因子分析与多层神经网络组合的酒驾辨识模型研究

发布时间：2018-03-02 16:23

  本文选题：酒后驾驶　切入点：驾驶行为　出处：《中国安全科学学报》2017年07期 　论文类型：期刊论文

【摘要】：为准确辨识驾驶员酒驾行为以及酒驾状态水平,提高酒驾治理效率,通过人因工程试验和驾驶模拟试验,采集并预处理驾驶员在正常、饮酒、醉酒3种驾驶状态下的驾驶行为数据(包括驾驶员的人、车、环境数据);对原始参数进行因子分析,提取特征参数并将其作为多层神经网络的输入向量,训练多层神经网络,建立基于因子分析和多层神经网络的酒驾行为辨识模型;选取75组测试样本数据输入模型,将模型的输出结果与实际情况比较,验证模型的有效性。研究表明:该模型的训练时间为0.905 s,最优验证均方误差(MSE)为0.034,识别准确率达92.41%,用该模型能较为快速、准确地识别酒后驾驶行为。
[Abstract]:In order to accurately identify the driver's behavior and the level of drinking driving, and to improve the efficiency of drinking driving, the drivers were collected and pretreated by human engineering test and driving simulation test. Driving behavior data (including driver's person, vehicle, environment data) under three driving states, factor analysis of original parameters, extracting characteristic parameters and using them as input vectors of multi-layer neural network, training multi-layer neural network, Based on factor analysis and multi-layer neural network, the identification model of drinking driving behavior is established, 75 groups of test sample data input model are selected, and the output results of the model are compared with the actual situation. The results show that the training time of the model is 0.905 s, the optimal mean square error (MSE) is 0.034, and the recognition accuracy is 92.41%. The model can be used to identify drunk driving behavior quickly and accurately.
【作者单位】： 山东理工大学交通与车辆工程学院;
【基金】：国家自然科学基金资助(61573009) 山东省自然科学基金资助(ZR2014FM027) 山东省高等学校科技计划(J15LB07) 汽车安全与节能国家重点实验室开放基金资助(KF16232)
【分类号】：U492.8
，

本文编号：1557272