CN109284664A - Driver assistance system and guardrail detection method - Google Patents
Driver assistance system and guardrail detection method Download PDFInfo
- Publication number
- CN109284664A CN109284664A CN201810787842.4A CN201810787842A CN109284664A CN 109284664 A CN109284664 A CN 109284664A CN 201810787842 A CN201810787842 A CN 201810787842A CN 109284664 A CN109284664 A CN 109284664A
- Authority
- CN
- China
- Prior art keywords
- guard bar
- bar structure
- vertical component
- guardrail
- information
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title description 8
- 238000000034 method Methods 0.000 claims abstract description 62
- 238000005520 cutting process Methods 0.000 claims abstract description 18
- 238000010801 machine learning Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000003892 spreading Methods 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 6
- 238000012706 support-vector machine Methods 0.000 claims description 6
- 230000000712 assembly Effects 0.000 claims description 4
- 238000000429 assembly Methods 0.000 claims description 4
- 238000013480 data collection Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000003447 ipsilateral effect Effects 0.000 claims description 3
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 claims description 2
- 230000004888 barrier function Effects 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 claims 1
- 230000000694 effects Effects 0.000 claims 1
- 238000009826 distribution Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/16—Anti-collision systems
- G08G1/165—Anti-collision systems for passive traffic, e.g. including static obstacles, trees
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30256—Lane; Road marking
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to the computer implemented methods for generating guardrail pattern comprising: a. captures a picture frame with a video camera；B. it determines that described image frame includes at least two guard bar structures of guardrail, and utilizes the guard bar structure in cutting box tracking described image frame, wherein the guard bar structure includes non-vertical component and vertical component；C. the information of cutting box is handled, to extract the angle gradient information of one of tracked guard bar structure；D. it is clustered using the diagonal gradient information of clustering method, to generate the angle information cluster of each section in relation to cutting box；E. the information on the non-vertical component of cutting box is extracted；F. step c to e is repeated, until the guard bar structure of all tracking is completed these steps；And g. is fitted a curve by the non-vertical component of tracked guard bar structure.The driver assistance system that the invention further relates to a kind of for implementing the method and a kind of for detecting the training machine learning algorithm of guard bar structure.
Description
Technical field
The present invention relates to machine learning fields, specifically, the present invention relates to a kind of method for detecting guardrail, it is a kind of
For implementing the driver assistance system of the method and a kind of for detecting the training machine learning algorithm of guard bar structure.
Background technique
Detecting roadside guardrail has particularly important meaning to increasingly automated drive.These structures (i.e. guardrail) are throughout generation
The road of various countries of boundary.They roadside exist for vehicle lateral movement provide the physical limit of free space.Therefore, guardrail
It detects extremely important.This problem (i.e. guardrail detection) can be handled by a video camera (such as B/W camera).
Summary of the invention
The present invention relates to a kind of for generating the computer implemented method of guardrail pattern, including:
A. a picture frame is captured with a video camera；
B. it determines that picture frame includes at least two guard bar structures of a guardrail, and utilizes the shield in cutting box tracking picture frame
Column structure, wherein guard bar structure includes non-perpendicular (i.e. "vertical") part (NVP) and vertical component (VP)；
C. the information of cutting box is handled, to extract the angle gradient information of one of tracked guard bar structure；
D. it is clustered using the diagonal gradient information of clustering method, to generate the letter of the angle of each section in relation to cutting box
Cease cluster；
E. the information on the non-vertical component (NVP) of cutting box is extracted；
F. step c to e is repeated, until the guard bar structure of all tracking all experienced these steps；And
G. a curve is fitted by the non-vertical component (NVP) of tracked guard bar structure.
Angle gradient information can be by providing angle gradient descriptor, and the direction gradient for preferably providing at least two guard bar structures is straight
Side's figure (HoG) descriptor extracts, and the descriptor is used to provide the vertical component and non-vertical component of guardrail single structure
Gradient line and angular orientation information about non-vertical component relative to object of reference.
