CN109886304A - A kind of complex road condition following peripheral vehicle behavior recognition methods based on HMM-SVM bilayer improved model - Google Patents
A kind of complex road condition following peripheral vehicle behavior recognition methods based on HMM-SVM bilayer improved model Download PDFInfo
- Publication number
- CN109886304A CN109886304A CN201910058190.5A CN201910058190A CN109886304A CN 109886304 A CN109886304 A CN 109886304A CN 201910058190 A CN201910058190 A CN 201910058190A CN 109886304 A CN109886304 A CN 109886304A
- Authority
- CN
- China
- Prior art keywords
- hmm
- svm
- vehicle
- bilayer
- model
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a kind of complex road condition following peripheral vehicle behavior recognition methods based on HMM-SVM bilayer improved model, including 1) off-line training: dividing exemplary perimeter vehicle behavior, extract the state characteristic information of nearby vehicle, this vehicle is reached by car networking, the state characteristic information under road co-ordinates system is converted into conjunction with this vehicle information, feature vector, X is generated, X is inputed to HMM and SVM parameter learning respectively；2) model refinement: disposing processor threshold among HMM and SVM, obtains best difference factor using NSGA-II algorithm optimization, obtains HMM-SVM bilayer improved model；3) on-line testing: this vehicle is distinguished using HMM-SVM bilayer improved model is tracked the affiliated behavior pattern of vehicle.The present invention utilizes the road co-ordinates system of high-precision map structuring, expands the application of nearby vehicle system of behavior；HMM excellent extremely strong two classification capacity of time series modeling ability and SVM is organically combined, and bilayer model is improved, improves the accuracy rate and recognition speed of vehicle behavior identification.
Description
Technical field
The invention belongs to Vehicular intelligent driving technology fields, and in particular to a kind of based on HMM-SVM bilayer improved model
The vehicle behavior recognition methods of complex road condition following peripheral.
Background technique
Road scene understanding is one of important component of vehicle drive assist system, while being also that Shape Of Things To Come is automatic
The fundamental importance of driving.Reality traffic environment be traffic participant more than one influence each other, the complication system of dynamic change,
Intelligent automobile will not only have visual perception, detection and tracking ability in this complicated system, it is also necessary to can pass through Che-road shape
The behavior of nearby vehicle is even predicted in the identification of state information, to improve perceived depth.
When carrying out Activity recognition to nearby vehicle, it is essential that the accurate practical nearby vehicle state of acquisition is characterized in
's.Existing vehicle behavior recognition methods is defaulted as straight trip section when obtaining vehicle-state feature, by road, is not particularly suited for
The identification of vehicle behavior under complex road condition (curved bend, U-bend etc.).High-precision map refers to the ground of high-precision, fining definition
Figure, it is reachable to cover data, the precision such as road network data, lane line data, lane line, the type of road edge and traffic sign
To Centimeter Level.And in recent years, with the rapid development of automatic Pilot industry, the high-precision map industry entry fast traffic lane of development,
Many mainstream electronic Map enterprises such as Gao De, Baidu etc. have also gradually externally opened for free the data of high-precision map, give
Its extensive use has welcome opportunity.It using high-precision map establishes road co-ordinates system, obtains vehicle in this coordinate system
State characteristic information breaches the limitation only identified in straight road to vehicle behavior, has expanded nearby vehicle system of behavior
Application.
The machine learning algorithm of related vehicle behavior identification can be roughly divided into Bayesian network, Hidden Markov mould at present
Type, support vector machines and Method Using Relevance Vector Machine etc..Wherein, Hidden Markov (HMM) model is because having powerful time series modeling ability
It is most widely used, but its influence for having ignored negative sample, and is only classified with maximum likelihood value, especially when output
When otherness is smaller between multidimensional probability, it be easy to cause higher false recognition rate.Algorithm of support vector machine (SVM) is asked in two classification
Have significant advantage in topic, the otherness between classification can be reflected to a greater degree, can by by lower dimensional space linearly not
The sample that can divide maps in higher dimensional space, is separated similar sample well with Euclidean distance as big as possible.Therefore,
For extremely strong two classification capacity of time series modeling ability and SVM for having made full use of HMM excellent, while considering the reality of recognition result
Shi Xingyu reliability, the invention propose a kind of HMM-SVM bilayer improved model, introduce difference factor, two kinds of classifiers of bringing into play
Respective advantage, to significantly improve the recognition capability of nearby vehicle system of behavior.
