高速公路驾驶员状态辨识与防撞预警系统的研究

发布时间：2018-07-31 12:31

【摘要】：近十年来我国高速交通网络得到快速的发展,伴随着人们出行便利的同时,由交通事故带来的安全隐患也变的越来越大。据美国国家高速公路安全专家对交通事故的调查分析,90%以上的事故都是驾驶员自身因素造成,而因车辆故障造成的交通事故仅占3%左右。通过对我国高速公路交通事故数据统计,由车辆与车辆之间造成的事故占高速公路事故总量的50%以上,车辆追尾造成的事故约占车与车事故的35%,其中追尾碰撞的主要原因是由驾驶员疏忽造成的。特别是在高速公路上,驾驶员一点点小失误可能就会给自己和他人带来沉重的灾难。对驾驶员驾驶行为和驾驶状态的判断和分析以及汽车防追尾预警技术的运用可以有效地降低因驾驶员行为造成的交通事故。本文对驾驶员状态辨识和安全预警模型进行了相关研究,具体内容如下:第一,根据已有的研究成果对驾驶员状态与驾驶员生理信号、驾驶行为及眼部信号之间的关系做了全面的归纳和分析,并确定了不同状态下驾驶员心率信号、转向盘信号及眼部特征的临界值,为驾驶员状态辨识提提供理论基础。第二,人眼检测算法的研究。驾驶员眼部特征的检测直接关系到驾驶员状态的判别,本文提出了一种混合肤色模型、积分投影及Canny边缘检测相结合的驾驶员眼睛快速定位的算法。首先根据人的肤色特征建立了混合肤色模型确定人脸位置,通过垂直积分投影和模板设计确定人脸区域,然后通过水平积分投影和投影曲线优化确定的眼睛范围,最后利用Canny算子对人眼区域进行边缘检测和形态学处理进行眼睛精确定位,通过仿真该算法的准确率达到90%以上。第三,高速公路驾驶员状态的辨识。高速公路单调的驾驶环境下驾驶员很容易疲劳和走神。通过模糊综合评价和DS证据理论对驾驶员眼部特征、驾驶行为和心率变化信息进行融合综合判断驾驶员状态,构建了基于模糊评价-DS证据理论的驾驶员状态的辨识模型并通过实例验证了其有效性。第四,建立了基于驾驶员状态的安全车距模型。首先通过驾驶舱模拟试验,获取不同状态下的驾驶员制动反应时间,并以这些试验数据为基础,根据不同状态下的驾驶员确定相应反应时间,然后通过分析车辆减速过程和前车运行状态,推导出车辆高速运动下安全车距模型并实时修正安全车距模型参数。最后,将修正后模型与典型安全车距模型进行了仿真对比,通过对比得到修正模型的可靠性更高。
[Abstract]:With the rapid development of high-speed transportation network in China in the past ten years, with the convenience of travel, the hidden dangers caused by traffic accidents are becoming more and more serious. According to the investigation and analysis of traffic accidents by national highway safety experts, more than 90% of accidents are caused by drivers' own factors, but only about 3% of traffic accidents are caused by vehicle failures. According to the statistics of expressway traffic accidents in China, the accidents caused by vehicles and vehicles account for more than 50% of the total number of highway accidents. The accidents caused by vehicle rearward account for about 35% of vehicle accidents, among which the main cause of rear-end collision is caused by the driver's negligence. Especially on the freeway, a small error by a driver can cause a heavy disaster for himself and others. The judgment and analysis of driver's driving behavior and driving state and the application of anti-rear-end warning technology can effectively reduce the traffic accidents caused by driver's behavior. This paper has carried on the correlation research to the driver state identification and the safety early warning model, the concrete contents are as follows: first, according to the existing research results, the driver state and the driver physiological signal, The relationship between driving behavior and eye signals is summarized and analyzed, and the critical values of heart rate signal, steering wheel signal and eye characteristics are determined in different states, which provides a theoretical basis for driver state identification. Second, the research of human eye detection algorithm. The detection of driver's eye features is directly related to the identification of driver's state. In this paper, a hybrid skin color model, integral projection and Canny edge detection are proposed to locate the driver's eyes quickly. Firstly, a mixed skin color model is established to determine the position of human face, and the vertical integral projection and template design are used to determine the face area. Then, the eye range is optimized by horizontal integral projection and projection curve. Finally, the Canny operator is used to detect the edge of the human eye region and to locate the eye accurately by morphological processing. The accuracy of the algorithm is over 90% through simulation. Third, the identification of the state of the motorway driver. In the monotonous driving environment of the freeway, the driver is easily tired and distracted. Through fuzzy comprehensive evaluation and DS evidence theory, the driver's eye characteristics, driving behavior and heart rate change information are fused to judge the driver's state. The identification model of driver state based on fuzzy evaluation-DS evidence theory is constructed and its validity is verified by an example. Fourthly, a safe distance model based on driver state is established. First of all, through the cockpit simulation test, the driver braking reaction time in different states is obtained, and based on these test data, the corresponding reaction time is determined according to the driver under different conditions. Then by analyzing the vehicle deceleration process and the running state of the front car, the safe distance model under the high speed motion of the vehicle is derived and the parameters of the safety distance model are revised in real time. Finally, the modified model is compared with the typical safety vehicle distance model, and the reliability of the modified model is higher.
【学位授予单位】：山东理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U463.6

【相似文献】