Angle gradient information can by implementation-facing view distributed type assemblies parallel calculating method (View-Oriented,
Distributed, Cluster-Based Approach to Parallel Computing, VODCA) it extracts.
The advantages of this method is the distribution of the reusable facing view for extracting histograms of oriented gradients (HoG) feature
Part cluster parallel computing method (VODCA), to mitigate the burden of computing hardware, and reusable point from facing view
The classifier part of cloth cluster parallel computing method (VODCA) frame.
Step g may include being inserted into a curve by the guard bar structure detected, to obtain the full curve of one guardrail of description.
The curve can be a multinomial of the gradient line intersection point of the non-vertical component (NVP) across at least two guard bar structures
Curve.
The step d of the above method can provide an average value mu and a corresponding spreading coefficient for each non-vertical component
σ；And spreading coefficient is used to Optimal Curve fitting.
In addition, the step d of the above method may also include removal wrong report guard bar structure.
It can determine that picture frame includes two guard bar structures on a road difference side.Therefore, two curves are produced.
Therefore, this method can detect multiple guardrails in a frame with particular advantage, i.e. it.To detected guardrail knot
The tracking of structure is performed with each suitable track algorithm or tracking, especially with a Kalman filtering.
Determine that described image frame includes two guard bar structures on road is not ipsilateral, it is contemplated that the angle of guard bar structure is believed
Breath.
In addition, the above method may also include the function of sounding a warning when driver is close to guardrail to it.
Therefore, this method has other advantage, that is, the method announced can lane auxiliary deviation system or/and radar
(accurately) road boundary is determined, because guardrail is always arranged along road boundary, vice versa.
The invention further relates to a kind of for detecting the driver assistance system of guard bar structure, which includes:
A. video camera, it can be used for capturing multiple images；
B. processor, it is for effectively connecting video camera, wherein processor can be used for implementing defined in the above method
Operation.
In addition, driver assistance system may also include radar, and processor is used for the information life by being provided by radar
At a guard barrier curve.
The invention further relates to a kind of vehicles including above-mentioned driver assistance system.
The vehicle can be automobile, motorcycle or truck.
The computer-implemented machine for the method that it includes guard bar structure that the invention further relates to a kind of for providing determining picture frame
Device learning method, this method comprises:
A. the flag data collection of at least one picture frame of guard bar structure is provided in a database, and as study mould
At least one machine of type；
B. at least one picture frame is handled to extract the angle gradient information of the guard bar structure of guardrail；
C. using angle gradient information training learning model, to complete point to the picture frame in picture frame including guard bar structure
Class.
Learning model can be any machine learning mode, method or model, and an e.g. support vector machines (SVM) or one is certainly
It adapts to enhancing machine (AdaBoost).
The step b of computer-implemented machine learning method may include the distributed type assemblies parallel computation for implementing facing view
Method (VODCA).
The computer system including memory and processor that the invention further relates to a kind of, wherein the processor can be used for
Execute the operation of above-mentioned computer-implemented machine learning method.
Detailed description of the invention
Fig. 1 illustrates the repeat pattern in a guardrail.
The single structure of Fig. 2 displaying guardrail.
The gradient of Fig. 3 displaying guardrail single structure.
Fig. 4 shows the feature description of histograms of oriented gradients (HoG) descriptor of a guard bar structure.
Fig. 5 shows the training flow chart of the machine learning algorithm for the guard bar structure that detects and classify.
Fig. 6 shows angle information contained in a cutting box.
The gauss hybrid models Density Distribution of angle information shown in Fig. 7 schematic thinking 6.
Fig. 8 is shown to the fitting of a polynomial on detected guard bar structure.
Fig. 9 is related to the flow chart of several steps method, detects guardrail knot with machine learning method (S1 to S5) including (1)
Structure, and (2) generate guardrail pattern from detected guard bar structure.
Figure 10 describes a computer system of the invention.
Specific embodiment
It reference will now be made in detail to the patent document that example is graphic in the accompanying drawings, is announced now, wherein identical attached drawing mark
Note always shows identical device.The various viewpoints of this publication are described below, to be explained with reference to the drawings wherein meaning.