Summary of the invention
The present invention is directed to the requirement of nearby vehicle Activity recognition real-time, reliability, proposes a kind of bis- based on HMM-SVM
The complex road condition following peripheral vehicle behavior recognition methods of layer improved model, can be real-time and accurately to complex road condition following peripheral vehicle
Behavior make identification, provide reference frame for the decision rule of intelligent vehicle.
The purpose of the present invention can be achieved through the following technical solutions, a kind of based on HMM-SVM bilayer improved model
The vehicle behavior recognition methods of complex road condition following peripheral, specifically includes:
Step (1): off-line training step
N kind exemplary perimeter vehicle behavior is divided, the state characteristic information of nearby vehicle is extracted, this is reached by car networking
Ben Che and the conversion of the state characteristic information of nearby vehicle are the state characteristic information under road co-ordinates system by vehicle, and raw
Feature vector, X (△ X, △ Y, the △ V of Cheng XinX, △ VY, △ aX, △ aY), by the feature vector, X of same category nearby vehicle behavior
It is input to HMM model as observation sequence, obtains the HMM of each vehicle behavior classification；Nearby vehicle behavior is one group two-by-two, will
The corresponding feature vector of two kinds of behavior classifications is defeated by each SVM model and carries out parameter learning in every group, obtains every group of SVM model；
Step (2): model refinement stage
By trained HMM model and SVM model group at bilayer model, and threshold process is disposed among two-layer model
Device then directly exports HMM layers of recognition result when the otherness between processor threshold obtains probability value is larger；Conversely, by general
The corresponding behavior of maximum two HMM models of rate value extracts, and selects corresponding SVM model to carry out two classification, exports SVM layers
Recognition result is optimized using NSGA-II algorithm, obtains best difference factor σ, finally show that HMM-SVM bilayer improves mould
Type；
Step (3): on-line testing stage
Tracked target vehicle gives this vehicle, this Che Liyong by car networking real-time Transmission from vehicle driving information for collected
Trained HMM-SVM bilayer improved model, which distinguishes, is tracked the affiliated behavior pattern of vehicle.
Further, the exemplary perimeter vehicle behavior includes with speeding, Zuo Huandao, right lane-change, overtakes other vehicles.
Further, road co-ordinates system is to hang down using the center line in lane as reference axis X with the tangent line of reference axis X
Straight line is reference axis Y.
Further, the feature vector under road co-ordinates system include this vehicle of t moment position (Xego, Yego),
Speed (VXego, VYego), acceleration (aXego, aYego) and the position (Xaro, Yaro) of nearby vehicle, speed (VXaro, VYaro)、
Acceleration (aXaro, aYaro)。
Further, in described eigenvector X, △ X=Xaro-Xego, △ Y=Yaro-Yego, △ VX=VXaro-
VXego、△VY=VYaro-VYego、△aX=aXaro-aXego、△aY=aYaro-aYego。
Further, the NSGA-II algorithm is using recognition correct rate and recognition time as optimization aim, the optimization aim letter
Number is f=T0-u1-u2-…uN, wherein u1, u2..., uNIt is the recognition correct rate of each behavior of training sample, T0To identify total time.
Further, the classifier equation of two classification is f (x)=sign (ωTX+b), wherein ω is adjustable power
It is worth vector, b is amount of bias.
