Background technology
The arriving in epoch is transported with highway wisdomization, the concept of intelligent vehicle becomes increasingly popular, driver is directed to vehicle master
The function demand for moving safety becomes more and more important.But the vehicle on highway is travelled at present, must still be dependent on driver's substantial length of operation row
Enter.Although Ministry of Communications constantly advocates the idea of highway traffic safety, the accident rate of road traffic accident still remains high, shows
The improvement effect of traffic safety has reached bottleneck.
Statistical information according to Ministry of Communications points out, the accident main cause of road traffic accident, with fatigue driving, drunk driving,
Driver diverts one's attention, does not notice that the cases such as surrounding road conditions are large.Additionally, traffic safety steering committee of Ministry of Communications more enters one
Step is usually likely to become the situation of the accident factor for normality driving behavior analysis.Driver at any time due to inherent mood with it is external
Ambient influnence road conditions recognition capability, it is difficult to which each point of each second is all absorbed in and notices surrounding road conditions, causes each driver institute
The sequencing noticed differs, and easily omits crucial traffic information, and the traffic of fully display road needs significantly to be carried
Rise.
The research of intelligent transport system field is also the core of forthcoming generations research.Particularly in computer technology, communication
Under the background that technology is continued to develop, the development for promoting intelligent transportation system plays an important roll, and can have been given play in future
Comprehensive, large-scale effect, promotes traffic control system to develop towards accurate, real-time and efficient direction.
In recent years, computer vision application has been gradually increased in intelligent transportation system, and in intelligent transportation system
Concrete application can be divided into two aspects.One is roadside video monitoring system;The second is vehicle-mounted automated driving system.Before wherein
Person refers to above road or video camera is installed in roadside, and main effect is by information such as vehicle location, speed and types
It is transferred in intelligent transportation system;The characteristics of the latter is that video camera is moved with vehicle, mainly can be to vehicle body
Situations such as fatigue state of situation around and driver, exercises supervision and is transferred in the middle of system.
For the problem and Present study further investigated of stereoscopic vision identification, how by the common picture of twin-lens image recognition
Plain feature simultaneously obtains feature depth of view information, will be the Main Bottleneck of stereoscopic vision recognizer process.Current development is explored to show
Condition, the applicable scope of stereoscopic vision extends to each field arround life extensively, but to all of feature in image is entered
Row stereoscopic vision recognizes that one will bring numerous and diverse unnecessary extra algorithm data, and two long algorithm times will be difficult to
Reach the system requirements of real time implementation visual identity and Service controll.
Therefore, for above-mentioned technical problem, it is necessary to provide a kind of DAS (Driver Assistant System) based on stereoscopic vision target and
Method.
The content of the invention
In view of this, it is an object of the invention to provide a kind of DAS (Driver Assistant System) based on stereoscopic vision target and side
Method.
To achieve these goals, technical scheme provided in an embodiment of the present invention is as follows:
A kind of DAS (Driver Assistant System) based on stereoscopic vision target, the DAS (Driver Assistant System) includes:
Image acquisition units, including located at 2 camera lenses of automobile upper position, the road conditions scene graph in front of driving is gathered respectively
Picture;
Feature Conversion unit, spatial coordinated information is converted to by the validity feature of road conditions scene image after collection;
Feature identification unit, the direction board for recognizing the mark graticule on road surface and front.
As a further improvement on the present invention, 2 optical center height h are 1.5 meters, 2 in described image collecting unit
Camera lens spacing b is 0.5 meter.
Technical scheme provided in an embodiment of the present invention is as follows:
A kind of auxiliary driving method based on stereoscopic vision target, methods described includes:
S1, by 2 camera lenses of automobile upper position gather respectively driving in front of road conditions scene image;
S2, the validity feature of road conditions scene image after collection is converted into spatial coordinated information;
The direction board of mark graticule and front on S3, identification road surface, the visual identity of auxiliary driver's reinforcing front road conditions.
As a further improvement on the present invention, the step S1 also includes:
Local edge cutting is carried out to road conditions scene image, so that level orientation and vertical orientations are all identical.