Before explaining in detail patent document, it should be understood that described in application of the invention is not limited to be described below
Or part shown in the drawings is designed and arranged details.The present invention can have other embodiments or be implemented in various ways or hold
Row.Moreover, it will also be appreciated that wording as used herein and term are that for the purpose of description, rather than limitation should be by
It is considered restrictive.Indefinite article " one " and " a certain " and definite article "the" equally also refer to corresponding plural number.Title is only used
In more fully understanding the present invention, and it is not necessarily to be construed as mutually distinguishing the content of each chapters and sections.
The detection of guard bar structure
Guardrail 101 is the linear structure for being present in highway and other 102 roadsides of road.
Guard bar structure is made of the repeat pattern for being shown as T1, T2......T6 in Fig. 1.It is important that, it is understood that herein
N number of this kind of duplicate pattern can be found by video camera (such as a B/W camera).
The repetition guardrail pattern for being referred to herein as guard bar structure (T1 to the T6 of example as shown in figure 1) can consist of two parts, i.e.,
Vertical component (VP) shown in Fig. 2 and non-vertical component (NVP).
The digital picture of guard bar structure can be handled, with image is provided angle gradient or similar characteristics descriptor (such as Fig. 3 institute
Show).These descriptors can be used for detecting this class formation.For example, Fig. 4 show with guard bar structure identical in Fig. 2 and Fig. 3, but
It is with a histograms of oriented gradients (HoG) image display descriptor of guard bar structure.It can be used gradient descriptor in angle (such as square
To histogram of gradients (HoG), edge orientation histogram, Scale invariant features transform descriptor and Shape context), because by angle
The angle information that gradient descriptor provides has carried out detailed description to target.
For detecting the training machine learning algorithm of guard bar structure
It can be used for providing the training machine study of detection guard bar structure for the above-mentioned angle gradient descriptor that guard bar structure provides
Algorithm.Housebroken machine learning model obtained can detect and classify guardrail and/or guard bar structure in image.
The method of the machine learning algorithm of one detection guard bar structure of training is illustrated in Fig. 5.This method can by computer-implemented,
And it can be by any suitable computer system such as personal computer, laptop, mainframe computer or including at least
Any other system of one processor and a memory executes.Training method and system can be not belonging to comprising in the car
Driver assistance system a part, therefore, can " offline " execute training, that is, method or system can be auxiliary not as driver
Aid method or system work." online " refers to the real-time operation of driver assistance method or system, that is, this method or system attempt
Guard bar structure and/or guardrail are detected during driving.
This method can include the following steps: one including, for example, guardrail and/or the exemplary database of guard bar structure ---
I.e. at least one includes digital picture --- middle offer data set, the step S501 of known guard bar structure.Database may include shield
The individual digit image of column structure or the image frame sequence (such as video) for describing guard bar structure.Can flag data collection, with display
The guard bar structure different with display of guard bar structure described in digital picture.
Data set extracts angle Gradient Features, step S502 in the case where can be used for stating angle gradient descriptor in use.Angle
Gradient descriptor can be input into an identifying system based on supervised learning.For example, identifying system can be such as support vector machines
(SVM) or adaptively enhance the learning model or classifier of machine (AdaBoost) etc..The output of this method is housebroken classification
Device, step S503, it can provide the inventory of contained guard bar structure in picture frame.This method also may include a tracking step, should be with
The detected structure of track step tracking, which can simultaneously be provided, defines frame or cutting box around guard bar structure.
Driver assistance method and system
1. detecting the method and system of guardrail
This housebroken model or classifier can be used for detecting the guard bar structure in any digital picture with online mode.
When the housebroken model of application on site or classifier, at least two guard bar structures tracked can be detected.If desired, can
The use of clustering algorithm by these structure verifications is real guard bar structure, and thus eliminates identified wrong report guard bar structure.This
Outside, clustering algorithm can be used to determine the angular distribution of categorized guard bar structure.
