CN107633213A - A kind of pilotless automobile traffic lights recognition methods - Google Patents
A kind of pilotless automobile traffic lights recognition methods Download PDFInfo
- Publication number
- CN107633213A CN107633213A CN201710787414.7A CN201710787414A CN107633213A CN 107633213 A CN107633213 A CN 107633213A CN 201710787414 A CN201710787414 A CN 201710787414A CN 107633213 A CN107633213 A CN 107633213A
- Authority
- CN
- China
- Prior art keywords
- traffic lights
- camera
- image
- information
- pilotless automobile
- 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
- Traffic Control Systems (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of pilotless automobile traffic lights recognition methods, including:Camera gathers traffic lights image information;Image information analysis is handled, turned right, keep straight on corresponding to traffic lights information;When camera gathers traffic lights image information, camera standard time for exposure T and time for exposure fluctuation threshold values Δ T are preset, in shooting image, the time for exposure of camera fluctuates for the standard exposure time fluctuates threshold values, as T Δs T~T+ Δs T.Advantages of the present invention is:Solve pilotless automobile and easily occur traffic lights display frequency and image acquisition time reverse operation in terms of IMAQ, cause camera gather image traffic lights region is all black picture sometimes, the defects of so as to cause camera blinding, it ensure that correctness of the pilotless automobile in traffic lights IMAQ;Adjusted by camera exposure, focal length short focus camera coordinate ensure that collection traffic lights image accuracy, for pilotless automobile by image acquisition traffic configured information basis is provided.
Description
Technical field
The present invention relates to unmanned car steering field, the more particularly to recognition methods of pilotless automobile traffic lights.
Background technology
For the autonomous driving vehicle of city operating mode, the identification of crossroad access signal lamp is with vehicle automatic start-stop with turning
To being the prerequisite function of pilotless automobile.Wherein, crossroad access signal lamp identification technology is whether the function may be used
With the key of normal work.Traffic lights identification technology relates generally to two key technology research:Traffic lights color segmentation with
Arrowhead-shaped indicator lamp directional information identification.Traffic lights color segmentation, which is mainly studied, is mapped to coloured image in different color skies
Between, the threshold value section being then distributed according to red light with green light extracts red light region and green light region.Arrowhead-shaped indicator lamp
The main research of directional information identification uses image processing techniques, and the left-hand rotation of arrowhead-shaped indicator lamp, right-hand rotation, straight trip information are identified
Come, for controlling vehicle crossroad to turn to.But it is obtaining of striving for that the premise of lights state, which is calculated, in analysis
To the picture of lights state.Prior art by processor obtains traffic lights after being gathered using camera by graphical analysis
State, so as to control motor turning start and stop, but traffic lights show the information such as time, arrow image using LED, due to can because
Traffic lights display frequency and image acquisition time reverse operation cause the image that camera gathers, and traffic lights region is completely black figure sometimes
Picture, so as to cause camera blinding, it can not correctly control vehicle start and stop and steering.I.e.:When traffic lights are shown, because light shows sometimes
The light on and off time shown and the time of IMAQ are on the contrary, it is all black that the traffic lights parts of images for causing camera to gather, which continues some frames,
Color, so as to have impact on the normal detection of traffic lights.
The content of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of pilotless automobile traffic lights identification side
Method, this method can reduce pilotless automobile due to can not correctly gather traffic lights image and caused by lights state identify
The defects of mistake.
To achieve these goals, the technical solution adopted by the present invention is:A kind of pilotless automobile traffic lights identification side
Method, including:
Camera gathers traffic lights image information;
Image information analysis is handled, turned right, keep straight on corresponding to traffic lights information;
It is characterized in that:When camera gathers traffic lights image information, camera standard time for exposure T and time for exposure ripple are preset
Dynamic threshold values Δ T, in shooting image, the time for exposure of camera fluctuates for the standard exposure time fluctuates threshold values, as T- Δs T
~T+ Δs T.
Camera for gathering image is two, respectively focal length camera and short focus camera.
Described focal length camera and short focus camera work simultaneously gathers the image information of traffic lights.