As a further improvement on the present invention, the step S2 includes:
By feature by three-dimensional coordinate system system conversion to two-dimensional pixel coordinate system;And/or
Known image common trait is tried to achieve into the depth of field by two-dimensional pixel coordinate system, is further changed to three-dimensional coordinate system
System.
As a further improvement on the present invention, the validity feature in the step S2 includes double Huang solid lines, right side red line, a left side
The pixel characteristic of side red line.
As a further improvement on the present invention, the step S3 is specially:
S31, with MedianFilter matrix disposals 10 times, then subtract each other with artwork and try to achieve image;
S32, Sobel filter are first with matrix SxTreatment filters the shade that road surface is likely to occur, then with matrix SxyTry to achieve second order side
Edge feature distribution image;
S33, the radio-frequency component in congeniality region is filtered;
S34, according to image procossing experience choose binary-state threshold;
S35, binary image are successively changed by artwork and negative film, are processed with labelization respectively, are set according to noise label size
High-pass filter threshold value, filters the noise and hole of small size;
Block after S36, noise filtering is processed with labeling again, according to the size of road surface region label, selectes the logical upper limit of filter
With lower threshold, road surface characteristic is leached;
, in the position of the plane of delineation, road surface region is seated image lower section for S37, foundation driving front position.
As a further improvement on the present invention, the binary-state threshold in the step S34 is 80% to 95% GTG point position.
As a further improvement on the present invention, the step S37 also includes:
Upper area is filtered, and gathers lower images region as ROI interest region.
The beneficial effects of the invention are as follows:
The present invention reads behavior and reappears to the process and visual information of driver's identification front side road conditions, by means of solid
Visual pattern carrys out the algorithm process method to road ahead scene Recognition of driving a vehicle, and simulates and reappear the vision identification processing of driver
Pattern, under different climatic environments, can recognize all kinds of direction boards of the mark graticule and front on road surface, aid in driver
The visual identity of reinforcing front road conditions.
Specific embodiment
In order that those skilled in the art more fully understand the technical scheme in the present invention, below in conjunction with of the invention real
The accompanying drawing in example is applied, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described implementation
Example is only a part of embodiment of the invention, rather than whole embodiments.Based on the embodiment in the present invention, this area is common
The every other embodiment that technical staff is obtained under the premise of creative work is not made, should all belong to protection of the present invention
Scope.
Shown in ginseng Fig. 1, the invention discloses a kind of DAS (Driver Assistant System) based on stereoscopic vision target, including:
Image acquisition units 10, including located at 2 camera lenses of automobile upper position, the road conditions scene graph in front of driving is gathered respectively
Picture;
Feature Conversion unit 20, spatial coordinated information is converted to by the validity feature of road conditions scene image after collection;
Feature identification unit 30, the direction board for recognizing the mark graticule on road surface and front.
Stereoscopic vision of the present invention is based on parallel camera lens, shown in ginseng Fig. 2, installed in 2 mirrors of automobile upper position
First 11 visions for replacing true people, road scene at the present invention selected, and pasted with constant spacing at camera lens road surface ahead
Green adhesive tape, with the position of label space coordinate, then with parallel twin-lens framework collection driving forward image.It is situated between by the present invention
The validity feature recognition strategy and depth information process for continuing, are changed to depth information, and estimate with scene by the reversion of image validity feature
The spatial coordinate location of survey carries out error comparison, assists assessment simulation effect.
Preferably, h=1.5 meters of optical center height, b=0.5 meters of twin-lens spacing, focal length f, coordinate direction in the present invention
In accordance with the right-hand rule of Outer Product of Vectors, with pin-hole imaging principle construction coordinate system.Driving front position, it is intended that relative to vehicle
The direction of straight trip track is in parallel orientation with the heave direction on road surface, as shown in Figure 2.Try to achieve driving front position and correspond to two
The pixel coordinate position of side mirror head, is beneficial to retain image lower section reservation ROI region, and provide the base of origin coordinate system transform
It is accurate.