This method or system especially may also include one for remove wrong report guard bar structure the step of.
It is as follows:
Wrong report can be eliminated by application clustering algorithm.Clustering algorithm can be gauss hybrid models, and (such as one-dimensional Gauss is mixed
Molding type (GMM-1D)) or a K average algorithm.
The step can be implemented in the following manner:
To eliminate wrong report, credibility check can be used.This can be realized by using following methods step.One tracker
It can be used for tracking structure detected.It is contemplated that guard bar structure (for further processing).Tracking length may more than one.With
Track length is related to the example quantity for being directed to a given tracking target detected.Under present circumstances, these targets are guardrails
Potential guard bar structure, i.e. target not yet passes through credibility check.
From the classifier of housebroken machine learning method, it can get and define frame comprising real guard bar structure.Then,
Direction gradient information can be extracted from the edge graph in that region defined in frame (similar to frame T1 to the T6 of Fig. 1；Fig. 6 is shown
Define the content of frame).
Since guard bar structure includes non-vertical component (NVP) and vertical component (VP) (referring to fig. 2 and Fig. 6), such as scheme
Shown in 6, we can obtain the zero degree of vertical component (VP) in direction gradient description or close to the edge of zero degree.Here, non-hang down
Straight part (NVP) edge component can facilitate non-zero angle gradient information.
The direction gradient value of information can be then transferred to clustering algorithm as defined above.Two dimension angular information can as a result,
It is converted into a line principal matrix or any for obtaining the proper method of one-dimensional data.It can be further point from clustering algorithm
Analysis selection such as gauss hybrid models (GMM), distribution and the m gaussian kernel function with maximum probability.One of Gaussian Profile
Vertical component (VP) can be corresponded to, and another will be referring to non-vertical component (NVP).
For example, determining m in a manner of threshold value T is arranged, and it is contemplated that all Gaussian Profiles higher than predefined thresholds T.
To demonstration section shown in Fig. 6, gauss hybrid models (GMM) the cluster output of n=4 is illustrated in Fig. 7
(quantity that n is gaussian kernel function).
Table 1
| Curve numberings, c | Average value, μ | Spreading coefficient, σ |
| 1 | -110.76 | 35.60 |
| 2 | -19.16 | 20.01 |
| 3 | 2.46 | 5.08 |
| 4 | 94.54 | 39.51 |
Table 1 shows the average value and spreading coefficient of guard bar structure as shown in Figure 7.It was noticed that table 1 is only used for figure
Solution and the explanation present invention.It can be seen that from Fig. 7 and corresponding table 1, gradient has the part centered on average value (also referred to as
Kernel) and by Sigma indicate spreading coefficient.It includes two high density peak values that window, which can be observed, i.e., respectively at 2.46 degree
With the c3 and c2 at -19.16 degree.
In addition, there is overlapping between these parts, therefore a part can be treated them as.Other two part, that is, c1 and
C4 indicates a pair of of line in [- 180,180] circular representation.This two parts belongs to guard bar structure, their angle can be determined.
For example, by checking gauss hybrid models (GMM) distribution a credibility check can be implemented, herein to refuse to report by mistake
Guard bar structure/define frame.Kernel ' one of average value (not all) is necessary for 0 ° of ± Δ (wherein Δ is predetermined threshold).Otherwise,
The frame detected can be rejected.
Identical operation (tracking by cutting box) can be executed on each guard bar structure detected.However, when executing
When the method for the invention, any suitable tracking (such as Kalman filtering) can be used.From these guard bar structures, we
The angle information of non-vertical component (NVP), i.e. average value mu can be extracted as described above.These average values may be relevant, because
Such as subsequent explanation, non-vertical component (NVP) helps to realize the continuity between guard bar structure.
2. generating guardrail pattern from the guard bar structure detected
In order to determine guardrail pattern, it would be desirable to which drawing a curve by series of points, (i.e. each guard bar structure is
This purpose is plus/one point of contribution).The intersection point that the point can be used as non-vertical component in detection block (NVP) gradient line obtains.
