CN108921060A - Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method - Google Patents
Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method Download PDFInfo
- Publication number
- CN108921060A CN108921060A CN201810632538.2A CN201810632538A CN108921060A CN 108921060 A CN108921060 A CN 108921060A CN 201810632538 A CN201810632538 A CN 201810632538A CN 108921060 A CN108921060 A CN 108921060A
- Authority
- CN
- China
- Prior art keywords
- clearance lamps
- motor vehicle
- license plate
- identified
- frame image
- 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
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
-
- 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
Abstract
The present invention provides a kind of motor vehicle based on deep learning and does not use according to regulations clearance lamps intelligent identification Method and system, storage medium, and this method includes：Traffic monitoring apparatus video flowing collected is obtained, the video flowing includes multiple image；Using Car license recognition neural network model, the license plate of each motor vehicle in each frame image is identified, and the position according to the license plate identified in each frame image, realizes the tracking to the corresponding motor vehicle of the license plate identified；Neural network model is identified using clearance lamps, the clearance lamps of each motor vehicle in each frame image is identified；The license plate that will identify that and the clearance lamps identified are matched, and according to state of the preset travel apart from interior clearance lamps in the trace of the corresponding motor vehicle of each license plate, judge whether the corresponding motor vehicle of the license plate uses according to regulations its clearance lamps.The present invention can increase the management intensity used to clearance lamps with intelligent recognition motor vehicle whether by proper use of clearance lamps is provided.
Description
Technical field
The present invention relates to field of intelligent transportation technology, and in particular to a kind of motor vehicle based on deep learning is not made by regulation
With clearance lamps intelligent identification Method and system, storage medium.
Background technique
Clearance lamps is the lamp of Chinese herbaceous peony, rear the extreme side, and on the roof of truck and also there is clearance lamps in side.Clearance lamps, light
Literally, " show " it is the meaning warned, " exterior feature " has the meaning of profile, so clearance lamps is a kind of car light of warning mark, uses
Come the signaling lamp for reminding other vehicles to pay attention to.It is mounted on the edge of vehicle roof, can indicate the height of vehicle but also is indicated
The width of vehicle.Safety standard provides that the motor vehicle for being higher than 3 meters in overall height must install clearance lamps, and the color of clearance lamps is preceding white
It is red afterwards.The relative position of clearance lamps and corresponding position lamp requires to be the most phase on apparent surface in the respective benchmark axis direction of two lamps
The adjacent o'clock projection spacing in a lateral vertical plane should be not less than 400mm.
Rearview mirror is seen when cloudy lane change, is basically shadow one if subsequent motor vehicle does not open side-marker lamp
Group, mixes with road surface, people is allowed to be difficult to see in rearview mirror, if especially accelerating the car to come up not open below shows width
Lamp is substantially difficult it is clear that being breakneck at this time in the case where heavy rain.
Currently, can not detect and identify automatically the scheme whether clearance lamps of motor vehicle is opened, and motor vehicle shows
The phenomenon that wide lamp is not opened by regulation largely exists in the motor vehicle of highway driving, and security risk is huge.
Summary of the invention
(1) the technical issues of solving
In view of the deficiencies of the prior art, it does not use according to regulations and shows the present invention provides a kind of motor vehicle based on deep learning
Whether wide lamp intelligent identification Method and system, storage medium, can be with intelligent recognition motor vehicle by the proper use of clearance lamps of regulation.
(2) technical solution
In order to achieve the above object, the present invention is achieved by the following technical programs：
In a first aspect, the present invention, which provides a kind of motor vehicle based on deep learning, does not use according to regulations clearance lamps intelligent recognition
Method, including：Traffic monitoring apparatus video flowing collected is obtained, the video flowing includes multiple image；Using preparatory training
Car license recognition neural network model, the license plate of each motor vehicle in each frame image is identified, and according to the vehicle identified
The tracking to the corresponding motor vehicle of the license plate identified is realized in position of the board in each frame image；Using showing for training in advance
Wide lamp identifies neural network model, identifies to the clearance lamps of each motor vehicle in each frame image；The license plate that will identify that and
The clearance lamps identified is matched, and is shown according to preset travel in the trace of the corresponding motor vehicle of each license plate
The state of wide lamp, judges whether the corresponding motor vehicle of the license plate uses according to regulations its clearance lamps.
