非接触式多源信息融合的疲劳驾驶极限学习机检测方法研究

Fatigue driving is one of the main causes of frequent traffic accidents and thus the research on detection method of fatigue driving is of great significance to improve drivers’ driving safety. The current detecting methods of fatigue driving have some certain limitations in terms of reliability and real-time performance. Facing the driving environment diversity, driver differences, uncertainty of characteristic parameters of fatigue driving, and not easy predictability of fatigue driving, this project is devoted to studying the non-contact fatigue driving evaluation method, which is based on the online LDA theory of the information fusion, regarding the driver fatigue sample data as the object, and setting the hardware-in-loop driving simulator as the experimental test bench. The detailed research contents are: Firstly, according to the information of driver’s physiological state, physiological response and driving behavior in fatigue situation, the detection and extraction methods of effective feature parameters of fatigue state are studied. Secondly, based on the acquisition of the feature parameters of the fatigue states, the method of the online update of the multisource information feature fusion is researched. Thirdly, the online sequential extreme learning machine based detection methodology is established，using the features after fusion. Without affecting the drivers’ driving comfort, not only the accuracy and reliability of the online detection for fatigue driving is enhanced, but also the basic theory research about the fatigue driving detection and its application is well executed.

疲劳驾驶是造成交通事故频繁发生的主要原因之一，疲劳驾驶的有效检测对提高行车安全具有重要意义。现有的疲劳驾驶检测方法在可靠性、实时性等方面存在一定的局限性，本项目针对驾驶环境多样性、驾驶员的差异性、疲劳驾驶状态特征参数的不确定性以及疲劳驾驶状态的不易预测性，以信息融合中在线LDA理论为基础，以驾驶员疲劳样本数据为对象，以模拟驾驶半实物仿真模型为实验手段，开展非接触式疲劳驾驶检测方法研究，内容包括：（1）依据疲劳状态下驾驶员生理状态、生理反应、驾驶行为等信息，研究疲劳状态有效特征参数的检测及提取方法；（2）在获取疲劳状态特征参数的基础上，研究多源信息特征级融合在线更新的方法；（3）结合融合后的特征，研究序贯极限学习机的疲劳驾驶检测方法。在不影响驾驶员驾驶舒适度的同时，提高了疲劳驾驶在线检测的实时性、准确性和可靠性，为疲劳驾驶检测及应用做好基础理论研究。

疲劳驾驶是造成交通事故频繁发生的主要原因之一，疲劳驾驶的有效检测对提高行车安全具有重要意义。现有的疲劳驾驶检测方法在可靠性、实时性等方面存在一定的局限性，本项目针对驾驶环境多样性、驾驶员的差异性、疲劳驾驶状态特征参数的不确定性以及疲劳驾驶状态的不易预测性，以信息融合中在线LDA理论为基础，以驾驶员疲劳样本数据为对象，以模拟驾驶半实物仿真模型为实验手段，开展非接触式疲劳驾驶检测方法研究，内容包括：（1）依据疲劳状态下驾驶员生理状态、生理反应、驾驶行为等信息，研究疲劳状态有效特征参数的检测及提取方法；（2）在获取疲劳状态特征参数的基础上，研究多源信息特征级融合在线更新的方法；（3）结合融合后的特征，研究序贯极限学习机的疲劳驾驶检测方法。在不影响驾驶员驾驶舒适度的同时，提高了疲劳驾驶在线检测的实时性、准确性和可靠性，为疲劳驾驶检测及应用做好基础理论研究。项目在执行期间在国内外重要学术期刊发表了论文21篇，其中SCI收录论文19篇；项目执行期间授权发明专利9项。