基于视觉的汽车驾驶员疲劳状况检测装置的研究
|
摘要
疲劳驾驶是导致交通事故的一个主要原因,有资料显示:驾驶在疲劳状态下的事故发生率与正常状态下相比是呈非线性上升趋势。因疲劳驾驶而产生的问题己经引起世人的关注,国内外的研究人员根据疲劳时驾驶员身体和驾驶状态发生的一系列指标变化已经研究出一些检测方法和装置,取得一定的进展,但市场上尚无成功的产品应用于实际中。故如何积极研究出能适应汽车技术发展的驾驶员疲劳状况检测装置具有重要的意义。
通过查阅国内外的文献,对目前已出现的疲劳驾驶检测方法有所了解。通过对比后,文中采用PERCLOS法作为该装置的检测方法,并在系统中加入近红外光源以满足应用环境的特殊性。利用人眼对近红外光的生理反应,采集前后瞳孔亮暗程度不同的两幅图像,通过差分处理,直接将目标区域定于人眼,避免了传统方法中找人脸定人眼的复杂过程。再经过顶帽变换、膨胀腐蚀等图像分析处理手段,结合人眼的几何特征找到眼睛区域,并在人眼区域内提取眼睛的特征参数,判断眼睛的睁闭状态,计算出PERCLOS值,由此判断驾驶员是否疲劳。
文中,整个检测装置是在以TI公司TMS320DM6437为核心的开发平台上进行的。DSP控制近红外照明系统,使CMOS摄像机实时采集符合要求的图像,通过视频解码芯片TVP5146把该模拟图像数字化,并存储在外部同步动态存储器SDRAM中,一旦主处理器DM6437判断驾驶员处于疲劳状态,则驱动报警电路,完成疲劳检测。该装置按处理过程可以分为四个模块,分别是图像采集模块、人眼定位模块、特征参数提取模块、疲劳分析及报警模块。
本文首先介绍了该装置的整体设计,然后具体介绍了各个模块的设计细节,程序的嵌入以及系统的整体调试。实验证明,整个系统的设计方案是可行的,研究的人眼定位和特征提取算法有较好的准确性。
For a long time, fatigue drive is a main reason of traffic accidents, and some investigateions indicate that traffic accident frequency will increase out of proportion in fatigue state.The problem of fatigue driving has drawn worldwide people’s attention. Researchers at home and abroad have developed some detection methods and devices ,according to a series of indicators’changes about the body and the driving state when the car dirver is fatigue. So far, the reseachers have made some progresses, but there is no successful product used in practice on the market. In a word, to develop the fatigue detecting device which is adapt to the development of automobile technology is great significant.
Through the domestic and foreign literatures, this author has learned the fatigue detections which have emerged until now. By contrast, this paper uses PERCLOS method as the device detection method, and the system is added near-infrared light to meet the special application environment.Using the physiological responses of human eye to the near-infrared light, the device collected the images that before and after ones which have different levels of bright and dark pupil.Through the differential treatment, the target region will be held directly on the human eye, avoiding the traditional methods that finding face first and then determining eyes, which is a complex process. After top-hat transformation, expansion, corrosion and other image processing tools, combining with the geometric characteristics of human eye, the system finds the eye region, and extracts the eyes’characteristic parameters, calculates PERCLOS value, thus to determines whether the car driver is fatigue.
In this paper, the whole device is on the development platform which core is TI’s TMS320DM6437. DSP controls the near-infrared light, and CMOS camera real-time collects images. Through the video decoder chip TVP5146, the analog images changed into digital format and stored in SDRAM, once the main processor DM6437 determines the driver is fatigue, and then the alarm will ring, which is all of the whole fatigue testing. According to the process, the device can be divided into four modules including image acquisition module, the human eye positioning module, feature extraction module, fatigue analysis and alarm module.
This paper first introduces the whole design of the equipment and presents the technological details of every module, program as well as the debug of the whole system. Experiments show that the system design is feasible, moreover the human eye positioning and feature extraction algorithms studied in this system have good accuracy.