Shown in ginseng Fig. 3, the invention also discloses a kind of auxiliary driving method based on stereoscopic vision target, the method bag
Include:
S1, by 2 camera lenses of automobile upper position gather respectively driving in front of road conditions scene image;
S2, the validity feature of road conditions scene image after collection is converted into spatial coordinated information;
The direction board of mark graticule and front on S3, identification road surface, the visual identity of auxiliary driver's reinforcing front road conditions.
The present invention does not consider the error that lens distortions are produced under the conditions of preferable projection imaging, can directly with similar three
Angular relation, also can be total to for known image to two-dimensional pixel coordinate system by three-dimensional coordinate system system conversion by feature
The depth of field is tried to achieve by two-dimensional pixel coordinate system with feature, further conversion to three-dimensional coordinate system is united.But in both coordinates
In the middle of linear transformation process between system, it is still necessary to correspond to the position of pixel coordinate by known space coordinates directional correction
Put, as the benchmark of origin coordinate system transform.
Except the benchmark of origin coordinate system transform, driving front position the position of pixel can also provide visual identity road surface with
In the image processing process of traffic lane line, in selection space coordinates, the interest region ROI treatment below horizontal plane.In the present invention
In the road scene of discussion, it has been determined that road surface is all located at below the horizontal line of space coordinates with traffic lane line, if therefore only considering
Image-region below treatment horizontal line, just it is advantageously ensured that the recognition effect of road surface and traffic lane line feature, it is to avoid effectively special
Levy region and the unnecessary process problem such as obscure and occur.Such as sky belongs to the relatively low region of second order radio-frequency component together with road surface, profit
ROI treatment is chosen with driving front position, sky can be avoided to be considered as a part for road surface characteristic.
The present invention gathers road scene according to analog architectures, to make twin-lens gather the level orientation and vertical orientations of image
All identical, image is carried out local edge cutting by the present invention, and condition is the same from vertical orientations to reach level orientation, therefore
The image length and width size that this Simulation identification is used is (736*592).System by parallel twin-lens acquired image, first
According to foregoing feature recognition strategy, belong to scene mode on daytime, by identification traffic lane line recognition strategy above, by pixel
Form switchs to HSV colour systems by RGB, and with selected HSV colour system threshold values, by central double Huang solid lines and both sides red graticule feature
Leach, as shown in Fig. 4 a, 4b.
In order to compare difference of the coordinate distribution results compared to actual estimation position that the feature depth of field is calculated, the present invention exists
Green adhesive tape is pasted by the traffic lane line on road surface as mark, is initial origin by lens shooting position, with front position of driving a vehicle
It is bearing of trend, every 5 meters of fixation is used as mark spacing.After parallel twin-lens gathers image, marked using image reading green
The v coordinate that note is located, such as Fig. 4 c, 4d, then the mark position that will be pasted on road surface, the v for corresponding to pixel by depth of field z coordinate sit
Mark is recorded one by one.Driving front position is seated the pixel coordinate position of both sides lens image, is also the origin of coordinate system.
In analog architectures of the invention, FOE is set to known coordinate in advance, and is locked in subsequent simulation treatment quiet
State image attributes, therefore the processing procedure inquired into simulation framework will be adopted based on still image, from CANON
IXUS960IS Digital Camera are gathered as still image and originated.Wherein road surface characteristic recognition strategy and algorithm flow be such as
Shown in Fig. 5, idiographic flow step is as follows:
S31, according to high light shade, with the uneven problem of adaptability brightness uniformity mechanism treatment scene illumination.First with
MedianFilter matrixes [5 × 5] are processed 10 times, then are subtracted each other with artwork and tried to achieve image;
S32, Sobel filter are first with matrix SxTreatment filters the shade that road surface is likely to occur, then with matrix SxyTry to achieve second order side
Edge feature distribution image;
S33, it is to retain distant place road surface validity feature, is first processed 5 times with Median Filter matrix sizes [3 × 3], then with square
Battle array [5 × 5] is processed 5 times, and the radio-frequency component in congeniality region is filtered, and is beneficial to follow-up binary-state threshold selection;
S34, according to image procossing experience, choose 80% to 95% GTG point position as binary-state threshold;
S35, binary image are successively changed by artwork and negative film, are processed with labelization respectively, are set according to noise label size
Size Filter high-pass filter threshold values, filter the noise and hole of small size;
Block after S36, noise filtering is processed with labeling again, according to the size of road surface region label, selectes Size
The logical upper limit of Filter filters and lower threshold value, you can leach road surface characteristic;
S37, depending on result, according to driving front position in the position of the plane of delineation, road surface region is seated under image
Side, to avoid top low-frequency image region from being leached in the lump, can filter upper area, and gather lower section figure in processing procedure
As region is used as ROI interest region.
