Development of steering behavior recognition method by using sensing data of drive recorder
Development of steering behavior recognition method by using sensing data of drive recorder
复制标题
DOI：
10.1109/iccas.2008.4694660
复制
发表时间：
2008-12
期刊：
影响因子：
--
通讯作者：
Hideki Tsunai;Kozo Maeda;Ryuzo Hayashi;P. Raksincharoensak;Masao Nagai
中科院分区：
文献类型：
--
作者：
Hideki Tsunai;Kozo Maeda;Ryuzo Hayashi;P. Raksincharoensak;Masao Nagai
提出了利用行车记录仪的传感数据进行车道保持、换道、识别等驾驶行为的方法。减少交通事故，不仅要提高主动安全技术，更要提高驾驶员的驾驶安全意识。提高驾驶员驾驶意识的方法之一是告知驾驶行为的安全程度。为了实现该系统，有必要开发基于行车记录仪数据的驾驶行为识别算法。因此，本文重点研究了车道变更识别方法，提出了基于Boosting框架的序贯标记法；Boosted条件随机场。在算法的开发过程中，重点对模型进行了四个特征的训练：速度、方向盘角度、移动方差和移动标准差。最后给出了识别结果，并对识别算法在机器学习过程中的最佳特征进行了检验。
This paper proposes the steering behavior, e.g. lane keeping, lane changing, recognition method by using sensing data of drive recorder. To reduce traffic accidents, it is necessary to improve not only active safety technology, but also the driver awareness about driving safety. One of the methods to increase the driver awareness is to inform the degree of safety of driving behavior. To realize the system, it is necessary to develop the driving behavior recognition algorithm by sensing data of drive recorder. Therefore, this study focuses on lane change recognition method and develops the algorithm by sequential labeling method based on boosting framework; Boosted Conditional Random Fields. To develop the algorithm, four features are focused here to train the model: those are, velocity, steering wheel angle, moving variance and moving standard deviation. Finally, the recognition results are shown, and the best features in machine learning process for recognition algorithm are examined.