通过查阅国内外的文献,对目前已出现的疲劳驾驶检测方法有所了解。通过对比后,文中采用PERCLOS法作为该装置的检测方法,并在系统中加入近红外光源以满足应用环境的特殊性。利用人眼对近红外光的生理反应,采集前后瞳孔亮暗程度不同的两幅图像,通过差分处理,直接将目标区域定于人眼,避免了传统方法中找人脸定人眼的复杂过程。再经过顶帽变换、膨胀腐蚀等图像分析处理手段,结合人眼的几何特征找到眼睛区域,并在人眼区域内提取眼睛的特征参数,判断眼睛的睁闭状态,计算出PERCLOS值,由此判断驾驶员是否疲劳。
文中,整个检测装置是在以TI公司TMS320DM6437为核心的开发平台上进行的。DSP控制近红外照明系统,使CMOS摄像机实时采集符合要求的图像,通过视频解码芯片TVP5146把该模拟图像数字化,并存储在外部同步动态存储器SDRAM中,一旦主处理器DM6437判断驾驶员处于疲劳状态,则驱动报警电路,完成疲劳检测。该装置按处理过程可以分为四个模块,分别是图像采集模块、人眼定位模块、特征参数提取模块、疲劳分析及报警模块。
本文首先介绍了该装置的整体设计,然后具体介绍了各个模块的设计细节,程序的嵌入以及系统的整体调试。实验证明,整个系统的设计方案是可行的,研究的人眼定位和特征提取算法有较好的准确性。
For a long time, fatigue drive is a main reason of traffic accidents, and some investigateions indicate that traffic accident frequency will increase out of proportion in fatigue state.The problem of fatigue driving has drawn worldwide people’s attention. Researchers at home and abroad have developed some detection methods and devices ,according to a series of indicators’changes about the body and the driving state when the car dirver is fatigue. So far, the reseachers have made some progresses, but there is no successful product used in practice on the market. In a word, to develop the fatigue detecting device which is adapt to the development of automobile technology is great significant.
Through the domestic and foreign literatures, this author has learned the fatigue detections which have emerged until now. By contrast, this paper uses PERCLOS method as the device detection method, and the system is added near-infrared light to meet the special application environment.Using the physiological responses of human eye to the near-infrared light, the device collected the images that before and after ones which have different levels of bright and dark pupil.Through the differential treatment, the target region will be held directly on the human eye, avoiding the traditional methods that finding face first and then determining eyes, which is a complex process. After top-hat transformation, expansion, corrosion and other image processing tools, combining with the geometric characteristics of human eye, the system finds the eye region, and extracts the eyes’characteristic parameters, calculates PERCLOS value, thus to determines whether the car driver is fatigue.
In this paper, the whole device is on the development platform which core is TI’s TMS320DM6437. DSP controls the near-infrared light, and CMOS camera real-time collects images. Through the video decoder chip TVP5146, the analog images changed into digital format and stored in SDRAM, once the main processor DM6437 determines the driver is fatigue, and then the alarm will ring, which is all of the whole fatigue testing. According to the process, the device can be divided into four modules including image acquisition module, the human eye positioning module, feature extraction module, fatigue analysis and alarm module.
This paper first introduces the whole design of the equipment and presents the technological details of every module, program as well as the debug of the whole system. Experiments show that the system design is feasible, moreover the human eye positioning and feature extraction algorithms studied in this system have good accuracy.
引文
[1]李志春,何仁,林谋有,李佩林.驾驶员疲劳检测技术的研究现状及发展趋势.农业化研究,2006,5:197-199.
[2]李贞,冯晓毅.基于传感器技术的驾驶疲劳检测方法综述.测控技术,2007,26(4):1-3.
[3]龚文.疲劳驾驶预警系统的研究:(硕士学位论文).上海:华东师范大学,2006.
[4]王磊.基于计算机视觉的驾驶员疲劳/瞌睡检测方法的研究:(硕士学位论文).山东:山东大学,2005.