The invention has the benefit that
(1) stage is obtained in all Che Yuben car state features, has introduced a kind of road connection using high-precision map structuring
Coordinate system is closed, the limitation only identified in straight road to vehicle behavior is can break through, has expanded answering for nearby vehicle system of behavior
Use occasion；
(2) on the basis of HMM single model identification nearby vehicle behavior, increase SVM model, form bilayer model to know
HMM excellent extremely strong two classification capacity of time series modeling ability and SVM is organically combined, is given full play to by other nearby vehicle behavior
Two kinds of respective advantages of classifier；
(3) using recognition correct rate and recognition time as optimization aim founding mathematical models, using NSGA-II algorithm to threshold value
Processor carries out multiple-objection optimization to obtain difference factor σ, is improved HMM-SVM bilayer model, and vehicle behavior is improved
The accuracy rate and recognition speed of identification.
Detailed description of the invention
Fig. 1 is the schematic diagram of establishing of road co-ordinates system, and Fig. 1 (a) is that straight way in co-ordinates system establishes schematic diagram,
Fig. 1 (b) is that bend in co-ordinates system establishes schematic diagram；
Fig. 2 is the complex road condition following peripheral vehicle behavior recognition methods block diagram based on HMM-SVM bilayer improved model.
Specific embodiment
Below in conjunction with attached drawing, further description of the technical solution of the present invention, but protection scope of the present invention is simultaneously
It is without being limited thereto.
Step1: state feature obtains and data processing
Conclude first and divide the typical nearby vehicle behavior of N kind, respectively with speeding, Zuo Huandao, right lane-change, overtake other vehicles.
It sets each laboratory vehicle and is equipped with the interface connecting with high-precision map, the high-precision map is in addition to containing a large amount of road
Information, an important feature for being different from traditional map is exactly precision, can reach the precision of Centimeter Level, and laboratory vehicle can be helped real
Existing high accuracy positioning.Each laboratory vehicle all has an independent ID, is accessed by OBD (On-Board Diagnostics) interface
The data extracted from vehicle are uploaded to cloud by car networking, by interacting in real time with other vehicles from the background.In view of data transmission
Sample frequency is set 50Hz by real-time and robustness, i.e. the acquisition front and back when a length of 0.02s between data twice.It is selected
One laboratory vehicle is this vehicle (Ego), around vehicle all around be nearby vehicle (Around1,2 ...).Here week
Side vehicle not only includes in the device-awares such as this vehicle driver's seat or vehicle-mounted vidicon, laser radar, millimetre-wave radar region
Vehicle, further include the vehicle in blind area, such as the vehicle that sky way turning mouth is blocked by massif.
For the target vehicle that identified in nearby vehicle, the shape of this vehicle download online nearby vehicle from car networking
State characteristic information.Accessed high-precision map is utilized, is naturally representation, pick-up road with reference to movement particles in rational mechanics
Center line be reference axis X, take the line vertical with the tangent line of reference axis X be reference axis Y, establish road co-ordinates system, such as Fig. 1
Shown, Fig. 1 (a) is establish schematic diagram of the straight way in co-ordinates system, and Fig. 1 (b) is that bend is illustrated in the foundation of co-ordinates system
Figure.By Ben Che and the nearby vehicle state characteristic information downloaded, the characteristic information being converted under road co-ordinates system, i.e. t
The truck position Shi Keben (Xego, Yego), speed (VXego, VYego), acceleration (aXego, aYego) and nearby vehicle position
(Xaro, Yaro), speed (VXaro, VYaro), acceleration (aXaro, aYaro), convert features described above information to new feature vector
X (△ X, △ Y, △ VX, △ VY, △ aX, △ aY), wherein △ X=Xaro-Xego, △ Y=Yaro-Yego, △ VX=VXaro-
VXego, △ VY=VYaro-VYego, △ aX=aXaro-aXego, △ aY=aYaro-aYego.Each typical case is obtained according to above-mentioned steps equivalent
The corresponding set of eigenvectors of nearby vehicle behavior classification is as training sample set, wherein according to behavior classification, training sample set can
It is set as B1, B2..., BN。
Step2: model training study
(1) Hidden Markov Model training study
Hidden Markov Model (Hidden Markov Model, HMM) is the probabilistic model about timing, by considerable
The variable measured removes research nonobservable variable.Hidden markov model can be set as a five-tuple (Q, V, A, B, π), wherein
Hidden state Q={ Q1, Q2..., QN, N is the number of hidden state；Observable state V={ V1, V2..., VM, M is observation shape
The number of state；Hidden state transition probability matrix A=[αij]N×NElement representation HMM model between each hidden state
Transition probability, αijBe t moment hidden state be Qi, t+1 moment hidden state be QjProbability, aij=P (It+1=Qj|It
=Qi), i=1,2 ..., N；J=1,2 ..., N, I are the status switch that length is T, and I={ I1,I2,…,IT}；Confusion matrix B
=[bj(k)]N×MElement representation HMM model in transition probability between each hidden state and observation state, bj(k) it indicates
In t moment, hidden state Qj, observation state OtProbability, bj(k)=P (Ot=Vk|It=Qj), k=1,2 ..., M；J=
1,2 ..., N, O are corresponding observation sequences；Initial state probabilities matrix π=(πi), wherein πi=P (I1=Qi), i=1,2 ...,
N indicates each hidden state Q of initial time t=1iProbability.
First to the identification HMM model initialization of each vehicle behavior, initial parameter N, M, A, B, π are obtained；It will be in Step1
The feature vector, X of the same category nearby vehicle behavior (with speeding, Zuo Huandao, right lane-change, overtaking other vehicles) of acquisition is as the defeated of HMM model
Enter, according to the parameter after model initialization, model λ=(A, B, π) parameter is adjusted using Baum-Welch iterative algorithm, is made general
Rate function maximization, progressive updating model parameter finally obtain the corresponding HMM of each vehicle behavior classification, complete first layer mould
The study of type.
(2) supporting vector machine model training study
Support vector machines (Support Vector Machine, SVM) model is current most popular two classifier,
Many distinctive advantages are shown in solution small sample, the identification of non-linear and high dimensional pattern.The basic principle is that finding one
Meet the optimal hyperlane of data classification requirement, so that hyperplane is in the case where ensuring nicety of grading, hyperplane and two class samples
This point distance is maximum.The hyperplane should meet ωTX+b=0, wherein ω is adjustable weight vector, and b is amount of bias, and X is spy
Vector is levied, T is the transposition symbol of matrix.Optimal hyperlane requires class interval to maximize, and the distance of two parallel hyperplane is
2/ | | ω | |, that is, require | | ω | | minimize, i.e., have minimum equation: φ (ω)=1/2 when being solved | | ω | |
=1/2 (ω, ω), to make all samples outside hyperplane, above formula should also meet constraint condition YiωT·Xi+ b > 1, Yi∈
{ -1,1 }, i=1,2 ... l, wherein YiIndicate sample class, l is number of samples.
Two-by-two it is one group by N kind exemplary perimeter vehicle behavior, is divided into h=N (N-1)/2 group, every group includes two kinds of behaviors
Classification i, j, a kind of to be used as positive sample, another kind is used as negative sample.The corresponding training sample set of positive negative sample will be obtained in Step1
Bi、BjMerge into the training sample set (B of the groupi, Bj), for training SVM model.Based under MATLAB environment, libsvm is used
Support vector machines tool, respectively by every group of training sample set (Bi, Bj) feature vector, X be defeated by each SVM model and carry out parametrics
It practises, finds out every group of optimal hyperlane result fk(x)=ωTX+b, k=1,2 ... h, so that obtaining each group carries out two classification
SVM model.
Step3: the double-deck improved model is formed
By trained HMM model and SVM model group at bilayer model, and threshold process is disposed among two-layer model
The corresponding training sample set of each exemplary perimeter vehicle behavior is formed total training sample set (B1, B2 ..., BN) by device, will be every
The feature vector, X of a sample is input to HMM model as observation sequence, and the multidimensional output probability of each nearby vehicle behavior
(P1, P2..., Pn) input as processor threshold.Processor threshold is compared each output probability, extracts maximum
Two probability Psmax-i、Pmax-ii；If σ is the difference factor in processor threshold, work as Pmax-ii/Pmax-iWhen≤σ, then HMM model is defeated
Otherness out between result is larger, directly exports the recognition result of the first layer model, i.e., each model output probability is maximum
(Pmax-i) corresponding to behavior.Work as Pmax-ii/Pmax-iWhen >=σ, then the otherness between HMM model output result is smaller, at this moment
If direct output probability maximum (Pmax-i) corresponding to behavior, be easy to produce higher identification error rate, need to carry out second
The identification of layer model.By Pmax-i、Pmax-iiCorresponding behavior Qi、QiiIt extracts, feature vector, X is input to behavior Qi、QiiPlace
In that group of SVM model, by SVM model classifiers Equation f (x)=sign (ωTX+b the behavior class where the sample) is exported
Not, using this result as final recognition result.The real-time that bilayer model is identified in view of the threshold value σ in processor threshold with
Reliability has a major impact, and the recognition result of training sample set is further analyzed in the present invention, takes each behavior of training sample
Recognition correct rate be u1, u2..., uN, identification total time is T0。
To improve to HMM-SVM bilayer model, mathematics is established using recognition correct rate and recognition time as optimization aim
Model carries out multiple-objection optimization to processor threshold using NSGA-II algorithm.II algorithm of NSGA- is most popular at present
One of multi-objective genetic algorithm has the advantages that the speed of service is fast, the convergence of disaggregation is good.Definition optimization object function is f=
T0-u1-u2-…uN, optimize difference factor σ for the improvement of recognition correct rate and recognition time, final acquisition HMM-SVM is bis-
Layer improved model.
Step4: on-line testing stage
Tracked target vehicle gives main vehicle by car networking real-time Transmission from vehicle driving information for collected, this vehicle combines
Feature of two vehicles under road co-ordinates system constitutes new feature vector, utilizes trained HMM-SVM bilayer improved model
It distinguishes and is tracked the affiliated behavior pattern of vehicle.
The embodiment is a preferred embodiment of the present invention, but present invention is not limited to the embodiments described above, not
In the case where substantive content of the invention, any conspicuous improvement that those skilled in the art can make, replacement
It all belongs to the scope of protection of the present invention.
Claims (8)
1. a kind of complex road condition following peripheral vehicle behavior recognition methods based on HMM-SVM bilayer improved model, which is characterized in that
It specifically includes:
Step (1): off-line training step
N kind exemplary perimeter vehicle behavior is divided, the state characteristic information of nearby vehicle is extracted, this vehicle is reached by car networking,
By the state characteristic information that Ben Che and the conversion of the state characteristic information of nearby vehicle are under road co-ordinates system, and generate new
Feature vector, X (△ X, △ Y, △ VX, △ VY, △ aX, △ aY), using the feature vector, X of same category nearby vehicle behavior as
Observation sequence is input to HMM model, obtains the HMM of each vehicle behavior classification；Nearby vehicle behavior is one group two-by-two, by every group
In the corresponding feature vector of two kinds of behavior classifications be defeated by each SVM model and carry out parameter learning, obtain every group of SVM model；
Step (2): model refinement stage
By trained HMM model and SVM model group at bilayer model, and processor threshold is disposed among two-layer model, when
When processor threshold show that the otherness between probability value is larger, then HMM layers of recognition result are directly exported；Conversely, by probability value
The corresponding behavior of maximum two HMM models extracts, and corresponding SVM model is selected to carry out two classification, and SVM layer of output identify
As a result, optimizing using NSGA-II algorithm, best difference factor σ is obtained, finally obtains HMM-SVM bilayer improved model；
Step (3): on-line testing stage
Tracked target vehicle gives this vehicle by car networking real-time Transmission from vehicle driving information for collected, this vehicle utilizes training
Good HMM-SVM bilayer improved model, which distinguishes, is tracked the affiliated behavior pattern of vehicle.
2. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 1
Recognition methods, which is characterized in that the exemplary perimeter vehicle behavior includes with speeding, Zuo Huandao, right lane-change, overtakes other vehicles.
3. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 1
Recognition methods, which is characterized in that road co-ordinates system is using the center line in lane as reference axis X, with cutting for reference axis X
The vertical line of line is reference axis Y.
4. a kind of complex road condition following peripheral vehicle row based on HMM-SVM bilayer improved model according to claim 1 or 3
For recognition methods, which is characterized in that the feature vector under road co-ordinates system include this vehicle of t moment position (Xego,
Yego), speed (VXego, VYego), acceleration (aXego, aYego) and the position (Xaro, Yaro) of nearby vehicle, speed (VXaro,
VYaro), acceleration (aXaro, aYaro)。
5. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 4
Recognition methods, which is characterized in that in described eigenvector X, △ X=Xaro-Xego, △ Y=Yaro-Yego, △ VX=VXaro-
VXego、△VY=VYaro-VYego、△aX=aXaro-aXego、△aY=aYaro-aYego。
6. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 1
Recognition methods, which is characterized in that the NSGA-II algorithm is using recognition correct rate and recognition time as optimization aim.
7. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 6
Recognition methods, which is characterized in that the optimization object function is f=T0-u1-u2-…uN, wherein u1, u2..., uNIt is trained sample
The recognition correct rate of this each behavior, T0To identify total time.
8. a kind of complex road condition following peripheral vehicle behavior based on HMM-SVM bilayer improved model according to claim 1
Recognition methods, which is characterized in that the classifier equation of two classification is f (x)=sign (ωTX+b), wherein ω is adjustable
Weight vector, b is amount of bias.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910058190.5A CN109886304B (en) | 2019-01-22 | 2019-01-22 | HMM-SVM double-layer improved model-based surrounding vehicle behavior recognition method under complex road conditions |
| PCT/CN2019/077584 WO2020151059A1 (en) | 2019-01-22 | 2019-03-11 | Method for recognizing surrounding vehicle behaviors in complex road conditions based on hmm-svm two-layer improved model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910058190.5A CN109886304B (en) | 2019-01-22 | 2019-01-22 | HMM-SVM double-layer improved model-based surrounding vehicle behavior recognition method under complex road conditions |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109886304A true CN109886304A (en) | 2019-06-14 |
| CN109886304B CN109886304B (en) | 2023-09-29 |
Family
ID=66926458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910058190.5A Active CN109886304B (en) | 2019-01-22 | 2019-01-22 | HMM-SVM double-layer improved model-based surrounding vehicle behavior recognition method under complex road conditions |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109886304B (en) |
| WO (1) | WO2020151059A1 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110288835A (en) * | 2019-06-28 | 2019-09-27 | 江苏大学 | A kind of nearby vehicle behavior real-time identification method based on kinematics predictive compensation mechanism |
| CN111354193A (en) * | 2020-02-26 | 2020-06-30 | 江苏大学 | Highway vehicle abnormal behavior early warning system based on 5G communication |
| WO2021093335A1 (en) * | 2019-11-14 | 2021-05-20 | Suzhou Zhijia Science & Technologies Co., Ltd. | Method for automatically labeling lane changing intention based on high-noise trajectory data of vehicle |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112069599B (en) * | 2020-08-28 | 2023-08-22 | 北京科技大学 | Method for predicting reliability of running performance of tracked vehicle |
| CN114299742B (en) * | 2022-01-20 | 2023-10-13 | 福建工程学院 | Speed limit information dynamic identification and update recommendation method for expressway |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971194A (en) * | 2017-02-16 | 2017-07-21 | 江苏大学 | A kind of driving intention recognition methods based on the double-deck algorithms of improvement HMM and SVM |
| CN107967486A (en) * | 2017-11-17 | 2018-04-27 | 江苏大学 | A kind of nearby vehicle Activity recognition method based on V2V communications with HMM-GBDT mixed models |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9784592B2 (en) * | 2015-07-17 | 2017-10-10 | Honda Motor Co., Ltd. | Turn predictions |
| US20180190377A1 (en) * | 2016-12-30 | 2018-07-05 | Dirk Schneemann, LLC | Modeling and learning character traits and medical condition based on 3d facial features |
| CN108470460B (en) * | 2018-04-11 | 2020-08-28 | 江苏大学 | Peripheral vehicle behavior identification method based on smart phone and RNN |
-
2019
- 2019-01-22 CN CN201910058190.5A patent/CN109886304B/en active Active
- 2019-03-11 WO PCT/CN2019/077584 patent/WO2020151059A1/en active Application Filing
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971194A (en) * | 2017-02-16 | 2017-07-21 | 江苏大学 | A kind of driving intention recognition methods based on the double-deck algorithms of improvement HMM and SVM |
| CN107967486A (en) * | 2017-11-17 | 2018-04-27 | 江苏大学 | A kind of nearby vehicle Activity recognition method based on V2V communications with HMM-GBDT mixed models |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110288835A (en) * | 2019-06-28 | 2019-09-27 | 江苏大学 | A kind of nearby vehicle behavior real-time identification method based on kinematics predictive compensation mechanism |
| CN110288835B (en) * | 2019-06-28 | 2021-08-17 | 江苏大学 | Surrounding vehicle behavior real-time identification method based on kinematic prediction compensation mechanism |
| WO2021093335A1 (en) * | 2019-11-14 | 2021-05-20 | Suzhou Zhijia Science & Technologies Co., Ltd. | Method for automatically labeling lane changing intention based on high-noise trajectory data of vehicle |
| CN111354193A (en) * | 2020-02-26 | 2020-06-30 | 江苏大学 | Highway vehicle abnormal behavior early warning system based on 5G communication |
| CN111354193B (en) * | 2020-02-26 | 2021-09-10 | 江苏大学 | Highway vehicle abnormal behavior early warning system based on 5G communication |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020151059A1 (en) | 2020-07-30 |
| CN109886304B (en) | 2023-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886304A (en) | A kind of complex road condition following peripheral vehicle behavior recognition methods based on HMM-SVM bilayer improved model | |
| CN107492251B (en) | Driver identity recognition and driving state monitoring method based on machine learning and deep learning | |
| CN108074244B (en) | Safe city traffic flow statistical method integrating deep learning and background difference method | |
| CN109460023A (en) | Driver's lane-changing intention recognition methods based on Hidden Markov Model | |
| CN105701479A (en) | Intelligent vehicle multi-laser radar fusion recognition method based on target features | |
| Rawashdeh et al. | Collaborative automated driving: A machine learning-based method to enhance the accuracy of shared information | |
| CN101226155A (en) | Intelligentize lung cancer early cell pathological picture recognition processing method | |
| CN107985189B (en) | Early warning method for lane changing depth of driver in high-speed driving environment | |
| CN103921743A (en) | Automobile running working condition judgment system and judgment method thereof | |
| CN109050535A (en) | A kind of quick landform industry and mining city method based on vehicle attitude | |
| CN110188797A (en) | A kind of intelligent automobile method for rapidly testing based on Bayes's optimization | |
| Zhang et al. | A framework for turning behavior classification at intersections using 3D LIDAR | |
| CN112462381A (en) | Multi-laser radar fusion method based on vehicle-road cooperation | |
| CN111474934A (en) | Personalized cruise control system based on high-precision map | |
| CN108470460B (en) | Peripheral vehicle behavior identification method based on smart phone and RNN | |
| CN110991377B (en) | Front mesh identification method of automobile safety auxiliary system based on monocular vision neural network | |
| CN109858951A (en) | New-energy automobile is powered on or changes the prediction of electricity demanding | |
| CN113537411B (en) | Improved fuzzy clustering method based on millimeter wave radar | |
| CN113611008B (en) | Vehicle driving scene acquisition method, device, equipment and medium | |
| Zheng et al. | Construction of driving conditions of Harbin urban passenger cars | |
| CN115129886A (en) | Driving scene recognition method and device and vehicle | |
| CN112131996B (en) | Road side image multi-scale pedestrian rapid detection method based on channel separation convolution | |
| CN115015908A (en) | Radar target data association method based on graph neural network | |
| CN114973195A (en) | Vehicle tracking method, device and system based on multi-information fusion | |
| CN110334608B (en) | Real-time traffic speed limit sign identification method of vehicle-mounted image |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |