摘要
不良驾驶行为识别对于减少交通事故的发生及实现智能交通具有重要意义。针对目前驾驶行为识别精度不高及系统稳定性不好问题,提出了一种基于多特征卷积神经网络和智能手机惯性传感器的新型驾驶行为识别方法。通过获取手机惯性传感器数据,利用多特征卷积神经网络对数据集进行特征提取和分析,来实现对多种驾驶行为的识别。最后,利用实际采集的车载数据进行不同方法试验对比,该算法对驾驶行为的分类精度达到97.14%,在识别精度上有一定优势。此外,MFCNN测试结果的方差仅为0.014391,小于其它3种网络模型,可见所研究的网络模型及方法更加稳定可靠。
The abnormal driving behaviors detection is of great significance to reduce the incidence of traffic accidents and realize intelligent transportation. In view that current driving behaviors detection system is not high enough in accuracy and stability, a novel driving behaviors detection method based on Multi-Feature Convolutional Neural Network (MFCNN) and smart phone inertial sensors is proposed. Based on the data collected via smart phone inertial sensors, the proposed MFCNN is used to extract the features of collected data and recognize the multiple driving behaviors. Finally, field test is conducted to verify the performance of the proposed scheme, which shows that the MFCNN is superior to Random Forest, D3 model and BP neural network. The MFCNN achieves recognition accuracy of 97.14% for driving behaviors. Furthermore, the variance of MFCNN test results is only 0.014 391, which is smaller than that of the other three network models, indicating that the MFCNN is more stable and reliable.
作者
谢非
汪壬甲
沈世斌
孙蕊
张斌
刘锡祥
XIE Fei;WANG Renjia;SHEN Shibin;SUN Rui;ZHANG Bin;LIU Xiyang(School of Electrical and Automation Engineering, Nanjing Normal University, Nanjing 210042, China;Collegeof Civil Aviation, Nanjing University of Aeronautics & Astronautics, Nanjing 210016, China;NARI Technology Co. Ltd., Nanjing 211000, China;College of Instrument Science & Engineering, Southeast University, Nanjing210096, China;Nanjing Industry Institute for Advanced Intelligent Equipment, Nanjing 210042, China)
出处
《中国惯性技术学报》
EI
CSCD
北大核心
2019年第3期288-294,共7页
Journal of Chinese Inertial Technology
基金
国家自然科学基金(61601228,41704022,61803208)
江苏省自然科学基金(BK20161021,BK20170780)
江苏省“六大人才高峰”人才支持计划(2016-DZXX-035)
江苏省高校自然基金(17KJB510031)
关键词
智能交通
惯性传感器
卷积神经网络
驾驶行为
intelligent transportation
inertial sensor
convolutional neural network
driving behaviors
