【作者】 刘宇；

【导师】 杨杰；

【作者基本信息】 武汉理工大学 ， 信号与信息处理， 2009， 硕士

【摘要】 疲劳驾驶是引发交通事故的主要原因之一,准确、及时检测出驾驶员的驾驶状态并对其做出相应的预警措施,可以减少甚至避免交通事故的发生。本文介绍的疲劳驾驶实时监测系统所完成的主要工作是利用实时信号处理平台,获得驾驶员疲劳状态的实时数据,并对数据进行分析。为了实现数据采样和分析过程的实时性,系统采用了以TI公司TMS320C6416DSP处理芯片为内核的数字信号处理平台作为硬件系统来完成具体算法的处理。系统中,处理器的时钟频率为600MHz,通过彩色CCD摄像头来完成视频信号的采集,将采集到的PAL制式YUV422视频信号输入到处理平台上进行处理,最后将处理结果输出到电脑显示器VGA或数字液晶显示器。疲劳状态判定算法则利用PERCLOS检测方法来具体实现。PERCLOS(percentage of eyelid closure over the public over time)检测是一种基于机器视觉的方法,该方法利用对眼睛状态的监控,完成驾驶员疲劳的检测。具体的软件实现步骤为:先结合运动检测中的帧差法和肤色模板匹配法来对采集到的视频信号完成驾驶员脸部检测和定位;再在以确定了的脸部区域内,利用由积分投影函数和微分投影函数所组成的混合投影算法来确定驾驶员眼睛的位置;然后利用边缘提取算法确定眼睑的坐标,利用上下眼睑间距判断驾驶员眼睛的状态;最后利用嵌入式实时操作系统DSP/BIOS来完成驾驶员在单位时间内眼睛眨动频率的计数和显示。为了能方便控制系统的运行,系统通过PCI接口实现了DSP和PC机的通信。利用PC机上的VC++6.0开发环境,本文设计了一个人控制系统,通过该系统向DSP发送各种消息。DSP在接受到PC端的消息后,通过消息响应函数实现消息的要求,同时向PC机发送一个回馈消息,用以确认PC机发送的消息准确收到和DSP端消息任务的完成。同时,系统通过DMA技术,实现了DSP和PC机之间的实时大批量数据交换,以保存DSP处理数据。本文所介绍的系统准确的定位了驾驶员的脸部,跟踪了驾驶员的眼睛,判断了眼睛状态,借助实时嵌入式操作系统DSP/BIOS完成了眼睑眨动次数的计算和显示,针对记录的眨动次数数据进行了分析。更多还原

【Abstract】 Driver inattention, particularly due to fatigue, is widely considered as one of the primary causes of traffic accident, as a result, detecting out the driver’s fatigue state accurately and timely, at the same giving alarm can reduce traffic accidents.The main task in the system designed in this paper is acquiring and analyze real-time data of driver’s fatigue state. With the purpose of real-time feature in analysis process, the DSP experimental platform in which core processor is TMS320C6416 manufactured by TI Corporation is used as hardware system to realize the processing algorithms. In the system, the processor’s clock frequency is 600MHz, the YUV422 PAL video signal is captured by color CCD, and then sent to DSP experimental platform, at last, the signal processed will be displaied in VGA.The PERCLOS method is adopted as core algorithm in fatigue state detecting. PERCLOS (percentage of eyelid closure over the public over time) detecting is method based on machine vision, this method realizes the fatigue state detecting in the way of monitoring the driver’s eye state. The concrete process is shown as following: firstly, the frame difference method in motion detection and template matching method of skin color are combined in the paper to complete the driver’s face detection and location; secondly, in the area where the face is located, the Hybrid Projection algorithm composed of Integral Projection Function and Differential Projection Function is used to locate the driver’s eyes; thirdly the eyelid position is located by method of edge extraction algorithm, subsequently, the eye state of dirver is judged based on distance between upper eyelid and lower eyelid; finally, the embedded real-time operating system—DSP/BIOS is utilized to realize the calculation and display of driver’s blink frequence in unit time. In this paper, various face and eyes detecting methods are introduced and analyzed. Through comparing the advantage and disadvantage of these methods, this paper illuminates the method choosed detaily.In order to control the system running conveniently, system achieves the communica-tion between DSP and PC through PCI interface. In PC, a control system is designed by use of VC++6.0 developing environment to send diverse messages to DSP. After receiving messages from PC, DSP realizes the requirements of message by message response functions, afterward, sends a feedback message to PC for affirming the receipt of message from PC and informing PC the message task completed. In the meaning time, the massive data exchanging between DSP and PC is implementd by use of DMA technology for saving of data processed in DSP.Through the experimental result, driver’s face and eyes can be located and traced accurately, by virtue of the embedded real-time operating system—DSP/BIOS and the analysis of eye state, the blink of eyelid is calculated and displayed. At last, the system performance is judged by means of analysis of blink time recorded.更多还原