[5]周玉彬,俞梦孙等.驾驶员瞌睡的视频检测研究.北京生物医学工程,2003,22(1):30-33.
[6]杨渝书,姚振强,李增勇等.心电图时频域指标在驾驶疲劳评价中的有效性研究.机械设计与制造,2002,5:94-95.
[7]翁茂荣,李强,花彩霞.机动车驾驶员疲劳检测系统的研究现状及发展趋势.浙江工贸职业技术学院学报,2006,6(1):52-56.
[8] N.Hernandez-Gress and D.Esteve. Driver drowsiness detection past, present and prospective work. Esteve. Traffic technology international. June/July 1997.
[9] Anneke Heitmann, Rainer Guttkuhn et al. Technologies for the Monitoring and Prevention of Driver Fatigue. Proceedings of the First International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design :81-86.
[10] W.Wierwille et al. Research on vehicle-based driver status/performance monitoring: development, validation, and refinement of algorithms for detection of driver drowsiness. Washington, DC, National Highway Traffic Safety Administration, 1994.
[11] Ellen M Ayoob, et al.. Copilot. http://www.ri.cmu.edu/projects/project_435.html.
[12] A Multi-Sensory Vision System for Monitoring Driver’s Vigilance, Active Grants and Projects, http://www.cs.unr.edu/~zhu_z/Project/project.html.
[13]钱录庆.日本发明防止驾驶员打瞌睡系统.南方农机,2004,5:31-34.
[14]郑培,宋正河,周一鸣.基于PERCLOS的机动车驾驶员驾驶疲劳的识别算法.中国农业大学学报,2002,7(2):104-109.
[15]李增勇,焦昆,陈铭.汽车驾驶员模拟精神负荷与心率变异性的相关性分析.北京生物医学工程,2002,3:49-51.
[16]刘志强.基于机器视觉的驾驶疲劳检测方法.应用研究,2006,2,35(3):64.
[17]葛如海,陈彦博,刘志强.基于计算机视觉的驾驶疲劳识别方法的研究.中国安全科学学报,2006,16(9):134-138.
[18]王光娟,詹永照,刘志强.基于DM642的嵌入式疲劳驾驶监测系统的实现.计算机应用,2007,27(10):2612-2614.
[19]吴林托,李芳,阙步军.基于红外LED的人眼快速定位与跟踪的研究.光学仪器,2006,28(1):85-88.
[20]田娥,莫易敏,廖张华.基于红外光源的驾驶员眼睛实时监测.计算机工程,2007,33(7):225-257.
[21]袁翔,黄博学,夏晶晶.疲劳驾驶检测方法研究现状.公路与汽运,2007,3:51-54.
[22]张广军.机器视觉.北京:科学出版社,2005.
[23]陈春波,王元庆.基于红眼效应的人眼探测.计算机工程与应用,2007,43(9):198-222.
[24]顾聚兴.近红外成像与热成像.红外,2005,12:40-42.
[25]金宝智.图像传感器CCD与CMOS的对比.现代电视技术.2005,5:80-81.
[26] Eastman Kodak Company,Image Sensor Solutions Terry Guy.机器视觉系统用图像传感器.电子产品世界.2004年7月下半月.
[27] TVP5146 Data Manual. http://www.ti.com. 2007.
[28] TMS320DM6437 Digital Media Processor. http://www.ti.com.2006.
[29]舒梅,董秀成.基于肤色和模板匹配的人眼定位.计算机工程与应用,2009,45(02):237-239.
[30]基于Hough变换圆检测的人眼精确定位方法.计算机工程与应用,2005,41(27):43-44.
[31]孙大瑞,吴乐南.基于Gabor变换的人眼定位.电路与系统学报,2001,6(4):29-32.
[32]闵祥龙,王江安,吴荣华.基于形态滤波与帧间差分的红外点目标检测算法.激光与红外,2008,38(7):720-729.
[33]数字图象处理(MATLAB版).北京:电子工业出版社,2005.