Traffic signals identification is carried out to the image of focal length camera, the collection of short focus camera respectively, if only one camera
Traffic lights information is identified, then with the transport information control pilotless automobile.
If both focal length camera, short focus camera collection image recognize traffic lights information, judge focal length camera with it is short
Whether burnt camera collection is the traffic lights image of same position, if so, judge whether the traffic lights information of identification is consistent, if one
Cause, gather any recognition result control pilotless automobile;If inconsistent, the traffic lights letter of two camera collections is obtained respectively
Stability is ceased, wherein traffic lights information stability judges according to the traffic lights information change number of continuous multiple frames image, chooses
The small traffic lights information control pilotless automobile of change frequency, when the image that change frequency is identical, and selection short focus camera gathers
To obtain traffic lights information.
If judge that focal length camera and short focus camera gather whether be same position traffic lights image result be it is no,
The distance selection of the traffic lights and automobile corresponding to traffic lights image gathered according to camera is wherein apart from the figure of small traffic lights
As obtaining traffic light status.
Focal length camera and short focus camera are judged according to the distance of traffic lights and automobile corresponding to the traffic lights image of collection
Collection whether be same position traffic lights image.
The advantage of the invention is that:Solve pilotless automobile and easily occur traffic lights display frequency in terms of IMAQ
Rate and image acquisition time reverse operation, causing the image that camera gathers, traffic lights region is all black picture sometimes, so as to cause
The defects of camera blinding, it ensure that correctness of the pilotless automobile in traffic lights IMAQ;Using focal length camera and short focus
The IMAQ strategy that camera coordinates, no matter automobile can obtain clearly as far as possible apart from traffic lights distance
Image information;Adjusted by camera exposure, focal length short focus camera coordinate ensure that collection traffic lights image accuracy, be nothing
People's driving obtains traffic configured information by image and provides basis.
Brief description of the drawings
Mark in the content and figure expressed below each width accompanying drawing of description of the invention is briefly described:
Fig. 1 is prior art pilotless automobile traffic lights identification process figure;
Fig. 2 is pilotless automobile focal length short focus principle schematic of the present invention;
Fig. 3 is that traffic lights image of the present invention obtains principle flow chart;
Embodiment
It is further detailed to the embodiment work of the present invention by the description to optimum embodiment below against accompanying drawing
Thin explanation.
As shown in figure 1, it is the traffic lights recognition methods flow chart of pilotless automobile in the prior art, first by phase
Machine gathers traffic lights image information;Secondly image information is analyzed and processed, turned right, keep straight on corresponding to traffic lights
Information;The entire car controller of pilotless automobile controls the start and stop steering operation of automobile according to obtained traffic lights information.It is existing
Camera collection traffic lights image information uses the camera that the time for exposure determines in technology, and is that single camera is used to control.Obtain
View data the color and directional information of traffic lights are obtained by image processing algorithm, so as to for pilotless automobile fortune
Row, which provides, turns to the support of start and stop data.
In a preferred embodiment of the invention, during camera collection traffic lights image information, camera standard exposure is preset
Time T and time for exposure fluctuation threshold values Δ T, in shooting image, the time for exposure of camera fluctuates for the standard exposure time
Fluctuate threshold values, as T- Δs T~T+ Δs T.In the image of continuous acquisition traffic lights, when shooting every time, the time for exposure of camera
It is different, its time for exposure is controlled by microprocessor, controls its time for exposure to be fluctuated between T- Δ T~T+ Δs T, standard
Time for exposure T is default threshold values for successive time for exposure, Δ T in the prior art, can be with 10% or so of T values, can be according to tool
Body demand is set.When taking pictures, multiple exposure values can be preset between T- Δ T~T+ Δs T, then be pre-stored in microprocessor
In, then shooting when, obtained at random by microprocessor multiple exposure values set in advance one of them be used for control camera
Image information is gathered using the exposure value.