Then when considering the angular range of structure detected, especially non-vertical component (NVP), pass through guardrail knot
Structure is fitted (such as multinomial) curve.
It may be selected to execute a continuity check.As described above, angle information (i.e. μ) can be obtained with spreading coefficient (i.e. σ).This
It can be in the case where considering about the information state of the angle (in σ) of each guard bar structure non-vertical component (NVP) to drawing/describe
Line is adjusted.Thus, it is possible to provide a sweep 801, the line can for example be shown in the image comprising road 802 (see Fig. 8)
Or it is used for other application.
Curve is drawn in the case where taking the angle of non-vertical component (NVP) into account ensures that proposed algorithm is including road
There is firm performance in the case of the guardrail of road two sides.In the case, driver assistance system or method can detect two sides
Guard bar structure.In such cases, the method and system announced causes the guard bar structure in road left and right side to have not
Same non-vertical component (NVP) numerical value.One of as a result, the method and system announced allows in a given frame simultaneously
Two guardrails of detection or/and drafting.
Fig. 9 illustrates the step summary of announced driver assistance system and method.
In step S901, frame (i.e. picture frame or video frame) is entered.By housebroken machine learning algorithm or method
Applied on frame to detect guard bar structure, step S902.Then, by any suitable tracking (such as Kalman filtering) with
The guard bar structure and offer (extraction) cutting box of track guardrail are for further processing.According to the information acquisition for including in cutting box
The angle information of structure in cutting box, including the angle information about non-vertical component (NVP), step S903.Application may be selected
Cluster, to eliminate wrong report, step S904.It obtains and the information in relation to non-vertical component (NVP) angle, step S905 is provided.?
It, can be (such as multinomial by each corresponding guard bar structure fitting one in the case of taking acquired non-vertical component (NVP) angle information into account
Formula) curve, step S906.
Figure 10 describes one for executing the computer system 1000 for the method for generating guardrail pattern of the present invention.It should
Computer system includes video camera 1002, and the video camera is for capturing the frame for staying in and handling in processor 1001.As option,
System 1000 includes detecting and the storage device 1003 of handled data for storing.
Computer-implemented machine learning method of the present invention can execute in same system, but in the case, take the photograph
Camera 1002 is option.
Computer system of the present invention can be personal computer, specific integrated circuit (ASIC) or field-programmable gate array
It arranges (FPGA).
Reference signs list
101 guardrails
T1-T6 guard bar structure
102 roads
S501 provides flag data collection for guard bar structure
S502 extracts angle Gradient Features
S503 classifying step
801 draw curve
802 roads
S901 input frame
S902 is using the machine learning algorithm for detecting guard bar structure
S903 extracts the angle information of shearing
S904 application cluster
The angle information of S905 extraction non-vertical component
S906 executes (multinomial) fitting while considering the angle information of non-vertical component
1000 computer systems
1001 processors
1002 video cameras
1003 storage devices
Claims (19)
1. the computer implemented method for generating guardrail pattern comprising:
A. a picture frame is captured with a video camera；
B. determine that described image frame includes at least two guard bar structures of guardrail, and using in cutting box tracking described image frame
Guard bar structure, wherein the guard bar structure includes non-vertical component (NVP) and vertical component (VP)；
C. the information of cutting box is handled, to extract the angle gradient information of one of tracked guard bar structure；
D. it is clustered using the diagonal gradient information of clustering method, to generate the angle information cluster of each section in relation to cutting box；
E. the information on the non-vertical component (NVP) of cutting box is extracted；
F. step c to e is repeated, until the guard bar structure of all tracking is completed these steps；And
G. a curve is fitted by the non-vertical component (NVP) of tracked guard bar structure.
2. according to the method described in claim 1, wherein, angle gradient information is by providing the angle of at least two guard bar structure
Gradient descriptor, preferred orientations gradient descriptors histogram extract, which provides the vertical component effect of guardrail single structure
Divide the angular orientation information with the gradient line of non-vertical component and about non-vertical component relative to object of reference.