[34]曲培树,董文会.基于眼睑曲率及模糊逻辑的人眼状态识别.计算机工程与科学,2007,29(8):50-51.
[35]江水郎.基于视觉的驾驶员疲劳检测研究:(硕士学位论文).上海:上海交通大学,2008.
[36]李英,赖剑煌,阮邦志.眼睛状态检测的组合方法.中国图象图形学报,2003,10(8):1140-1145.
[37] Laurence H Tim H, Nick M. Review of fatigue detection and prediction technologies. Western Australia: National Road Transport Commission, 2000.
[38]刘志强,汪旸.基于机器视觉的驾驶疲劳检测方法.中国制造业信息化,2006,35(3):63-66.
[39]韩相军,关永,王雪立.基于DSP的疲劳驾驶实时监测系统研究.计算机技术与发展,2006,16(2):47-49.
[40]高永萍,秦华标.驾驶员疲劳检测系统.仪表技术与传感器,2007,(01):60-62.
[41] TMS320DM643x DMP General-Purpose Input/Output(GPIO). http://www.ti.com.2006.
[42] TMS320DM643x DMP Video Processing Front End (VPFE). http://www.ti.com.2007.
[43] TMS320DM643x DMP Video Processing Back End (VPBE). http://www.ti.com.2007.
[44] TMS320DM643x DMP Inter-Integrated Circuit (I2C) Peripheral. http:// www.ti.com.2007.
[45]彭启琼,管庆等.DSP集成开发环境一CCS及DSP/BIOS的原理与应用.北京:电子工业出版社,2004:176-200.
[46]张勇.C/C++语言硬件程序设计.西安:西安电子科技大学出版社,2003.
[47] Texas Instruments Incorporated著.徐盛,胡剑凌编译. TMS320 DSP算法标准.北京:清华大学出版社,2007.
[2]李贞,冯晓毅.基于传感器技术的驾驶疲劳检测方法综述.测控技术,2007,26(4):1-3.
[3]龚文.疲劳驾驶预警系统的研究:(硕士学位论文).上海:华东师范大学,2006.
[4]王磊.基于计算机视觉的驾驶员疲劳/瞌睡检测方法的研究:(硕士学位论文).山东:山东大学,2005.
[5]周玉彬,俞梦孙等.驾驶员瞌睡的视频检测研究.北京生物医学工程,2003,22(1):30-33.
[6]杨渝书,姚振强,李增勇等.心电图时频域指标在驾驶疲劳评价中的有效性研究.机械设计与制造,2002,5:94-95.
[7]翁茂荣,李强,花彩霞.机动车驾驶员疲劳检测系统的研究现状及发展趋势.浙江工贸职业技术学院学报,2006,6(1):52-56.
[8] N.Hernandez-Gress and D.Esteve. Driver drowsiness detection past, present and prospective work. Esteve. Traffic technology international. June/July 1997.
[9] Anneke Heitmann, Rainer Guttkuhn et al. Technologies for the Monitoring and Prevention of Driver Fatigue. Proceedings of the First International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design :81-86.
[10] W.Wierwille et al. Research on vehicle-based driver status/performance monitoring: development, validation, and refinement of algorithms for detection of driver drowsiness. Washington, DC, National Highway Traffic Safety Administration, 1994.
[11] Ellen M Ayoob, et al.. Copilot. http://www.ri.cmu.edu/projects/project_435.html.
[12] A Multi-Sensory Vision System for Monitoring Driver’s Vigilance, Active Grants and Projects, http://www.cs.unr.edu/~zhu_z/Project/project.html.
[13]钱录庆.日本发明防止驾驶员打瞌睡系统.南方农机,2004,5:31-34.
[14]郑培,宋正河,周一鸣.基于PERCLOS的机动车驾驶员驾驶疲劳的识别算法.中国农业大学学报,2002,7(2):104-109.
[15]李增勇,焦昆,陈铭.汽车驾驶员模拟精神负荷与心率变异性的相关性分析.北京生物医学工程,2002,3:49-51.
