CN113159096A - Driving intention modeling method and recognition method based on simulated driver - Google Patents
Driving intention modeling method and recognition method based on simulated driver Download PDFInfo
- Publication number
- CN113159096A CN113159096A CN202110142108.4A CN202110142108A CN113159096A CN 113159096 A CN113159096 A CN 113159096A CN 202110142108 A CN202110142108 A CN 202110142108A CN 113159096 A CN113159096 A CN 113159096A
- Authority
- CN
- China
- Prior art keywords
- driver
- data
- driving intention
- test
- simulated
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a driving intention modeling method and a driving intention identification method based on a simulated driver, wherein the modeling method comprises the following steps: selecting a driver to use a simulation driver to carry out a simulation driving test, collecting a plurality of groups of driver face information data by using face recognition equipment, and adding a driver operation type label into each group of driver face information data to obtain original test data; preprocessing test data, wherein the preprocessing comprises data normalization and data division, and a training set and a test set are obtained after the data division; training a driving intention recognition model for predicting driver behavior prediction based on an SVM, wherein during training, facial information data of a driver are used as input variables, and a driver operation behavior label is used as an output variable; and testing the obtained SVM-based driving intention recognition model. The invention enables the driving intention to identify the model, the modeling process of the model is easy to implement, the model calculation amount is small, the requirement on calculation is low, and the prediction accuracy on the driver behavior is high.
Description
Technical Field
The invention relates to the field of traffic safety, in particular to a driving intention modeling method and an identification method based on a simulated driver.
Background
With the continuous development of the auxiliary driving technology and even the unmanned technology, more and more intelligent automobiles can be driven on the road in the visible future. However, due to the high development cost, the high price of related sensors and other devices, and the relatively backward traffic infrastructure, the fully unmanned vehicle can only be tested or used in a small range on a few roads where the unmanned vehicle is manually driven, such as specially developed test road sections or ports with few people. Although some unmanned vehicles of the vehicle enterprises are subjected to drive tests in urban roads, security personnel are still required to take over at any time to prevent accidents. It can be inferred that during a future period of time, it will occur that the driverless vehicle is traveling on the highway simultaneously with the driverless vehicle. In order to improve the safety of the unmanned vehicle, it is necessary to share data of the travel intention of the manned vehicle so that the unmanned vehicle can make a correct preparation for the operation. The intersection is a road section where traffic accidents often occur in real life, and the behaviors of drivers are various. Therefore, it is necessary to model the behavior of the driver at an intersection to predict its behavior and provide a reference for decision making of unmanned vehicles. The Chinese invention patent with the application number of CN201911398412.4 and the name of 'a method, a device and equipment for identifying driver behavior' provides a driver behavior identification method using an image identification technology, the method mainly designs and uses image information mining to obtain the driver action, the method provided by the patent excessively depends on the image identification technology, has higher requirement on computing capacity, and is not beneficial to large-scale popularization and use.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a driving intention modeling method and a driving intention recognition method based on a simulated driver, which are more convenient in data acquisition and have lower requirements for computing power, and establish a driving intention recognition model based on driver facial information data obtained by performing a simulated driving test using the simulated driver using an SVM algorithm.
In order to achieve the above object, the present invention firstly provides a driving intention modeling method based on a simulated driver, comprising the steps of:
selecting a driver to use a simulation driver to carry out a simulation driving test, collecting a plurality of groups of driver face information data by using face recognition equipment, and adding a driver operation type label into each group of driver face information data to obtain original test data;
preprocessing test data, wherein the preprocessing comprises data normalization and data division, and a training set and a test set are obtained after the data division;
training a driving intention recognition model for predicting driver behavior prediction based on an SVM, wherein during training, facial information data of a driver are used as input variables, and a driver operation behavior label is used as an output variable;
and testing the obtained SVM-based driving intention recognition model.
Further, each set of the driver facial feature data includes a pitch angle, a yaw angle, and a roll angle of the driver's face.
Furthermore, in the simulated driving test, the road model for simulating the driving environment adopts an urban road, the total length is not less than 2km, and at least 2 crossroads are provided.