3. the method according to any claim in the claims, wherein angle gradient information be by implement towards
What the distributed type assemblies parallel calculating method (VODCA) of view extracted.
4. the method according to any claim in the claims, wherein step g includes the guardrail by detecting
One curve of structure interpolation, to obtain the full curve of one guardrail of description.
5. the method according to any claim in the claims, wherein the curve is across described at least two
The polynomial curve of the gradient line intersection point of the non-vertical component (NVP) of a guard bar structure.
6. the method according to any claim in the claims, wherein step d provides for each non-vertical component
One average value mu and corresponding spreading coefficient σ；And carry out the fitting of Optimal Curve using spreading coefficient.
7. the method according to any claim in the claims, wherein step d further includes removal wrong report guardrail knot
Structure.
8. the method according to any claim in the claims, wherein to detected guard bar structure with
Track is to be implemented with any suitable track algorithm or method especially with Kalman filtering algorithm.
9. the method according to any claim in the claims, wherein determine that picture frame includes that road is not ipsilateral
Two guard bar structures and/or generate two curves.
10. according to the method described in claim 9, wherein, by considering the angle information of guard bar structure, determining that picture frame includes
Two guard bar structures on road is not ipsilateral.
It further include when driver is close to guardrail to driving 11. the method according to any claim in the claims
The person of sailing is alerted.
12. a kind of driver assistance system (1000) for detecting guard bar structure comprising:
A. video camera (1002), for capturing multiple images；
B. processor (1001), for camera operation connecting, wherein the processor is wanted for implementing in aforesaid right
Operation defined in asking.
13. driver assistance system according to claim 12 further includes radar, and processor is used for by by radar
The information of offer generates guard barrier curve.
14. a kind of vehicle including driver assistance system described in claim 12 or 13.
15. vehicle according to claim 14, wherein the vehicle is automobile, motorcycle or truck.
16. for providing the computer-implemented machine learning method for the method that determining picture frame includes guard bar structure, the machine
Learning method includes:
A. in the database provide guard bar structure at least one picture frame flag data collection and as learning model extremely
A few machine；
B. at least one described picture frame is handled to extract the guard bar structure angle gradient information of guardrail；
C. using angle gradient information training learning model, to complete to the image frame classification in picture frame including guard bar structure.
17. according to the method for claim 16, wherein the learning model is support vector machines (SVM) or adaptive increasing
Strong machine (AdaBoost) or any suitable machine learning method.
18. method according to claim 16 or 17, wherein step b include implement facing view distributed type assemblies simultaneously
Row calculation method (VODCA).
19. a kind of computer system (1000) comprising memory (1003) and processor (1001), wherein the processor
For implementing operation defined in any claim in claim 1 to 11 or 16 to 18.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102017212418.0 | 2017-07-20 | ||
| DE102017212418.0A DE102017212418A1 (en) | 2017-07-20 | 2017-07-20 | DRIVER ASSISTANCE SYSTEM AND METHOD FOR GUIDANCE PLANK RECOGNITION |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109284664A true CN109284664A (en) | 2019-01-29 |
| CN109284664B CN109284664B (en) | 2023-10-27 |
Family
ID=64951408