In some embodiments, it is described before the clearance lamps to each motor vehicle in each frame image and license plate identify
Method further includes：According to pre-set perspective inverse transformation parameter, perspective inverse transformation is carried out to each frame image.
In some embodiments, the setting up procedure of the perspective inverse transformation parameter includes：From the traffic monitoring apparatus
Detection zone is selected in monitoring area, the detection zone is opposite with the clearance lamps identification input data of neural network model
It answers；Using least square method, the perspective inverse transformation parameter of each pixel in the detection zone is determined.
In some embodiments, described to select detection zone out of the traffic monitoring apparatus monitoring area, including：?
In the monitoring area along lane line choose 4 pixels, four endpoint phase mappings of 4 pixels and a rectangle,
The size of the rectangle is identical as the clearance lamps identification number of input data of neural network model.
In some embodiments, the training process of the Car license recognition neural network model includes：Building the first training sample
This collection, first training sample concentrate the first picture including multiple label license plates, and first picture is set by traffic monitoring
Standby acquisition in advance；Multiple first pictures that first training sample is concentrated are trained using yolo algorithm, obtain license plate
Identify each model parameter of neural network model.
In some embodiments, the training process of the clearance lamps identification neural network model includes：The second training of building
Sample set, second training sample concentrate the second picture including multiple label clearance lamps and its state, the second picture
It is acquired in advance by traffic monitoring apparatus；It is instructed using multiple second pictures that yolo algorithm concentrates second training sample
Practice, obtains each model parameter of clearance lamps identification neural network model.
Second aspect, the present invention provide a kind of motor vehicle based on deep learning and do not use according to regulations clearance lamps intelligent recognition
System, including：
Module is obtained, for obtaining traffic monitoring apparatus video flowing collected, the video flowing includes multiple image；
First identification module, for using Car license recognition neural network model trained in advance, to each in each frame image
The license plate of motor vehicle is identified, and the position according to the license plate identified in each frame image, is realized to the vehicle identified
The tracking of the corresponding motor vehicle of board；
Second identification module, for identifying neural network model using clearance lamps trained in advance, to every in each frame image
The clearance lamps of one motor vehicle is identified；
Judgment module, the license plate for will identify that are matched with the clearance lamps identified, and according to each license plate pair
State of the preset travel apart from interior clearance lamps in the trace for the motor vehicle answered, judges whether the corresponding motor vehicle of the license plate is pressed
Regulation uses its clearance lamps.
In some embodiments, system further includes：
Inverse transform block, for license plate of first identification module to each motor vehicle in each frame image carry out identification and
Before second identification module identifies the clearance lamps of each motor vehicle in each frame image, according to pre-set
Depending on inverse transformation parameter, perspective inverse transformation is carried out to each frame image.
The third aspect, the present invention provide a kind of storage medium, are stored thereon with computer program, in the computer program
Method as above can be realized when being executed by processor.
(3) beneficial effect
The embodiment of the invention provides a kind of motor vehicles based on deep learning not to use according to regulations clearance lamps intelligent recognition
Method and system, storage medium identify the license plate in each frame image using Car license recognition neural network model, and use and show exterior feature
Lamp identification neural network model identifies the clearance lamps in each frame image, then matches clearance lamps and license plate, Jin Ergen
Whether clearance lamps is used according to regulations with motor vehicle corresponding to the matched license plate of clearance lamps according to the state judgement of clearance lamps.As it can be seen that
The present invention can increase the management power used to clearance lamps with intelligent recognition motor vehicle whether by proper use of clearance lamps is provided
Degree, prevention reduce all kinds of traffic accidents caused by the not proper use of clearance lamps of turnpike driving, reduce traffic injuries and deaths and
Mitigate traffic injury, ensures the security of the lives and property of country, society and broad masses of the people conscientiously.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 a and Fig. 1 b, which show the motor vehicle in one embodiment of the invention based on deep learning and do not use according to regulations, shows exterior feature
The flow diagram of lamp intelligent identification Method；
Fig. 2 shows the views that motor line in one embodiment of the invention sails to a frame image nearby；
Fig. 3 shows the schematic diagram that motor line in one embodiment of the invention sails to a frame image of distant place；
Fig. 4 shows the schematic diagram for selecting detection zone in one embodiment of the invention in monitoring area；
Fig. 5 shows the flow diagram of offline part in one embodiment of the invention；
Fig. 6 shows the motor vehicle in one embodiment of the invention based on deep learning and does not use according to regulations clearance lamps intelligently knowledge
The structural block diagram of other system.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
In a first aspect, the embodiment of the present invention, which provides a kind of motor vehicle based on deep learning, does not use according to regulations clearance lamps intelligence
Energy recognition methods, as illustrated in figs. 1A and ib, this method includes：
S1, traffic monitoring apparatus video flowing collected is obtained, which includes multiple image；
It will be appreciated that traffic monitoring apparatus can be, but not limited to be arranged in the multiple area of expressway traffic accident, highway
The regions such as complicated highway section, high speed bridge floor entrance, city expressway accident-prone area.