[16]刘志强.基于机器视觉的驾驶疲劳检测方法.应用研究,2006,2,35(3):64.
[17]葛如海,陈彦博,刘志强.基于计算机视觉的驾驶疲劳识别方法的研究.中国安全科学学报,2006,16(9):134-138.
[18]王光娟,詹永照,刘志强.基于DM642的嵌入式疲劳驾驶监测系统的实现.计算机应用,2007,27(10):2612-2614.
[19]吴林托,李芳,阙步军.基于红外LED的人眼快速定位与跟踪的研究.光学仪器,2006,28(1):85-88.
[20]田娥,莫易敏,廖张华.基于红外光源的驾驶员眼睛实时监测.计算机工程,2007,33(7):225-257.
[21]袁翔,黄博学,夏晶晶.疲劳驾驶检测方法研究现状.公路与汽运,2007,3:51-54.
[22]张广军.机器视觉.北京:科学出版社,2005.
[23]陈春波,王元庆.基于红眼效应的人眼探测.计算机工程与应用,2007,43(9):198-222.
[24]顾聚兴.近红外成像与热成像.红外,2005,12:40-42.
[25]金宝智.图像传感器CCD与CMOS的对比.现代电视技术.2005,5:80-81.
[26] Eastman Kodak Company,Image Sensor Solutions Terry Guy.机器视觉系统用图像传感器.电子产品世界.2004年7月下半月.
[27] TVP5146 Data Manual. http://www.ti.com. 2007.
[28] TMS320DM6437 Digital Media Processor. http://www.ti.com.2006.
[29]舒梅,董秀成.基于肤色和模板匹配的人眼定位.计算机工程与应用,2009,45(02):237-239.
[30]基于Hough变换圆检测的人眼精确定位方法.计算机工程与应用,2005,41(27):43-44.
[31]孙大瑞,吴乐南.基于Gabor变换的人眼定位.电路与系统学报,2001,6(4):29-32.
[32]闵祥龙,王江安,吴荣华.基于形态滤波与帧间差分的红外点目标检测算法.激光与红外,2008,38(7):720-729.
[33]数字图象处理(MATLAB版).北京:电子工业出版社,2005.
[34]曲培树,董文会.基于眼睑曲率及模糊逻辑的人眼状态识别.计算机工程与科学,2007,29(8):50-51.
[35]江水郎.基于视觉的驾驶员疲劳检测研究:(硕士学位论文).上海:上海交通大学,2008.
[36]李英,赖剑煌,阮邦志.眼睛状态检测的组合方法.中国图象图形学报,2003,10(8):1140-1145.
[37] Laurence H Tim H, Nick M. Review of fatigue detection and prediction technologies. Western Australia: National Road Transport Commission, 2000.
[38]刘志强,汪旸.基于机器视觉的驾驶疲劳检测方法.中国制造业信息化,2006,35(3):63-66.
[39]韩相军,关永,王雪立.基于DSP的疲劳驾驶实时监测系统研究.计算机技术与发展,2006,16(2):47-49.
[40]高永萍,秦华标.驾驶员疲劳检测系统.仪表技术与传感器,2007,(01):60-62.
[41] TMS320DM643x DMP General-Purpose Input/Output(GPIO). http://www.ti.com.2006.
[42] TMS320DM643x DMP Video Processing Front End (VPFE). http://www.ti.com.2007.
[43] TMS320DM643x DMP Video Processing Back End (VPBE). http://www.ti.com.2007.
[44] TMS320DM643x DMP Inter-Integrated Circuit (I2C) Peripheral. http:// www.ti.com.2007.
[45]彭启琼,管庆等.DSP集成开发环境一CCS及DSP/BIOS的原理与应用.北京:电子工业出版社,2004:176-200.
[46]张勇.C/C++语言硬件程序设计.西安:西安电子科技大学出版社,2003.
[47] Texas Instruments Incorporated著.徐盛,胡剑凌编译. TMS320 DSP算法标准.北京:清华大学出版社,2007.