西北工业大学 | Remote sensing images explanatory note generation method based on multiple dimensioned cutting mechanism |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105160340A (en) * | 2015-08-31 | 2015-12-16 | 桂林电子科技大学 | Vehicle brand identification system and method |
| CN106127747A (en) * | 2016-06-17 | 2016-11-16 | 史方 | Car surface damage classifying method and device based on degree of depth study |
| CN106503710A (en) * | 2016-10-26 | 2017-03-15 | 北京邮电大学 | A kind of automobile logo identification method and device |
| CN108491797A (en) * | 2018-03-22 | 2018-09-04 | 河北省科学院应用数学研究所 | A kind of vehicle image precise search method based on big data |
| CN109002744A (en) * | 2017-06-06 | 2018-12-14 | 中兴通讯股份有限公司 | Image-recognizing method, device and video monitoring equipment |
-
2019
- 2019-01-04 CN CN201910008768.6A patent/CN109766819A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105160340A (en) * | 2015-08-31 | 2015-12-16 | 桂林电子科技大学 | Vehicle brand identification system and method |
| CN106127747A (en) * | 2016-06-17 | 2016-11-16 | 史方 | Car surface damage classifying method and device based on degree of depth study |
| CN106503710A (en) * | 2016-10-26 | 2017-03-15 | 北京邮电大学 | A kind of automobile logo identification method and device |
| CN109002744A (en) * | 2017-06-06 | 2018-12-14 | 中兴通讯股份有限公司 | Image-recognizing method, device and video monitoring equipment |
| CN108491797A (en) * | 2018-03-22 | 2018-09-04 | 河北省科学院应用数学研究所 | A kind of vehicle image precise search method based on big data |
Non-Patent Citations (1)
| Title |
|---|
| 张春玉等: "《一种有效的基于卷积神经网络的车辆检索算法》", 《科技创新与应用》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119726A (en) * | 2019-05-20 | 2019-08-13 | 四川九洲视讯科技有限责任公司 | A kind of vehicle brand multi-angle recognition methods based on YOLOv3 model |
| CN110321864A (en) * | 2019-07-09 | 2019-10-11 | 西北工业大学 | Remote sensing images explanatory note generation method based on multiple dimensioned cutting mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104573646B (en) | Chinese herbaceous peony pedestrian detection method and system based on laser radar and binocular camera | |
| CN105046196B (en) | Front truck information of vehicles structuring output method based on concatenated convolutional neutral net | |
| CN104217217B (en) | A kind of vehicle mark object detecting method and system based on two layers of classified | |
| WO2021238185A1 (en) | Object detection method and apparatus, electronic device, storage medium and program | |
| CN108710865A (en) | A kind of driver's anomaly detection method based on neural network | |
| CN103198315B (en) | Based on the Character Segmentation of License Plate of character outline and template matches | |
| CN106127747A (en) | Car surface damage classifying method and device based on degree of depth study | |
| CN107491753A (en) | A kind of parking offense detection method based on background modeling | |
| JP2018173756A (en) | Detection device, detection method, and detection program | |
| CN107862340A (en) | A kind of model recognizing method and device | |
| CN106407981A (en) | License plate recognition method, device and system | |
| CN105740886B (en) | A kind of automobile logo identification method based on machine learning | |
| CN108073926A (en) | A kind of licence plate recognition method and device | |
| CN109766819A (en) | Testing vehicle register identification method and device | |
| CN110226170A (en) | A kind of traffic sign recognition method in rain and snow weather | |
| CN105930803A (en) | Preceding vehicle detection method based on Edge Boxes and preceding vehicle detection device thereof | |
| EP3657386B1 (en) | Method and system for supporting an automated moving vehicle | |
| CN103530608A (en) | Vehicle type judgment method and vehicle type judgment device | |
| CN110135327B (en) | Driver behavior identification method based on multi-region feature learning model | |
| CN108205649A (en) | Driver drives to take the state identification method and device of phone | |
| JP2016200910A (en) | Driver state determination device | |
| CN108986255A (en) | Electronic device, driving behavior methods of marking and storage medium | |
| CN105893942A (en) | eSC and HOG-based adaptive HMM sign language identifying method | |
| CN109767298A (en) | The matched method and system of passenger's driver safety | |
| CN113706842B (en) | Method, equipment, storage medium and device for reminding article in car of leaving behind |
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: 20190517 |