基于多特征的驾驶员不安全行为检测的研究

发布时间：2018-08-14 17:08

【摘要】：自中国加入世界贸易组织以来,中国经济得到蓬勃发展,全国的基础设施建设逐渐完善,人民生活水平得到很大提升,私家车的数量逐年递增,从而造成交通事故频繁发生,给人民的生命财产造成重大影响,严重影响了交通运输效率。因此,开发驾驶员的安全行为督导系统有助于加强驾驶人员的安全意识并进行安全驾驶操作,有极高的实用价值和社会意义。针对目前的基于计算机视觉的驾驶员安全行为辅助系统主要集中于驾驶员的精神状态和疲劳驾驶方向,而行车是否安全主要取决于驾驶员的驾驶行为是否符合安全操作规程,因此,本文以驾驶员的驾驶行为本身为切入点,对不安全的驾驶行为进行检测。本文主要的研究内容如下所示:(1)从驾驶员开车过程中的具体行为入手,把驾驶员的手部动作行为分成两类:驾驶员的手部运动趋势和手部在方向盘的位置。通过车辆的转向信息和档位信息来判断驾驶员的手部运动趋势,通过计算机视觉技术来定位驾驶员的手部位置。(2)利用椭圆检测技术来定位采集图像中的方向盘,从而得到感兴趣区域,通过大量统计分析,建立高斯肤色模型,根据得到的高斯肤色模型计算图像中像素点属于肤色区域的概率,通过大津阈值分割把驾驶员的手部从背景中分离出来,为了使肤色检测结果可以应对不同的外部环境,提出了自适应的高斯肤色模型。(3)根据采集得到的驾驶员行为特征,训练驾驶员行为分类决策树来进行驾驶行为识别,并通过大量样本验证方法可行性。
[Abstract]:Since China's entry into the World Trade Organization (WTO), China's economy has developed vigorously, the national infrastructure has been gradually improved, the standard of living of the people has been greatly improved, and the number of private cars has increased year by year, resulting in frequent traffic accidents. The life and property of the people have a major impact, seriously affecting the efficiency of transportation. Therefore, it is of great practical value and social significance to develop the driver's safety behavior supervision system, which is helpful to strengthen the driver's safety consciousness and carry out the safe driving operation. At present, the driver safety behavior assistant system based on computer vision mainly focuses on the driver's mental state and fatigue driving direction, and whether the driving safety depends on whether the driver's driving behavior conforms to the safe operation rules or not. Therefore, in this paper, the driver's driving behavior itself as a starting point, unsafe driving behavior detection. The main contents of this paper are as follows: (1) starting with the driver's specific behavior during driving, the author divides the driver's hand movement behavior into two categories: the driver's hand movement trend and the position of the hand on the steering wheel. The steering information and gear information of the vehicle are used to judge the movement trend of the driver's hand, and the position of the driver's hand is located by the computer vision technology. (2) the steering wheel in the captured image is located by using the elliptical detection technology. In order to get the region of interest, through a lot of statistical analysis, the Gao Si skin color model is established. According to the obtained Gao Si skin color model, the probability of pixels belonging to the skin color region in the image is calculated. The driver's hand is separated from the background by the threshold segmentation of Otsu. In order to make the skin color detection results adapt to different external environments, an adaptive Gao Si skin color model is proposed. (3) according to the driver's behavior characteristics collected, the skin color model is proposed. Driver behavior classification decision tree is trained to recognize driving behavior, and a large number of samples are used to verify the feasibility of the method.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U463.6;TP391.41

【相似文献】