S2, using Car license recognition neural network model trained in advance, to the license plate of each motor vehicle in each frame image into
Row identification, and the position according to the license plate identified in each frame image, are realized to the corresponding motor vehicle of the license plate identified
Tracking；
It will be appreciated that above-mentioned multiple image is inputted above-mentioned Car license recognition neural network model, just appeared in
State the license plate of each motor vehicle in multiple image；
What it is due to traffic monitoring apparatus monitoring is a distance, it is therefore desirable to track to motor vehicle, pass through vehicle here
Board is tracked, specially：It first identifies license plate, then the change in location according to license plate in front of and after frames image, realizes to machine
The tracking of motor-car.
Wherein, the training process of Car license recognition neural network model may include following process：
S21, the first training sample set of building, first training sample concentrate the first figure including multiple label license plates
Piece, first picture are acquired in advance by traffic monitoring apparatus；
S22, multiple first pictures that first training sample is concentrated are trained using yolo algorithm, obtain license plate
Identify each model parameter of neural network model.
It will be appreciated that S22 obtains each model parameter of Car license recognition neural network model to get trained vehicle is arrived
Board identifies neural network model.
For example, it as shown in figure 5, label license plate picture is marked the license plate in the first picture, then uses
Yolo algorithm is trained, and obtains the parameter of the model for car plate detection.
S3, neural network model is identified using clearance lamps trained in advance, exterior feature is shown to each motor vehicle in each frame image
Lamp is identified；
It will be appreciated that the identification to clearance lamps, is the identification to clearance lamps position in each frame image, clearance lamps
State can determine that the state of clearance lamps includes opening and closing two states according to the pixel of position.
Wherein, the training method of clearance lamps identification neural network model can refer to the instruction of Car license recognition neural network model
Practice method, can specifically include following process：
S31, the second training sample set of building, it includes multiple label clearance lamps and its state that second training sample, which is concentrated,
Second picture, the second picture acquires in advance by traffic monitoring apparatus；
As shown in Figures 2 and 3, second picture can be used as to the picture that clearance lamps is marked using rectangle frame.
S32, it is trained using multiple second pictures that yolo algorithm concentrates second training sample, obtains showing exterior feature
Each model parameter of lamp identification neural network model.
For example, then right as shown in figure 5, label car light picture is marked clearance lamps in second picture
The picture marked is trained using yolo algorithm, is obtained for the car light i.e. parameter of the model of clearance lamps detection.
S4, the license plate that will identify that and the clearance lamps identified are matched, and according to the corresponding motor vehicle of each license plate
Trace in state of the preset travel apart from interior clearance lamps, judge whether the corresponding motor vehicle of the license plate uses according to regulations it
Clearance lamps.
It will be appreciated that carrying out matched process to the license plate and clearance lamps identified here, will actually be located at same
Clearance lamps and license plate on motor vehicle carry out corresponding process.
It will be appreciated that the preset travel distance in the trace of each motor vehicle, refers to that traffic monitoring apparatus is supervised
The stretch line in motor-driven vehicle going route controlled, this section of route can be pre-set effective operating range.For example, being supervised
The motor-driven vehicle going route controlled is 500 meters, then preset travel distance is 100 meters in this 500 meters.
It will be appreciated that operating motor vehicles are not travelled on a highway by regulation under low visibility meteorological condition, if
Do not accomplish following three, it will be by 6 points of punishment：
In bad weather, it should accomplish：
1, when visibility is less than 200 meters, fog lamp, dipped headlight, clearance lamps and front and back fog lamp are opened, speed must not exceed per small
When 60 kilometers, kept at a distance from 100 meters or more with same lane front truck；
2, when visibility is less than 100 meters, fog lamp, dipped headlight, clearance lamps, front and back fog lamp and hazard warning lamp are opened,
Speed must not exceed 40 kilometers per hour, keep at a distance from 50 meters or more with same lane front truck；
3, when visibility is less than 50 meters, fog lamp, dipped headlight, clearance lamps, front and back fog lamp and hazard warning lamp, vehicle are opened
Speed must not exceed 20 kilometers per hour, and sail out of highway as early as possible from nearest outlet.
As it can be seen that, when low visibility is in 200 meters, should just open clearance lamps according to traffic law.
For example, when air visibility be 150 meters when, a license plate be xxxxx motor vehicle in monitoring area most
The state of clearance lamps is in off state in 100 meters afterwards, it may be considered that the motor vehicle does not use clearance lamps according to the rules.
Certainly, it needs the case where opening clearance lamps, no longer to illustrate one by one here there is also other.
Method provided by the invention identifies the license plate in each frame image using Car license recognition neural network model, and adopts
The clearance lamps in each frame image is identified with clearance lamps identification neural network model, then matches clearance lamps and license plate,
And then whether the motor vehicle according to corresponding to the judgement of the state of clearance lamps and clearance lamps matched license plate uses according to regulations clearance lamps.
As it can be seen that the present invention can increase the pipe used to clearance lamps with intelligent recognition motor vehicle whether by proper use of clearance lamps is provided
Reason dynamics, prevention reduce all kinds of traffic accidents caused by the not proper use of clearance lamps of turnpike driving, reduce traffic accident injury
Traffic injury is died and mitigated, ensures the security of the lives and property of country, society and broad masses of the people conscientiously.
It in some embodiments, can also be right according to pre-set perspective inverse transformation parameter before executing S2 and S3
Each frame image carries out perspective inverse transformation.
As shown in Fig. 2, car light usually nearby is larger, information compares redundancy；As shown in figure 3, the car light of distant place is smaller, depending on
Wild openr but useless information is also more, and key message is less.General vehicle reaches at a distance, and projection of the car light in camera is non-
Often small, detection and identification for car light can be relatively difficult, and carrying out perspective inverse transformation to image here can remove nearby
Information useless in the broad horizon of distant place is removed, and enhances the key messages such as car light by redundancy, conducive to the knowledge of clearance lamps
Not.
Wherein, since perspective inverse transformation parameter is pre-set, such as offline setting, specific setting up procedure include：From
Detection zone is selected in the monitoring area of the traffic monitoring apparatus, the detection zone and the clearance lamps identify neural network
The input data of model is corresponding；Using least square method, the perspective inverse transformation ginseng of each pixel in the detection zone is determined
Number.
For example, as shown in figure 5, along choosing lane detection zone, the detection zone of selection and neural network model
It inputs corresponding, the perspective inverse transformation parameter of the pixel in detection zone is then determined using least square method.
Above-mentioned detection zone is a part of monitoring area, can choose the region where important information as detection zone
Domain can neglect garbage in this way.Region where important information is generally the region between two lane lines, and for
Place too far can also be ignored due to being unfavorable for detecting, therefore the selection course of detection zone may include：Such as Fig. 4
It is shown, 4 pixels, four endpoints of 4 pixels and a rectangle are chosen along lane line in the monitoring area
Phase mapping, the size of the rectangle are identical as the clearance lamps identification number of input data of neural network model.
For example, the number of the input data of clearance lamps identification neural network model is 600*600, then the four of the rectangle mapped
A endpoint can be respectively (0,0), (600,0), (600,600), (0,600).
The mapping relations between the position in pixel and world coordinate system in each frame image can be expressed as：
In above formula, X, Y, Z are each point in world coordinate system；X, y is each pixel in the first image, and mij is
Penetrate transformation parameter.Projection transformation parameter can be obtained using least square method.
