基于多算法融合的驾驶员疲劳状态检测

发布时间：2018-07-01 11:10

本文选题：疲劳检测 + YCbCr　；参考：《郑州大学》2014年硕士论文

【摘要】：交通的快速发展，带来了很多负面效应，其中疲劳驾驶带来的交通事故已经成为一个重要的影响因素。近年来，汽车的设计越来越满足人类对舒适度的要求，疲劳驾驶就更容易发生。为了减少这种事故的发生，找出一种准确的检测疲劳状态并且及时有效的给驾驶员一个警惕成为了一个必要的手段。本文在保证驾驶舒适度的基础上，采用非接触式疲劳检测算法检测出疲劳状态。具体的疲劳检测算法分开来讲，有以下几部分构成：人脸检测部分，在选取了合适的人脸库基础上，根据判断肤色在色彩空间上的特征，选取了基于YCbCr肤色特征的人脸检测算法。首先对肤色进行预处理，然后对得到的处理进行积分投影后最终分离出人脸区域。人眼检测部分，采用自适应边缘特征提取的人眼定位检测算法。首先对提取出人脸图像进行Robert算子自适应边缘特征提取，经过求梯度、计算复杂度等的处理后，基本上可以判断出人的眼睛位置。在此基础上提出对人眼进行二次特征提取，经过再次计算梯度和复杂度处理，得到一个眼睛的定位结果。实验结果也证明了这种方法能够很大程度上改善人眼检测的检测率，并且能够提高检测速度。人眼状态判断部分，本文采用国际上通用的PERCLOS状态识别方法对人眼进行判断。通过计算眼睛面积的大小，与标准里眼睛睁开度比较判断眼睛的睁闭状态。论文中的人脸检测算法、人眼定位检测算法以及疲劳判断算法是在MATLAB图像函数库实现。实验证明改进后的算法对人脸检测和人眼检测率都有了明显的提高，并且在一定的程度上提高了实时性，，所以本文采用的算法是可行的，具有一定的实际意义。
[Abstract]:The rapid development of traffic has brought a lot of negative effects, among which the traffic accident caused by fatigue driving has become an important factor. In recent years, the design of cars more and more meet the requirements of human comfort, fatigue driving is more likely to occur. In order to reduce the occurrence of this kind of accident, it is necessary to find out an accurate detection of fatigue state and to alert the driver timely and effectively. In this paper, on the basis of ensuring driving comfort, non-contact fatigue detection algorithm is used to detect fatigue state. The specific fatigue detection algorithm is divided into the following parts: the face detection part, on the basis of selecting the appropriate face database, judging the color features in color space, A face detection algorithm based on YCbCr color feature is selected. Firstly, the skin color is pretreated, then the face region is separated by integral projection. In the part of human eye detection, adaptive edge feature extraction algorithm is used to detect human eye location. Firstly, the human face image is extracted by Robert operator adaptive edge feature extraction. After processing the gradient and computational complexity, we can basically judge the position of human eyes. On the basis of this, the second feature extraction of human eye is proposed. After calculating the gradient and complexity again, the location result of one eye is obtained. The experimental results also show that this method can greatly improve the detection rate of human eye detection and improve the detection speed. In the part of human eye state judgment, this paper uses the universal Percos state recognition method to judge the human eye. By calculating the size of the eye area, the open state of the eye is judged by comparing it with the standard degree of eye opening. The face detection algorithm, human eye location detection algorithm and fatigue judgment algorithm are realized in MATLAB image function library. The experimental results show that the improved algorithm improves the face detection rate and the human eye detection rate obviously, and improves the real-time performance to a certain extent, so the algorithm adopted in this paper is feasible and has certain practical significance.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.254;U463.6

【参考文献】