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810787842.4A Active CN109284664B (en) | 2017-07-20 | 2018-07-18 | Driver assistance system and guardrail detection method |
Country Status (3)
| Country | Link |
|---|---|
| JP (1) | JP7164342B2 (en) |
| CN (1) | CN109284664B (en) |
| DE (1) | DE102017212418A1 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111241900A (en) * | 2019-04-12 | 2020-06-05 | 泰州阿法光电科技有限公司 | Traffic environment field maintenance method |
| CN112446314A (en) * | 2020-11-19 | 2021-03-05 | 武汉中海庭数据技术有限公司 | Method and system for extracting guardrail elevation based on projection drawing |
| CN117090133A (en) * | 2023-08-23 | 2023-11-21 | 青岛迪乐普精密机械有限公司 | Guardrail and detection method thereof |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110068321B (en) * | 2019-03-27 | 2020-04-24 | 清华大学 | UAV relative pose estimation method of fixed-point landing sign |
| KR102628027B1 (en) | 2019-09-10 | 2024-01-24 | 주식회사 에이치엘클레무브 | Apparatus and Method for recognizing an object |
| CN111081033B (en) * | 2019-11-21 | 2021-06-01 | 北京百度网讯科技有限公司 | Method and device for determining orientation angle of vehicle |
| CN113344874B (en) * | 2021-06-04 | 2024-02-09 | 温州大学 | Pedestrian boundary crossing detection method based on Gaussian mixture modeling |
| CN114299308B (en) * | 2021-12-20 | 2024-09-06 | 中用科技有限公司 | Intelligent detection system and method for damaged road guardrail |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020021937A1 (en) * | 1999-09-23 | 2002-02-21 | Ochoa Carlos M. | Guardrail beam with improved edge region and method of manufacture |
| JP2005322002A (en) * | 2004-05-07 | 2005-11-17 | Kazuo Iwane | Compound artificial intelligence device |
| US20060041448A1 (en) * | 2004-08-20 | 2006-02-23 | Patterson Robbie L | Number of new and unique manufacturing and assembley methods and processes to cost effectively refit and market legacy implements like "The Gilhoolie" presently names "The Wili Grip" TM |
| JP2007183912A (en) * | 2005-12-06 | 2007-07-19 | Nissan Motor Co Ltd | Apparatus and method for detecting lane line |
| JP2011010010A (en) * | 2009-06-25 | 2011-01-13 | Sanyo Electric Co Ltd | Image reproduction device and imaging apparatus |
| CN103226835A (en) * | 2013-04-15 | 2013-07-31 | 北京大学深圳研究生院 | Target tracking method and system based on on-line initialization gradient enhancement regression tree |
| CN105488811A (en) * | 2015-11-23 | 2016-04-13 | 华中科技大学 | Depth gradient-based target tracking method and system |
| FR3035725A1 (en) * | 2015-04-28 | 2016-11-04 | Bluecime | METHOD FOR DETECTING GARDECORPS POSITION OF TELESIE NACELLES BY IMAGE ANALYSIS |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011243154A (en) | 2010-05-21 | 2011-12-01 | Mitsubishi Electric Corp | Device and method for detecting vehicle |
| EP2963633A4 (en) | 2013-02-27 | 2016-12-07 | Hitachi Automotive Systems Ltd | Object sensing device |
| JP5888275B2 (en) | 2013-03-29 | 2016-03-16 | アイシン・エィ・ダブリュ株式会社 | Road edge detection system, method and program |
| JP6295859B2 (en) | 2014-07-08 | 2018-03-20 | 株式会社デンソー | In-vehicle system, information processing apparatus, and program |
-
2017
- 2017-07-20 DE DE102017212418.0A patent/DE102017212418A1/en active Pending
-
2018
- 2018-07-18 CN CN201810787842.4A patent/CN109284664B/en active Active
- 2018-07-19 JP JP2018135652A patent/JP7164342B2/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020021937A1 (en) * | 1999-09-23 | 2002-02-21 | Ochoa Carlos M. | Guardrail beam with improved edge region and method of manufacture |
| JP2005322002A (en) * | 2004-05-07 | 2005-11-17 | Kazuo Iwane | Compound artificial intelligence device |
| US20060041448A1 (en) * | 2004-08-20 | 2006-02-23 | Patterson Robbie L | Number of new and unique manufacturing and assembley methods and processes to cost effectively refit and market legacy implements like "The Gilhoolie" presently names "The Wili Grip" TM |
| JP2007183912A (en) * | 2005-12-06 | 2007-07-19 | Nissan Motor Co Ltd | Apparatus and method for detecting lane line |
