基于视觉通道的疲劳驾驶检测

发布时间：2018-04-11 17:44

本文选题：疲劳驾驶 + 人脸检测　；参考：《南昌大学》2016年硕士论文

【摘要】：在机动车辆日增的今天,车辆成为我们必不可少的代步工具,随之而来的交通事故也是不可避免的出现。而疲劳驾驶已被确认为交通事故中的“头号公路杀手”。因此,判断驾驶员是否疲劳驾驶并报警提示是非常有意义的课题。文章提取了眼睛和嘴巴的特征参数进行疲劳状态检测。论文的主要工作有：首先,人脸检测。文章选用不受人脸表情、姿势、角度影响的HSV色彩空间快速定位人脸,得到人脸图像的初步定位。同时利用肤色模型检测到人脸区域为ASM提供初始位置,解决了ASM初始定位难的问题。其次,基于ASM算法进行人眼和嘴巴跟踪获得眼睛与嘴巴区域。为了提高面部关键部位的精确度,文章改进了AAM算法,采用48点模型,将HSV和局部AAM算法结合,降低了搜索时间同时增强了人脸定位鲁棒性。再次,进行眼睛状态的判断。为了更加精确的判断眼睛状态,选择了Canny算子提取眼睛轮廓。文章选择眼睛上眼睑到内外眼角水平线段的距离作为判断眼睛闭合程度的依据,计算PERCLOS值,判断眼睛是否疲劳。然后对嘴巴状态进行判断。改进的局部AAM采用19点的嘴巴模型,提供出较为精确的嘴巴区域,足够进行嘴巴张开度的判断,计算PMECLOS值。最后,文章提出了融合眼睛疲劳参数和嘴巴疲劳参数,并且给出了驾驶员精神评估模型。将单一疲劳特征参数变为双疲劳参数,提高了疲劳检测的准确度,并得到较好的检测效果。
[Abstract]:With the increasing number of motor vehicles, vehicles have become an indispensable means of transportation, and traffic accidents are inevitable.Fatigue driving has been recognized as the top road killer in traffic accidents.Therefore, it is very meaningful to judge whether the driver is tired and alert.The characteristic parameters of eyes and mouth are extracted for fatigue state detection.The main work of this paper is as follows: first, face detection.In this paper, HSV color space, which is not affected by facial expression, pose and angle, is used to locate the face quickly, and the initial location of the face image is obtained.At the same time, the skin color model is used to detect the initial position of ASM, which solves the problem of initial location of ASM.Secondly, eye and mouth regions are obtained by human eye and mouth tracking based on ASM algorithm.In order to improve the accuracy of the key parts of the face, this paper improves the AAM algorithm, adopts a 48-point model, combines the HSV algorithm with the local AAM algorithm, reduces the search time and enhances the robustness of face location.Third, judge the state of the eyes.In order to judge the eye state more accurately, the Canny operator is chosen to extract the eye contour.The distance from the upper eyelid to the horizontal line of the angle of the eye was chosen as the basis for judging the degree of eye closure. The PERCLOS value was calculated to determine whether the eyes were tired or not.Then the state of the mouth is judged.The improved local AAM uses the 19:00 mouth model to provide a more accurate mouth area, which is sufficient to judge the opening of the mouth and calculate the PMECLOS value.Finally, this paper proposes the fusion of eye fatigue parameters and mouth fatigue parameters, and gives the driver spirit evaluation model.The single fatigue characteristic parameter is changed into double fatigue parameter, the accuracy of fatigue detection is improved, and a better detection effect is obtained.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U463.6;TP391.41

【参考文献】