The number of the selected drivers is not less than 10, and each driver repeatedly performs left-turn, right-turn and forward operation of the intersection for at least 30 times respectively; the data acquisition range is the range from the vehicle being 30m in front of the intersection to the vehicle being 50m behind the intersection. Specifically, when data is collected, in the process that a vehicle drives through the intersection, a range from a position within 30m of the vehicle driving to a position within 50m of the vehicle driving through a sidewalk stop line to a position just before the vehicle drives through the sidewalk stop line is an effective data range.
When the operation types of the drivers are classified, the adopted operation type labels of the drivers correspond to left turning, right turning and advancing respectively. In a preferred embodiment, the driver operation type labels are represented by the numbers "1", "2", "3" for left turn, right turn and forward, respectively.
Further, when the test data is normalized, the normalized test data is calculated by using the following formula:
in the formula, i is a data number; j is a variable type number; x represents a variable value; y represents data of the correlation variable after normalization; max is the maximum value of the correlation variable; min is the minimum value of the relevant variable.
Further, the normalized test data is divided into a training set and a testing set according to a random classification mode. In a preferred embodiment, the training set and test set have a 4:1 ratio of data quantities. In other embodiments, it may be in different ratios, such as 7:3, or 9:1, etc.
Further, when the model test is carried out, the driver face information data of the test data points in the test set is used as an input variable, the obtained SVM-based driving intention recognition model is input, and a predicted driver operation type label is obtained through calculation; if the driver operation type label predicted by the model is consistent with the actual driver operation type label corresponding to the test data point, the test data point is successfully predicted, otherwise, the prediction fails.
If the prediction success rate of the obtained SVM-based driving intention recognition model exceeds 85% for the whole test set, the model is acceptable; otherwise, more simulation driving tests need to be carried out.
The invention also provides a driver intention recognition method based on the simulated driver, which is used for recognizing the operation intention of the driver using the simulated driver according to the obtained SVM-based driving intention recognition model. Further, the step of recognizing the operation intention of the driver using the simulated driver using the SVM-based driving intention recognition model includes:
collecting face information data of a driver;
normalizing the data;
inputting the normalized data serving as an input variable into the SVM-based driving intention recognition model, and obtaining a driver operation type label through model calculation;
and correspondingly obtaining the driver operation type according to the driver operation type label.
Due to the adoption of the technical scheme, the invention achieves the following technical effects: the method is based on driver facial information data obtained by using a simulation driver to perform a simulation driving test, and establishes the relationship between the driver facial information data and the driver operation behavior by using an SVM algorithm, so that a high-reliability SVM-based driving intention recognition model can be obtained; the method has the advantages of easy implementation of the modeling process, convenient and fast data acquisition, high model calculation speed, lower requirement on calculation capacity and high identification accuracy rate of the driving intention of the driver.
Drawings
Fig. 1 is a flow chart diagram of a driving intention modeling method based on a simulated driver according to the invention.
Detailed Description
In order to make the technical solution of the embodiments of the present invention better understood, the technical solution of the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by equivalent changes and modifications by one skilled in the art based on the embodiments of the present invention, shall fall within the scope of the present invention.
Referring to fig. 1, the present embodiment provides a driving intention modeling method based on a simulated driver, including the following steps:
s1, carrying out test and collecting data
Selecting a driver to use a simulation driver to carry out a simulation driving test, using a face recognition device to collect a plurality of groups of driver face information data, and adding a driver operation type label into each group of driver face information data to obtain original test data. In the simulation driving test, the road model for simulating the driving environment adopts an urban road, the total length is not less than 2km, and at least 2 crossroads are provided. The number of the selected drivers is not less than 10, and each driver repeatedly performs left-turn, right-turn and forward operation of the intersection for at least 30 times respectively; the data acquisition range is the range from the vehicle being 30m in front of the intersection to the vehicle being 50m behind the intersection. Specifically, when data is collected, in the process that a vehicle drives through the intersection, a range from a position within 30m of the vehicle driving to a position within 50m of the vehicle driving through a sidewalk stop line to a position just before the vehicle drives through the sidewalk stop line is an effective data range. In this embodiment, the device used for acquiring the face information is an OptiTrack face information capture system. Each set of the driver facial feature data includes a pitch angle, a yaw angle, and a roll angle of the driver's face. In the simulated driving test, the frequency of the collected data is 50Hz, and about 30 ten thousand groups of test data are obtained.
For each group of driver face information data, when the driver operation types are classified, the adopted driver operation type labels respectively correspond to left turning, right turning and advancing. In the present embodiment, the driver operation type labels are represented by numerals "1", "2", "3" for left turn, right turn, and forward, respectively.
The invention also provides a driver intention recognition method based on the simulated driver, which is used for recognizing the operation intention of the driver using the simulated driver according to the obtained SVM-based driving intention recognition model. Further, the step of recognizing the operation intention of the driver using the simulated driver using the SVM-based driving intention recognition model includes:
collecting face information data of a driver;
normalizing the data;
inputting the normalized data serving as an input variable into the SVM-based driving intention recognition model, and obtaining a driver operation type label through model calculation;
and correspondingly obtaining the driver operation type according to the driver operation type label.
S2, data preprocessing
And preprocessing the test data, wherein the preprocessing comprises data normalization and data division, and a training set and a test set are obtained after the data division.
When the test data are normalized, the normalized test data are obtained by calculation by using the following formula:
in the formula, i is a data number; j is a variable type number; x represents a variable value; y represents data of the correlation variable after normalization; max is the maximum value of the correlation variable; min is the minimum value of the relevant variable.
The normalized test data is divided into a training set and a testing set according to a random classification mode. In this embodiment, the ratio of the number of data in the training set to the test set is 4: 1.
And testing the obtained SVM-based driving intention recognition model.
S3, training driving intention recognition model
Training an SVM-based driving intention recognition model for predicting driver behavior prediction based on an SVM algorithm by using a training set. During training, the face information data of the driver is used as an input variable, and the operation behavior label of the driver is used as an output variable. And taking the pitch angle, the yaw angle and the roll angle of the face of the driver of the data points in the training set as independent variables, and taking the corresponding driver operation type labels in the data points as dependent variables. After training, an SVM-based driving intention recognition model for predicting the operation type of the driver is obtained.
The example was trained using the Hewlett packard Z1G6 workstation, and the total training time of the model was 4 hours and 36 minutes.
S4, testing driving intention recognition model
When model testing is carried out, the driver face information data of the test data points in the test set is used as an input variable, the obtained driving intention recognition model based on the SVM is input, and a predicted driver operation type label is obtained through calculation; if the driver operation type label predicted by the model is consistent with the actual driver operation type label corresponding to the test data point, the test data point is successfully predicted, otherwise, the prediction fails.
If the prediction success rate (accuracy) of the obtained SVM-based driving intention recognition model exceeds 85% for the whole test set, the model is acceptable; otherwise, more simulation driving tests need to be carried out. In this example, the prediction accuracy of the entire test set was 87.3%, and the model was acceptable.
According to the model prediction success rate, when the model prediction success rate exceeds a preset threshold value by 85%, the judgment that the model precision meets the requirement is made, and the SVM-based driving intention recognition model obtained in the embodiment is acceptable. Otherwise, a supplementary driving simulation test is performed.
The embodiment also provides a driving intention recognition method based on the simulated driver, which adopts the SVM-based driving intention recognition model obtained in the modeling process to recognize the driving intention of the driver using the simulated driver. The method comprises the following specific steps:
collecting driver face information data: detecting a pitch angle, a yaw angle and a roll angle of the face of the driver in real time by using an OptiTrack face information capturing system;
data normalization: according to the normalization formula of the embodiment, the pitch angle, the yaw angle and the roll angle are normalized;
inputting a driving intention recognition model based on the SVM, which is obtained by modeling, by using the normalized data as an input variable, and obtaining a driver operation type label through model calculation;
and correspondingly obtaining the driver operation type according to the driver operation type label. When the driver operation type tag is "1", it indicates a left turn; when the driver operation type label is "2", it indicates a right turn; when the driver operation type flag is "3", it indicates straight traveling. According to the verification of the actual operation behavior and the recognized operation type of the driver, the driving intention recognition test is carried out on 10 drivers for 300 times in the embodiment, and the accuracy rate reaches 86%.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention; also, the above description should be understood as being readily apparent to those skilled in the relevant art and can be implemented, and therefore, other equivalent changes and modifications without departing from the concept disclosed herein are intended to be included within the scope of the present invention.
Claims (10)
1. A driving intention modeling method based on a simulated driver is characterized by comprising the following steps:
selecting a driver to use a simulation driver to carry out a simulation driving test, collecting a plurality of groups of driver face information data by using face recognition equipment, and adding a driver operation type label into each group of driver face information data to obtain original test data;
preprocessing test data, wherein the preprocessing comprises data normalization and data division, and a training set and a test set are obtained after the data division;
training a driving intention recognition model for predicting driver behavior prediction based on an SVM, wherein during training, facial information data of a driver are used as input variables, and a driver operation behavior label is used as an output variable;
and testing the obtained SVM-based driving intention recognition model.
2. The simulated driver based driving intent modeling method according to claim 1, wherein each set of said driver facial feature data comprises pitch angle, yaw angle, and roll angle of the driver's face.
3. The modeling method of driving intention based on the simulated driver according to claim 1, characterized in that in the simulated driving test, the road model for simulating the driving environment is an urban road, the total length is not less than 2km, and at least 2 crossroads are provided; the number of the selected drivers is not less than 10, and each driver repeatedly performs the operations of turning left, turning right and advancing at the intersection for at least 30 times.
4. The simulated driver-based driving intention modeling method according to claim 1, characterized in that the collected data range is a range from a vehicle at 30m in front of the intersection to 50 m; the driver operation type tags include type tags corresponding to left turn, right turn, and forward, which are represented by different numbers.
5. The driving intention modeling method based on a simulated driver as claimed in claim 1, characterized in that when the test data is normalized, the normalized test data is calculated using the following formula:
in the formula, i is a data number; j is a variable type number; x represents a variable value; y represents data of the correlation variable after normalization; max is the maximum value of the correlation variable; min is the minimum value of the relevant variable.
6. The simulated driver-based driving intention modeling method according to claim 1 or 5, characterized in that the normalized experimental data is divided into a training set and a testing set in a random classification manner.
7. The simulated driver-based driving intent modeling method according to claim 6, wherein the training set and the test set have a data quantity ratio of 4: 1.
8. The simulated driver-based driving intention modeling method according to claim 1, characterized in that, in performing model testing, the obtained SVM-based driving intention recognition model is inputted using the driver's facial information data of the test data points in the test set as input variables, and the predicted driver operation type label is calculated; if the driver operation type label predicted by the model is consistent with the actual driver operation type label corresponding to the test data point, the test data point is successfully predicted, otherwise, the prediction fails.
9. The simulated driver based driving intention modeling method according to claim 8, wherein if the prediction success rate of the obtained SVM based driving intention recognition model exceeds 85% for the entire test set, the model is acceptable; otherwise, more simulation driving tests need to be carried out.
10. A driver's intention recognition method based on a simulated driver, characterized in that the driver's operation intention using the simulated driver is recognized according to the SVM-based driving intention recognition model obtained in any one of claims 1 to 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110142108.4A CN113159096A (en) | 2021-02-02 | 2021-02-02 | Driving intention modeling method and recognition method based on simulated driver |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110142108.4A CN113159096A (en) | 2021-02-02 | 2021-02-02 | Driving intention modeling method and recognition method based on simulated driver |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113159096A true CN113159096A (en) | 2021-07-23 |
Family
ID=76879171
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110142108.4A Pending CN113159096A (en) | 2021-02-02 | 2021-02-02 | Driving intention modeling method and recognition method based on simulated driver |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113159096A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104494600A (en) * | 2014-12-16 | 2015-04-08 | 电子科技大学 | SVM (support vector machine) algorithm-based driver intention recognition method |
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 |
CN107316032A (en) * | 2017-07-06 | 2017-11-03 | 中国医学科学院北京协和医院 | One kind sets up facial image identifier method |
US20190129436A1 (en) * | 2017-10-28 | 2019-05-02 | TuSimple | System and method for real world autonomous vehicle trajectory simulation |
CN110427850A (en) * | 2019-07-24 | 2019-11-08 | 中国科学院自动化研究所 | Driver's super expressway lane-changing intention prediction technique, system, device |
CN112288023A (en) * | 2020-11-03 | 2021-01-29 | 浙江天行健智能科技有限公司 | Modeling method for aggressive driving recognition based on simulated driver and SVM algorithm |
-
2021
- 2021-02-02 CN CN202110142108.4A patent/CN113159096A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104494600A (en) * | 2014-12-16 | 2015-04-08 | 电子科技大学 | SVM (support vector machine) algorithm-based driver intention recognition method |
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 |
CN107316032A (en) * | 2017-07-06 | 2017-11-03 | 中国医学科学院北京协和医院 | One kind sets up facial image identifier method |
US20190129436A1 (en) * | 2017-10-28 | 2019-05-02 | TuSimple | System and method for real world autonomous vehicle trajectory simulation |
CN110427850A (en) * | 2019-07-24 | 2019-11-08 | 中国科学院自动化研究所 | Driver's super expressway lane-changing intention prediction technique, system, device |
CN112288023A (en) * | 2020-11-03 | 2021-01-29 | 浙江天行健智能科技有限公司 | Modeling method for aggressive driving recognition based on simulated driver and SVM algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110364008B (en) | Road condition determining method and device, computer equipment and storage medium | |
CN109242251B (en) | Driving behavior safety detection method, device, equipment and storage medium | |
CN105070053A (en) | Intelligent traffic monitoring camera for identifying vehicle illegal movement modes | |
CN114077541A (en) | Method and system for validating automatic control software for an autonomous vehicle | |
CN110077398B (en) | Risk handling method for intelligent driving | |
CN106781503A (en) | Method and apparatus for monitoring driving behavior | |
CN113033840B (en) | Method and device for judging highway maintenance | |
CN112249009B (en) | Vehicle speed control method, device and system and electronic equipment | |
Zhang et al. | A graded offline evaluation framework for intelligent vehicle’s cognitive ability | |
CN113851017A (en) | Pedestrian and vehicle identification and early warning multifunctional system based on road side RSU | |
CN111209880A (en) | Vehicle behavior identification method and device | |
CN111882924A (en) | Vehicle testing system, driving behavior judgment control method and accident early warning method | |
CN112069279A (en) | Map data updating method, device, equipment and readable storage medium | |
CN115782905B (en) | Automatic driving safety degree quantification system for driving vehicle | |
CN113511204A (en) | Vehicle lane changing behavior identification method and related equipment | |
CN115291515A (en) | Automatic driving simulation test system and evaluation method based on digital twinning | |
Li et al. | Research on lane change prediction model based on GBDT | |
Komol et al. | Deep transfer learning based intersection trajectory movement classification for big connected vehicle data | |
CN116978200A (en) | Method and system for monitoring and early warning of high-speed illegal vehicles | |
CN113159096A (en) | Driving intention modeling method and recognition method based on simulated driver | |
CN106504542A (en) | Speed intelligent monitoring method and system | |
CN112686127A (en) | GM-HMM-based driver overtaking intention identification method | |
CN115114786A (en) | Evaluation method, system and storage medium for traffic flow simulation model | |
Klitzke et al. | Extraction and analysis of highway on-ramp merging scenarios from naturalistic trajectory data | |
Hamad et al. | How Do Drivers Behave at Roundabouts in a Mixed Traffic? A Case Study Using Machine Learning |
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 |