【作者】 郭纯宏；

【导师】 宋凯；

【作者基本信息】 沈阳理工大学 ， 通信与信息系统， 2011， 硕士

【摘要】 根据统计分析,疲劳驾驶是引发交通事故的主要原因之一,因此及时有效地检测出驾驶员的疲劳状态,以减少此类交通事故的发生,有着重要的现实意义。本文通过CCD摄像头实时采集驾驶员的头部图像,设计了一种基于TMS320DM642 DSP芯片的人眼信息检测系统,用于检测驾驶员的疲劳状态。本文重点对检测系统中的人眼定位、人眼跟踪和人眼状态识别模块及其算法在DSP中的实现作了研究。首先提出了一种基于blob分析的肤色人脸检测方法,这种算法能够有效去除彩色图像中的类肤色噪声;接着进行人眼定位,采用了基于灰度投影法和改进型Hough变换法实现人眼的精确定位;在跟踪算法上,采用了虹膜跟踪和基于Kalman预测的模板匹配法相结合的方法实现眼睛的实时跟踪;在接下来的人眼状态识别中,提出了基于Gabor变换提取眼睛特征和支持向量机相结合的方法;最后综合PERCLOS和AECS参数对驾驶员的疲劳度进行判别。在理论研究的基础之上,基于SEED-DM642开发平台,在CCS环境下采用C语言编程实现人眼信息检测系统并且调试完成。实验结果表明,该系统能够快速、准确地定位人眼并且识别人眼状态,从而有效地判定出驾驶员的疲劳状态。更多还原

【Abstract】 According to statistical analysis,fatigue driving is one of the main reasons that caused traffic accidents, so detecting driver fatigue timely and effectively and reduce the traffic accident caused by driver fatigue has important practical significance. Real-time collecting driver’s images through CCD camera, this paper designs a detection system of eye information based on the TMS320DM642 DSP chip. The system is used to detect driver fatigue.This paper mainly research on the eye location, eye tracking and eye state recognition in detection system and theirs algorithm implementation in the DSP. First skin color face detection algorithm based on blob analysis was proposed, which can remove the background noise whose color is similar to human skin effectively; Then to locate eye, locating eyes accurately using gray projection algorithm and modified Hough transform; In the eye tracking, using iris tracking and the template matching algorithm based on Kalman prediction to tracking eyes real-time; In the eye state recognition, the method of combining Gabor transform extracting eye characteristic and SVM was proposed to recognizing eye state, open or closed; finally using PERCLOS and AECS parameters to detect driver fatigue.This paper use C language to program and debug in the CCS environment based on the theoretical study and SEED-DM642 system development platform. The experiment results show that the system can locate the eye and recognize eye state quickly and accurately, so it can detect drowsiness effectively.更多还原