Second aspect, the embodiment of the present invention also provide a kind of motor vehicle based on deep learning and do not use according to regulations clearance lamps
Intelligent identifying system, as shown in fig. 6, the system includes：
Module is obtained, for obtaining traffic monitoring apparatus video flowing collected, the video flowing includes multiple image；
First identification module, for using Car license recognition neural network model trained in advance, to each in each frame image
The license plate of motor vehicle is identified, and the position according to the license plate identified in each frame image, is realized to the vehicle identified
The tracking of the corresponding motor vehicle of board；
Second identification module, for identifying neural network model using clearance lamps trained in advance, to every in each frame image
The clearance lamps of one motor vehicle is identified；
Judgment module, the license plate for will identify that are matched with the clearance lamps identified, and according to each license plate pair
State of the preset travel apart from interior clearance lamps in the trace for the motor vehicle answered, judges whether the corresponding motor vehicle of the license plate is pressed
Regulation uses its clearance lamps.
In some embodiments, system further includes：
Inverse transform block, for license plate of first identification module to each motor vehicle in each frame image carry out identification and
Before second identification module identifies the clearance lamps of each motor vehicle in each frame image, according to pre-set
Depending on inverse transformation parameter, perspective inverse transformation is carried out to each frame image.
In some embodiments, further include：
Parameter setting module, for selecting detection zone, the detection out of the traffic monitoring apparatus monitoring area
Region is corresponding with the clearance lamps identification input data of neural network model；Using least square method, the detection is determined
The perspective inverse transformation parameter of each pixel in region.
In some embodiments, parameter setting module is specifically used for：4 pictures are chosen along lane line in the monitoring area
Four endpoint phase mappings of vegetarian refreshments, 4 pixels and a rectangle, the size of the rectangle and clearance lamps identification nerve
The number of the input data of network model is identical.
In some embodiments, system further includes：
First model construction module, for constructing the first training sample set, it includes multiple that first training sample, which is concentrated,
The first picture of license plate is marked, first picture is acquired in advance by traffic monitoring apparatus；Using yolo algorithm to described first
Multiple first pictures that training sample is concentrated are trained, and obtain each model parameter of Car license recognition neural network model.
In some embodiments, system further includes：
Second model construction module, for constructing the second training sample set, it includes multiple that second training sample, which is concentrated,
The second picture of clearance lamps and its state is marked, the second picture is acquired in advance by traffic monitoring apparatus；Using yolo algorithm
Multiple second pictures concentrated to second training sample are trained, and obtain each of clearance lamps identification neural network model
Model parameter.
It will be appreciated that system provided by the invention is corresponding with method provided by the invention, explanation, act in relation to content
The contents such as example and beneficial effect can be with reference to the corresponding portion in the above method.
The third aspect, the embodiment of the present invention provide a kind of storage medium, are stored thereon with computer program, in the calculating
Machine program can realize the above method when being executed by processor.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
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
Invention is explained in detail, those skilled in the art should understand that：It still can be to aforementioned each implementation
Technical solution documented by example is modified or equivalent replacement of some of the technical features；And these modification or
Replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (9)
1. a kind of motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method, which is characterized in that including：
Traffic monitoring apparatus video flowing collected is obtained, the video flowing includes multiple image；
Using Car license recognition neural network model trained in advance, the license plate of each motor vehicle in each frame image is identified,
And the position according to the license plate identified in each frame image, realize the tracking to the corresponding motor vehicle of the license plate identified；
Neural network model is identified using clearance lamps trained in advance, the clearance lamps of each motor vehicle in each frame image is known
Not；
The license plate that will identify that and the clearance lamps identified are matched, and according to the tracking road of the corresponding motor vehicle of each license plate
State of the preset travel apart from interior clearance lamps in line, judges whether the corresponding motor vehicle of the license plate uses according to regulations its clearance lamps.
2. the method as described in claim 1, which is characterized in that clearance lamps and license plate to each motor vehicle in each frame image into
Before row identification, the method also includes：According to pre-set perspective inverse transformation parameter, perspective contravariant is carried out to each frame image
It changes.
3. according to the method described in claim 2, it is characterized in that, the setting up procedure of the perspective inverse transformation parameter includes：
Detection zone, the detection zone and clearance lamps identification mind are selected out of the traffic monitoring apparatus monitoring area
Input data through network model is corresponding；
Using least square method, the perspective inverse transformation parameter of each pixel in the detection zone is determined.
4. according to the method described in claim 3, it is characterized in that, described select out of the traffic monitoring apparatus monitoring area
Detection zone is selected, including：
4 pixels, four endpoint phases of 4 pixels and a rectangle are chosen along lane line in the monitoring area
Mapping, the size of the rectangle are identical as the clearance lamps identification number of input data of neural network model.
5. method according to claims 1 to 4, which is characterized in that the Car license recognition neural network model was trained
Journey includes：
Construct the first training sample set, first training sample concentrate include multiple label license plates the first picture, described the
One picture is acquired in advance by traffic monitoring apparatus；
Multiple first pictures that first training sample is concentrated are trained using yolo algorithm, obtain Car license recognition nerve
Each model parameter of network model.
6. described in any item methods according to claim 1~5, which is characterized in that the clearance lamps identifies neural network model
Training process include：
The second training sample set is constructed, second training sample concentrates the second figure including multiple label clearance lamps and its state
Piece, the second picture are acquired in advance by traffic monitoring apparatus；
It is trained using multiple second pictures that yolo algorithm concentrates second training sample, obtains clearance lamps identification mind
Each model parameter through network model.
7. a kind of motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identifying system, which is characterized in that including：
Module is obtained, for obtaining traffic monitoring apparatus video flowing collected, the video flowing includes multiple image；
First identification module, for using Car license recognition neural network model trained in advance, to each motor-driven in each frame image
The license plate of vehicle is identified, and the position according to the license plate identified in each frame image, is realized to the license plate pair identified
The tracking for the motor vehicle answered；
Second identification module, for identifying neural network model using clearance lamps trained in advance, to each machine in each frame image
The clearance lamps of motor-car is identified；
Judgment module, the license plate for will identify that are matched with the clearance lamps identified, and corresponding according to each license plate
Whether state of the preset travel apart from interior clearance lamps in the trace of motor vehicle judges the corresponding motor vehicle of the license plate by regulation
Use its clearance lamps.
8. system according to claim 7, which is characterized in that further include：
Inverse transform block, for license plate of first identification module to each motor vehicle in each frame image carry out identification and it is described
It is anti-according to pre-set perspective before second identification module identifies the clearance lamps of each motor vehicle in each frame image
Transformation parameter carries out perspective inverse transformation to each frame image.
9. a kind of storage medium, is stored thereon with computer program, which is characterized in that held in the computer program by processor
Claim 1~6 described in any item methods can be realized when row.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810632538.2A CN108921060A (en) | 2018-06-20 | 2018-06-20 | Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810632538.2A CN108921060A (en) | 2018-06-20 | 2018-06-20 | Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108921060A true CN108921060A (en) | 2018-11-30 |
Family
ID=64420629
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810632538.2A Pending CN108921060A (en) | 2018-06-20 | 2018-06-20 | Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921060A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI684920B (en) * | 2018-12-05 | 2020-02-11 | 財團法人資訊工業策進會 | Headlight state analysis method, headlight state analysis system, and non-transitory computer readable media |
| DE102021200000A1 (en) | 2021-01-03 | 2022-07-07 | Volkswagen Aktiengesellschaft | Method for generating an image of a motor vehicle on a display device of a user interface for an infotainment system of the motor vehicle and motor vehicle for carrying out the method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102221358A (en) * | 2011-03-23 | 2011-10-19 | 中国人民解放军国防科学技术大学 | Monocular visual positioning method based on inverse perspective projection transformation |
| CN102567380A (en) * | 2010-12-28 | 2012-07-11 | 沈阳聚德视频技术有限公司 | Method for searching vehicle information in video image |
| CN204440645U (en) * | 2015-02-16 | 2015-07-01 | 北京精英智通科技股份有限公司 | A kind of driving at night illegal use high beam behavior automatic evidence-collecting equipment |
| CN105430798A (en) * | 2015-12-14 | 2016-03-23 | 山西省交通科学研究院 | Daytime running outline marker lamp system of motor vehicle |
| CN106462762A (en) * | 2016-09-16 | 2017-02-22 | 香港应用科技研究院有限公司 | Detection, tracking and positioning of vehicle based on enhanced inverse perspective mapping |
| CN106934378A (en) * | 2017-03-16 | 2017-07-07 | 山东建筑大学 | A kind of dazzle light identifying system and method based on video depth study |
-
2018
- 2018-06-20 CN CN201810632538.2A patent/CN108921060A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567380A (en) * | 2010-12-28 | 2012-07-11 | 沈阳聚德视频技术有限公司 | Method for searching vehicle information in video image |
| CN102221358A (en) * | 2011-03-23 | 2011-10-19 | 中国人民解放军国防科学技术大学 | Monocular visual positioning method based on inverse perspective projection transformation |
| CN204440645U (en) * | 2015-02-16 | 2015-07-01 | 北京精英智通科技股份有限公司 | A kind of driving at night illegal use high beam behavior automatic evidence-collecting equipment |
| CN105430798A (en) * | 2015-12-14 | 2016-03-23 | 山西省交通科学研究院 | Daytime running outline marker lamp system of motor vehicle |
| CN106462762A (en) * | 2016-09-16 | 2017-02-22 | 香港应用科技研究院有限公司 | Detection, tracking and positioning of vehicle based on enhanced inverse perspective mapping |
| CN106934378A (en) * | 2017-03-16 | 2017-07-07 | 山东建筑大学 | A kind of dazzle light identifying system and method based on video depth study |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI684920B (en) * | 2018-12-05 | 2020-02-11 | 財團法人資訊工業策進會 | Headlight state analysis method, headlight state analysis system, and non-transitory computer readable media |
| DE102021200000A1 (en) | 2021-01-03 | 2022-07-07 | Volkswagen Aktiengesellschaft | Method for generating an image of a motor vehicle on a display device of a user interface for an infotainment system of the motor vehicle and motor vehicle for carrying out the method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102117546B (en) | On-vehicle traffic light assisting device | |
| CN105216797B (en) | Method of overtaking and system | |
| CN100583125C (en) | Vehicle intelligent back vision method | |
| CN102768801B (en) | Method for detecting motor vehicle green light follow-up traffic violation based on video | |
| CN101984478B (en) | Abnormal S-type driving warning method based on binocular vision lane marking detection | |
| KR101671428B1 (en) | Intelligent Monitoring System For Violation Vehicles in crossroads | |
| CN104537841A (en) | Unlicensed vehicle violation detection method and detection system thereof | |
| CN103308056A (en) | Road marking detection method | |
| CN102521983A (en) | Vehicle violation detection system based on high definition video technology and method thereof | |
| CN201397576Y (en) | Device for automatically shooting picture of the illegal turning of vehicles at crossings | |
| CN102963294A (en) | Method for judging opening and closing states of high beam of vehicle driving at night | |
| CN101469991A (en) | All-day structured road multi-lane line detection method | |
| CN110782673A (en) | Vehicle violation identification and detection system based on unmanned aerial vehicle shooting cloud computing | |
| CN104008650A (en) | Motor vehicle high beam monitoring device and identification method thereof | |
| CN105654073A (en) | Automatic speed control method based on visual detection | |
| CN103832357B (en) | A kind of lane-departure warning system and method based on machine vision | |
| CN112329553B (en) | Lane line marking method and device | |
| CN111231833A (en) | Automobile auxiliary driving system based on combination of holographic projection and AR | |
| CN114387785A (en) | Safety management and control method and system based on intelligent highway and storable medium | |
| CN108921060A (en) | Motor vehicle based on deep learning does not use according to regulations clearance lamps intelligent identification Method | |
| CN108986476A (en) | Motor vehicle does not use according to regulations high beam recognition methods, system and storage medium | |
| Sisiopiku | Variable speed control: technologies and practice | |
| CN108510754A (en) | Violation driving behavior alarming device and method | |
| CN112532921A (en) | Water conservancy system intelligent monitoring implementation method and system | |
| CN208781404U (en) | A kind of traffic light control system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181130 |