摘要
与传统基于驾驶员行为的疲劳检测手段相比,基于驾驶员生理指标的驾驶疲劳检测是一种更加客观准确的检测方法,但由于生理信息复杂度高,传统生理指标疲劳检测模型效果不佳且实时性差。随机森林是一种收敛速度快,可处理复杂特征向量样本,高效精准的分类算法。文章在驾驶员生理指标检测基础上,提出一种应用随机森林模型进行疲劳驾驶检测识别的方法,并通过粒子群优化算法和设置阈值修剪错误决策树方式对随机森林模型进行优化,以提高精度和效率。仿真实验结果表明,优化后检测精准确度高达98%,运行效率提高50%。
Compared with the traditional detection methods,the driver's physiological indicators of fatigue driving test is a more objective and accurate detection method.Due to the complexity of physiological information,fatigue testing model of traditional physiological indicators is not good enough to be real-time.Random Forest(RF)is a fast-convergence,efficient and accurate classification algorithm,it can handle complex eigenvector samples.Based on the detection of physiological indicators of pilots,this paper proposes a method to detect fatigue driving based on Random Forest model.To improve the detection accuracy and operational efficiency,set the threshold to prune the error decision tree of the Random Forest,and the Particle swarm optimization(PSO)is also used to optimize the parameters of model.The simulation results show that the optimized detection accuracy reaches 98%and the running speed increases by 50%.
作者
叶建芳
刘强
李雪莹
Ye Jianfang;Liu Qiang;Li Xueying(Engineering Research Center of Digital Textile and Garment Technology,Shanghai 201620;College of Information Science&Technology,Donghua University,Shanghai 201620)
出处
《汽车实用技术》
2018年第13期39-43,共5页
Automobile Applied Technology
基金
东华大学非线性科学研究所基金资助项目(20160905-3)
关键词
疲劳驾驶
生理指标检测
复杂特征向量
随机森林
粒子群优化
fatigue driving
physiological index detection
complex eigenvector
random forest
particle swarm optimization