As shown in Fig. 2 in a preferred embodiment of the present invention, camera uses two, respectively focal length camera and short focus
Camera, two cameras work simultaneously gathers traffic lights image information.Focal length camera, short focus camera are all connected with microprocessor, camera
Work controlled by microprocessor, image identification unit, graphics processing unit are set in microprocessor, for knowing to image
Not, traffic lights information is obtained, arrow corresponding to red green information and red green information of the traffic lights information including traffic lights points to letter
Breath, and range information of the traffic lights apart from camera position corresponding to the image of the collection identified by monocular camera.
Focal length camera and the traffic lights image of short focus camera collection handle identification through the image processing software in microprocessor
After obtain traffic lights information.If the image for only having a camera in both focal length camera, short focus camera is known by microprocessor
When not going out traffic lights information i.e. traffic signal light condition, the traffic lights information identified using this is as pilotless automobile
The foundation of control.Microprocessor identifies through traffic lights image recognition software, obtains the value of feedback of the result of both identifications, if only
One recognition result feedback, then the foundation control automobile controlled using the result as pilotless automobile open according to traffic lights information
Stop travelling.
If the recognition result for receiving acquisition is two, i.e., focal length camera, the traffic lights of short focus camera collection are identified respectively
Image receives recognition result feedback, then needs to judge traffic light position information.If focal length camera, short focus phase
When traffic lights location matches in machine are unanimously that position is identical, then the friendship identified in focal length camera, short focus camera is determined whether
Whether traffic light status are identical in logical lamp information, if identical, using any of which camera recognition result, if traffic light status
It is inconsistent, then it is used as according to identification traffic behavior stability and selects race's foundation.Traffic behavior stability can (N values be excellent according to continuous N frames
Selecting 5-10) image traffic light status change frequencies judges.In lasting N two field pictures, if traffic lights identification state changes
It is bigger, show that identification information is more unstable, choose stable recognition result as automobile control signal.If traffic lights identification is steady
Qualitatively judge identical, then the magazine state of preferred short focus is as final identification information.If the traffic of two camera calibrations
Lamp positional information can not match, then select wherein nearest traffic light status as recognition result.
As shown in figure 3, if the image of both focal length camera, short focus camera collection recognizes traffic lights information, judge to grow
The collection of burnt camera and short focus camera whether be same position traffic lights image, the friendship according to corresponding to the traffic lights image of collection
The distance of logical lamp and automobile come judge focal length camera and the collection of short focus camera whether be same position traffic lights image, it is long
Defocus distance is that S1 short focus distances are S2, during apart from identical S1=S2, then traffic lights position phase corresponding to the traffic lights image that gathers
Together.
If judgement is the traffic lights image of same position, judges whether the traffic lights information of identification is consistent, identify traffic
Lamp information includes arrow directional information corresponding to lights state, traffic lights, and when both are consistent, traffic lights information is just consistent.
If judged result is consistent, pilotless automobile is controlled using the recognition result of the image of any camera collection;If inconsistent,
The traffic lights information stability of two camera collections is obtained respectively, the good traffic lights information of stability is then used to control automobile, its
Middle traffic lights information stability judges that change frequency is small, stablizes according to the traffic lights information change number of continuous multiple frames image
Property it is good, choose change frequency it is small traffic lights information control pilotless automobile, when change frequency is identical, stability is unanimously chosen
The image of short focus camera collection obtains traffic lights information.
If judge that focal length camera and short focus camera gather whether be same position traffic lights image result be it is no,
The distance selection of the traffic lights and automobile corresponding to traffic lights image gathered according to camera is wherein apart from the figure of small traffic lights
As obtaining traffic lights information control pilotless automobile.The image for the corresponding camera collection that numerical value is small i.e. in S1, S2 is as vapour
Car identifies the image of traffic lights, and its traffic lights result identified is used to control pilotless automobile.
The distance of object and camera in picture is calculated according to the picture that camera gathers, is using monocular camera range measurement principle
It can be achieved.Can be with camera and traffic lights as the distance of traffic lights and automobile corresponding to the traffic lights image in the present embodiment
Distance represents, and the distance of camera and traffic lights can realize according to camera location algorithm.Prior art is according to camera ranging
It is mature technology, is described briefly here.Traffic lights ranging according in the obtained image information of detection traffic lights laterally away from
From calculating.Camera installs fixation on vehicle, according to the installation site of camera and the focus information of camera, calculates the internal reference of camera
And external parameter information.Then can calculates the image pixel number that traffic lights horizontal direction takes in image and handed over true
The corresponding relation of logical lamp and vehicle distances.
Obviously present invention specific implementation is not subject to the restrictions described above, as long as employing the methodology and skill of the present invention
The improvement for the various unsubstantialities that art scheme is carried out, within protection scope of the present invention.
Claims (7)
1. a kind of pilotless automobile traffic lights recognition methods, including:
Camera gathers traffic lights image information;
Image information analysis is handled, turned right, keep straight on corresponding to traffic lights information;
It is characterized in that:When camera gathers traffic lights image information, camera standard time for exposure T and time for exposure flutter valve are preset
It is worth Δ T, in shooting image, the time for exposure of camera fluctuates for the standard exposure time fluctuates threshold values, as T- Δs T~T+
ΔT。
A kind of 2. pilotless automobile traffic lights recognition methods as claimed in claim 1, it is characterised in that:For gathering image
Camera be two, respectively focal length camera and short focus camera.
A kind of 3. pilotless automobile traffic lights recognition methods as claimed in claim 2, it is characterised in that:Described focal length phase
Machine and short focus camera work simultaneously gathers the image information of traffic lights.
A kind of 4. pilotless automobile traffic lights recognition methods as claimed in claim 2 or claim 3, it is characterised in that:Respectively to length
Burnt camera, the image of short focus camera collection carry out traffic signals identification, if only one camera identifies traffic lights information,
Pilotless automobile is then controlled with the transport information.
A kind of 5. pilotless automobile traffic lights recognition methods as claimed in claim 4, it is characterised in that:If focal length camera,
The image of both short focus cameras collection recognizes traffic lights information, and whether judge that focal length camera and short focus camera gather is phase
With the traffic lights image of position, if so, judge whether the traffic lights information of identification is consistent, if unanimously, using any recognition result
Control pilotless automobile;If inconsistent, the traffic lights information stability of two camera collections, wherein traffic lights are obtained respectively
Information stability judges according to the traffic lights information change number of continuous multiple frames image, chooses the small traffic lights letter of change frequency
Breath control pilotless automobile, when change frequency is identical, the image of short focus camera collection is chosen to obtain traffic lights information.
A kind of 6. pilotless automobile traffic lights recognition methods as claimed in claim 5, it is characterised in that:If judge focal length phase
Whether what machine and short focus camera gathered is that the result of image of traffic lights of same position is traffic lights that is no, being gathered according to camera
The distance selection of traffic lights and automobile corresponding to image wherein obtains the control of traffic lights information apart from the image of small traffic lights
Pilotless automobile.
A kind of 7. pilotless automobile traffic lights recognition methods as described in claim 5 or 6, it is characterised in that:According to collection
Traffic lights image corresponding to the distance of traffic lights and automobile come judge that focal length camera and short focus camera gather whether be identical
The image of the traffic lights of position, apart from identical, then traffic lights position corresponding to the traffic lights image gathered is identical.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710787414.7A CN107633213A (en) | 2017-09-04 | 2017-09-04 | A kind of pilotless automobile traffic lights recognition methods |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710787414.7A CN107633213A (en) | 2017-09-04 | 2017-09-04 | A kind of pilotless automobile traffic lights recognition methods |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107633213A true CN107633213A (en) | 2018-01-26 |
Family
ID=61100397
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710787414.7A Pending CN107633213A (en) | 2017-09-04 | 2017-09-04 | A kind of pilotless automobile traffic lights recognition methods |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107633213A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109559536A (en) * | 2018-12-10 | 2019-04-02 | 百度在线网络技术(北京)有限公司 | Traffic lights, traffic light recognition method, device, equipment and storage medium |
CN109635640A (en) * | 2018-10-31 | 2019-04-16 | 百度在线网络技术(北京)有限公司 | Traffic light recognition method, device, equipment and storage medium based on cloud |
CN110688992A (en) * | 2019-12-09 | 2020-01-14 | 中智行科技有限公司 | Traffic signal identification method and device, vehicle navigation equipment and unmanned vehicle |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103213540A (en) * | 2012-01-18 | 2013-07-24 | 富士重工业株式会社 | Vehicle driving environment recognition apparatus |
CN104506779A (en) * | 2014-12-24 | 2015-04-08 | 浙江宇视科技有限公司 | Traffic lamp color correcting method and image pickup equipment |
CN104766071A (en) * | 2015-04-28 | 2015-07-08 | 重庆邮电大学 | Rapid traffic light detection algorithm applied to pilotless automobile |
US9305223B1 (en) * | 2013-06-26 | 2016-04-05 | Google Inc. | Vision-based indicator signal detection using spatiotemporal filtering |
CN106394406A (en) * | 2015-07-29 | 2017-02-15 | 株式会社万都 | Camera device for vehicle |
-
2017
- 2017-09-04 CN CN201710787414.7A patent/CN107633213A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103213540A (en) * | 2012-01-18 | 2013-07-24 | 富士重工业株式会社 | Vehicle driving environment recognition apparatus |
US9305223B1 (en) * | 2013-06-26 | 2016-04-05 | Google Inc. | Vision-based indicator signal detection using spatiotemporal filtering |
CN104506779A (en) * | 2014-12-24 | 2015-04-08 | 浙江宇视科技有限公司 | Traffic lamp color correcting method and image pickup equipment |
CN104766071A (en) * | 2015-04-28 | 2015-07-08 | 重庆邮电大学 | Rapid traffic light detection algorithm applied to pilotless automobile |
CN106394406A (en) * | 2015-07-29 | 2017-02-15 | 株式会社万都 | Camera device for vehicle |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109635640A (en) * | 2018-10-31 | 2019-04-16 | 百度在线网络技术(北京)有限公司 | Traffic light recognition method, device, equipment and storage medium based on cloud |
CN109559536A (en) * | 2018-12-10 | 2019-04-02 | 百度在线网络技术(北京)有限公司 | Traffic lights, traffic light recognition method, device, equipment and storage medium |
CN110688992A (en) * | 2019-12-09 | 2020-01-14 | 中智行科技有限公司 | Traffic signal identification method and device, vehicle navigation equipment and unmanned vehicle |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6447729B2 (en) | Signal recognition device and signal recognition method | |
US9740942B2 (en) | Moving object location/attitude angle estimation device and moving object location/attitude angle estimation method | |
JP4914234B2 (en) | Leading vehicle detection device | |
JP5962850B2 (en) | Signal recognition device | |
CN106494395B (en) | A kind of automatic parking control method and device based on data fusion | |
CN108305504A (en) | A kind of parking stall air navigation aid and system | |
RU2703440C1 (en) | Method and device for controlling movement | |
US20180204077A1 (en) | Traffic Light Recognition Device and Traffic Light Recognition Method | |
CN107633213A (en) | A kind of pilotless automobile traffic lights recognition methods | |
US9683861B2 (en) | Estimated route presentation apparatus and estimated route presentation method | |
WO2007111220A1 (en) | Road division line detector | |
KR102176773B1 (en) | Parking system of vehicle | |
US11610317B2 (en) | Image processor and image processing method | |
JP2010006172A (en) | Auto lighting system | |
US20120249795A1 (en) | Signal recognizing device, signal recognizing method and signal recognizing program | |
WO2013035828A1 (en) | Device and method for predicting turning of vehicle | |
JP2020067698A (en) | Partition line detector and partition line detection method | |
JP2016038757A (en) | Traffic light recognition apparatus and traffic light recognition method | |
US20210180963A1 (en) | Onboard device | |
WO2023056720A1 (en) | Visual boundary detection-based robot control system and method | |
JP7068017B2 (en) | Vehicle travel path recognition device and travel control device | |
US11620908B2 (en) | Image processor and image processing method | |
JP3674316B2 (en) | Travel position determination device | |
WO2021094800A1 (en) | Traffic signal recognition method and traffic signal recognition device | |
JP5652374B2 (en) | Vehicle headlamp control device |
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: 20180126 |