The inner parameter after correction is tried to achieve after feature recognition, you can according to pin-hole imaging principle, with similar triangles
Geometry derivation concept, Projection Character is changed in pixel coordinate known to the image of both sides with by matrix equality.Feature passes through square
Battle array conversion will be presented conversion coordinate and be illustrated with emulation to the simulation result of space coordinates distribution with form and map file.It is sequentially only
The vertical double Huang solid lines of display, right side red line, the pixel characteristic identification respective coordinates of left side red line, and each for the spy of scene estimation
Levying coordinate position carries out error comparison analysis.
According to simulation result, analysis image feature recognition position is preferable compared to scene estimation feature in space coordinates
Offset error between position.The distance between ideal position error is estimated as characteristics of image identification position and scene using d, such as
Lower expression:
Stereoscopic vision recognition result according to double yellow solid line features is analyzed.Due to depth of field coordinate 80 and 85 meters of level
Line and without double yellow solid lines, thus fails to obtain clear and definite characteristic coordinates information.Feature at 75 meters of depth of field coordinate, due to position
End in double yellow solid lines, shape it is slightly biased it is irregular influenceed, add the not good reason of distant place pixel resolution, cause bimirror
The common trait that head image is gathered, the precision of horizontal offset (Disparity) collection is not enough, the space changed by algorithm
Coordinate position also thus substantially distortion.And the closely feature at 15 meters of coordinate, because the image range that right side camera lens catches has
Limit, only collects local double Huang solid lines, therefore cannot try to achieve real double Huang solid line centers.
Height is 0.7 meter to camera lens from the ground in the present embodiment, the left and right sides image difference after validity feature identifying processing
Such as Fig. 6 a, 6b.The validity feature that finally 4 horizontal lines are searched, calculates the space coordinates of double Huang solid lines and right side red line respectively
Information, collection horizontal line searches the depth information z coordinate of two point features, and tries to achieve average value.The top of the present embodiment
Horizontal line, due to right side, both sides image can not search validity feature and cannot try to achieve its depth information, therefore with double Huangs
The depth information of solid line is presented.
Daytime and preferable weather scene mode, identification road surface and the robustness of traffic lane line, can reach required mark
It is accurate.
By above-mentioned technical proposal as can be seen that the invention has the advantages that:
The present invention reads behavior and reappears to the process and visual information of driver's identification front side road conditions, by means of solid
Visual pattern carrys out the algorithm process method to road ahead scene Recognition of driving a vehicle, and simulates and reappear the vision identification processing of driver
Pattern, under different climatic environments, can recognize all kinds of direction boards of the mark graticule and front on road surface, aid in driver
The visual identity of reinforcing front road conditions.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie
In the case of without departing substantially from spirit or essential attributes of the invention, the present invention can be in other specific forms realized.Therefore, no matter
From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power
Profit requires to be limited rather than described above, it is intended that all in the implication and scope of the equivalency of claim by falling
Change is included in the present invention.Any reference in claim should not be considered as the claim involved by limitation.
Moreover, it will be appreciated that although the present specification is described in terms of embodiments, not each implementation method is only wrapped
Containing an independent technical scheme, this narrating mode of specification is only that for clarity, those skilled in the art should
Specification an as entirety, the technical scheme in each embodiment can also be formed into those skilled in the art through appropriately combined
May be appreciated other embodiment.