| JP2011010010A (en) * | 2009-06-25 | 2011-01-13 | Sanyo Electric Co Ltd | Image reproduction device and imaging apparatus |
| CN103226835A (en) * | 2013-04-15 | 2013-07-31 | 北京大学深圳研究生院 | Target tracking method and system based on on-line initialization gradient enhancement regression tree |
| FR3035725A1 (en) * | 2015-04-28 | 2016-11-04 | Bluecime | METHOD FOR DETECTING GARDECORPS POSITION OF TELESIE NACELLES BY IMAGE ANALYSIS |
| CN105488811A (en) * | 2015-11-23 | 2016-04-13 | 华中科技大学 | Depth gradient-based target tracking method and system |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111241900A (en) * | 2019-04-12 | 2020-06-05 | 泰州阿法光电科技有限公司 | Traffic environment field maintenance method |
| CN112446314A (en) * | 2020-11-19 | 2021-03-05 | 武汉中海庭数据技术有限公司 | Method and system for extracting guardrail elevation based on projection drawing |
| CN112446314B (en) * | 2020-11-19 | 2022-07-29 | 武汉中海庭数据技术有限公司 | Method and system for extracting guardrail elevation based on projection drawing |
| CN117090133A (en) * | 2023-08-23 | 2023-11-21 | 青岛迪乐普精密机械有限公司 | Guardrail and detection method thereof |
| CN117090133B (en) * | 2023-08-23 | 2024-06-11 | 青岛迪乐普精密机械有限公司 | Guardrail and detection method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| DE102017212418A1 (en) | 2019-01-24 |
| CN109284664B (en) | 2023-10-27 |
| JP2019021317A (en) | 2019-02-07 |
| JP7164342B2 (en) | 2022-11-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109284664A (en) | Driver assistance system and guardrail detection method | |
| Hur et al. | Multi-lane detection in urban driving environments using conditional random fields | |
| US10217007B2 (en) | Detecting method and device of obstacles based on disparity map and automobile driving assistance system | |
| CN109541583B (en) | Front vehicle distance detection method and system | |
| Xu et al. | Detection of sudden pedestrian crossings for driving assistance systems | |
| US9633436B2 (en) | Systems and methods for multi-dimensional object detection | |
| Bedruz et al. | Real-time vehicle detection and tracking using a mean-shift based blob analysis and tracking approach | |
| Zhou et al. | Vehicle classification using transferable deep neural network features | |
| CN106157283A (en) | The detection method of lane segmentation thing and device | |
| EP3261018A1 (en) | Device and method for object recognition of an input image for a vehicle | |
| Pyo et al. | Front collision warning based on vehicle detection using CNN | |
| CN113505638A (en) | Traffic flow monitoring method, traffic flow monitoring device and computer-readable storage medium | |
| El Jaafari et al. | A novel approach for on-road vehicle detection and tracking | |
| CN105512633A (en) | Power system dangerous object identification method and apparatus | |
| Lu et al. | Monocular multi-kernel based lane marking detection | |
| Dhanakshirur et al. | A framework for lane prediction on unstructured roads | |
| Chincholkar et al. | TRAFFIC SIGN BOARD DETECTION AND RECOGNITION FOR AUTONOMOUS VEHICLES AND DRIVER ASSISTANCE SYSTEMS. | |
| Changalasetty et al. | Moving vehicles classification in WEKA | |
| Zhang et al. | Accurate object tracking based on homography matrix | |
| Khan et al. | Multiple moving vehicle speed estimation using Blob analysis | |
| Mandlik et al. | A review on lane detection and tracking techniques | |
| Teknomo et al. | Tracking algorithm for microscopic flow data collection | |
| Shin et al. | Vehicle classification by road lane detection and model fitting using a surveillance camera | |
| Pirzada et al. | Single camera vehicle detection using edges and bag-of-features | |
| Han et al. | Neuromorphic convolutional recurrent neural network for road safety or safety near the road |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20230110 Address after: Germany Ingolstadt Applicant after: Continental Zhixing Germany Co.,Ltd. Address before: Hannover Applicant before: CONTINENTAL AUTOMOTIVE GmbH |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |