﻿ 改进的滑动窗口算法与SVM在换道行为识别中的应用
 计算机系统应用  2019, Vol. 28 Issue (1): 113-118 PDF

Application of Improved Sliding Window Algorithm and SVM in Vehicle Lane Change Behavior Recognition
LI Huan, CHENG Han-Han, WANG Ji-Wu, AN Yi-Sheng
School of Information Engineering, Chang’an University, Xi’an 710064, China
Foundation item: National Natural Science Foundation of China (61703053)
Abstract: In order to reduce the probability of accidents caused by bad lane changing behavior, it is necessary to identify the lane changing behavior during the actual driving of the vehicle. In this paper, the IOS intelligent device is used to collect data, and the corresponding feature vector is established. The vehicle lane change behavior recognition model based on support vector machine is proposed. An improved N-δ sliding window interception algorithm is proposed for the recognition of continuous lane change behavior so as to divide the data containing multiple behaviors quickly, the sample data is used to verify the feasibility of the N-δ sliding window interception algorithm and the validity of the classifier.
Key words: N-δ sliding window interception algorithm     Support Vector Machine (SVM)     vehicle lane change behavior recognition     feature extraction

 图 1 车辆换道过程

1 支持向量机概论

1.1 线性分类

 图 2 线性分类的最优超平面

H在样本空间可表示为:

 ${{{\omega }}^{\rm T}}{{x}} + b = 0$ (1)

 $f\left( {{x}} \right) = {{{\omega }}^{\rm T}}{{x}} + b$ (2)

 ${y_i}\left( {{{{\omega }}^{\rm T}}{{{x}}_i} + b} \right) \geqslant 1, \;\; i = 1,2,\cdots,m$ (3)

 $\gamma = \frac{2}{{\parallel {{\omega }}\parallel }}$ (4)

SVM寻找最优超平面H即寻找满足约束(3)中的 ${{\omega }}$ b, 使得 $\gamma$ 最大. 推导可得SVM的基本型:

 $\left\{\begin{array}{l}\mathop {\min }\limits_{{{\omega }},b} {\rm{ }}\frac{1}{2}\parallel {{\omega }}{\parallel ^2}\\{\rm{s}}{\rm{.t}}{\rm{. }}\;{y_i}\left( {{{{\omega }}^{\rm T}}{{{x}}_i} + b} \right) \ge 1, \;\; i = 1,2,\cdots, m\end{array}\right.$ (5)

 $\begin{array}{l}f\left( {{x}} \right) = {{{\omega }}^{\rm T}}{{x}} + b\\\;\;\;\;\;\;\;\;\;{\rm{ = }}\displaystyle\sum\limits_{i = 1}^m {{\alpha _i}{y_i}{{x}}_i^{\rm T}{{x}} + b} \end{array}$ (6)

1.2 非线性分类

 $\begin{array}{l}f\left( {{x}} \right) = {{{\omega }}^{\rm T}}\phi \left( {{x}} \right) + b\\\;\;\;\;\;\;\;\;\;{\rm{ = }}\displaystyle\sum\limits_{i = 1}^m {{\alpha _i}{y_i}\kappa \left( {{{{x}}_i},{{{x}}_j}} \right) + b} \end{array}$ (7)
1.3 近似线性可分

 $\begin{array}{l}\mathop {\min }\limits_{{{\omega }},b} \;\;\;\frac{1}{2}\parallel {{\omega }}{\parallel ^2}{{ + C}}\sum\limits_{i = 1}^m {{\xi _i}} \\{\rm{s}}{\rm{.t}}{\rm{.}}\;\;\;{y_i}\left( {{{{\omega }}^{\rm T}}{{{x}}_i} + b} \right) \ge 1 - {\xi _i}\\\;\;\;\;\;\;{\xi _i} \ge 0,\;\;\;\;\;\;i = 1,2,\cdots,m\end{array}$ (8)
2 数据模块 2.1 数据获取

IOS智能设备中嵌入的加速计和陀螺仪能够感知宿主设备的状态. 我们将iPad固定在车辆上, 获取车辆行驶过程中加速计和陀螺仪的数据来代表宿主车辆的行为数据, 借此实现了低成本的数据采集. 提取数据特征来构造特征向量, 使用SVM进行模型的训练和测试.

2.2 数据预处理

 $y'(i) = y(i) - \overline y$ (9)

 y'(i) = \left\{ {\begin{aligned} & {y(i) ,\quad\quad\quad\quad\quad\quad\quad\;\; i = 0} \\ &{0.5y(i) + 0.5y'(i - 1) ,\quad i > 0{\rm{ }}} \end{aligned}} \right. (10)

2.3 特征提取

 图 3 向左换道加速计曲线图

 图 4 向左换道陀螺仪曲线图

 图 5 向右换道加速计曲线图

3 模型建立

 图 6 向右换道陀螺仪曲线图

3.1 单一换道行为模型

1) 选择由换道时和正常行驶时陀螺仪Z轴的数据得到的特征向量构成训练样本集;

2) 基于1)得到的训练样本构造支持向量机;

3) 使用拉格朗日乘子法求解2)中构造的支持向量机.

3.2 连续换道行为模型

 $MES(i) = \frac{1}{k}(g{(i)^2} + g{(i + 1)^2} + \cdots + g{(i + k - 1)^2})$ (11)

 图 7 车辆换道行为识别整体架构图

1　 N←第1个样本点到第p个样本点的数据窗口;

2　 TcN窗口内最后δ个样本点的能量值;

3　 while(剩余样本点数>250)

4　　 while( $\scriptstyle {T_c} \geqslant {\bar T_s}$ )

5　　　 N←第1个样本点到第N-δ个样本点的数据窗口;

6　　　 TcN窗口内最后δ个样本点的能量值;

7　　 end while

8　　 M←第N+1个样本点到第N+p个样本点的数据窗口;

9　　 TcM窗口内最后δ个样本点的能量值;

10　　 while( $\scriptstyle {T_c} \geqslant {\bar T_e}$ )

11　　　 M←第N+1个样本点到第M+δ个样本点的数据窗口;

12　　　 TcM窗口内最后δ个样本点的能量值;

13　　 end while

14　　　 if(剩余样本点数<250)then

15　　　　output 截取完成的测试集else

16　　　　 N←第M+1个样本点到第M+p个样本点的数据窗口;

17　　　　 TcN窗口内最后δ个样本点的能量值;

18　　　 end else

19　　 end if

20　 end while

4 模型分析

5 主要结论

 [1] 黄贤丞, 唐阳山, 刘光健, 等. 驾驶员换道行为分析. 汽车实用技术, 2017(18): 147-148. [2] 邱小平, 刘亚龙, 马丽娜, 等. 基于贝叶斯网络的车辆换道模型. 交通运输系统工程与信息, 2015, 15(5): 67-73. DOI:10.3969/j.issn.1009-6744.2015.05.010 [3] 马雪兰. 城市路段转向换道模型研究[硕士学位论文]. 合肥: 合肥工业大学, 2017. [4] 王畅. 基于隐形马尔科夫模型的驾驶员意图辨识方法研究[硕士学位论文]. 长春: 吉林大学, 2011 [5] Wu BF, Chen WH, Chang CW, et al. A new vehicle detection with distance estimation for lane change warning systems. Proceedings of the IEEE Intelligent Vehicles Symposium. Istanbul, Turkey. 2007. 698–703. [6] Bhatt A, Dave V, Panchamia Y, et al. Analyzing behavioral attributes of drivers and implementing safe driving model. Proceedings of 2017 IEEE International Conference on Vehicular Electronics and Safety (ICVES). Vienna, Austria. 2017. 228–232. [7] 杨龙海, 罗沂, 徐洪. 基于GPS定位数据的高速公路换道特征分析与行为识别. 北京交通大学学报, 2017, 41(3): 39-46. DOI:10.3969/j.issn.1672-8106.2017.03.005 [8] Li KQ, Wang X, Xu YC, et al. Lane changing intention recognition based on speech recognition models. Transportation Research Part C: Emerging Technologies, 2016, 69: 497-514. DOI:10.1016/j.trc.2015.11.007 [9] Cao K, Yu SY, Yu SW. Maneuver behavior identification and prediction for vehicles based on multi-hidden Markov model. Information and Control, 2014, 34(4): 506-512. [10] Leonhardt V, Wanielik G. Neural network for lane change prediction assessing driving situation, driver behavior and vehicle movement. Proceedings of the 20th IEEE International Conference on Intelligent Transportation Systems. Yokohama, Japan. 2017. 1–6. [11] Zhu LL, Liu L, Zhao XP, et al. Driver behavior recognition based on support vector machine. Journal of Transportation Systems Engineering and Information Technology, 2017, 17(1): 91-97. [12] Moridpour S, Sarvi M, Rose G. Modeling the lane-changing execution of multiclass vehicles under heavy traffic conditions. Transportation Research Record: Journal of the Transportation Research Board, 2010, 2161(1): 11-19. DOI:10.3141/2161-02 [13] Yang D, Zhu LL, Yang F, et al. Modeling and analysis of lateral driver behavior in lane-changing execution. Transportation Research Record: Journal of the Transportation Research Board, 2015, 2490: 127-137. DOI:10.3141/2490-14 [14] Zhao FK, Zhao YB, Tian YT. Driver lane changing intention recognition based on multi-class information. Proceedings of 2018 Chinese Control and Decision Conference (CCDC). Shenyang, China. 2018. 3439–3444. [15] Chen RB, Ye XE, Zhang FY, et al. Research on recognition method of vehicle lane-change behavior based on video image. Proceedings of 2015 International Conference on Applied Science and Engineering Innovation. Jinan, China. 2015, 12. 5–10