| 引用本文: | 姜思羽,吴斌,邱少健,羊梅君.驾驶员疲劳检测技术的算法设计与硬件实现[J].哈尔滨工业大学学报,2014,46(5):95.DOI:10.11918/j.issn.0367-6234.2014.05.015 |
| JIANG Siyu,WU Bin,QIU Shaojian,YANG Meijun.Algorithm design and hardware implementation of driver fatigue monitoring[J].Journal of Harbin Institute of Technology,2014,46(5):95.DOI:10.11918/j.issn.0367-6234.2014.05.015 |
|
| 摘要: |
| 为了减少由疲劳驾驶而引起的交通事故,提出了一种基于非接触式的驾驶员疲劳驾驶检测方法．利用摄像头对驾驶员的面部图像进行采集,经过图像前处理,采用在线识别的方法对驾驶员的面部特征进行识别,获取其疲劳状态;硬件采用DSP系列的TMS320DM642作为核心处理器,在判定驾驶员疲劳之后发出报警．实验结果表明,使用脉冲耦合神经网络方法对图像进行增强与在线识别,可以有效地确定人脸与人眼区域,主控芯片TMS320DM642的运算处理能力满足系统的要求,摄像头焦距为8 mm时,系统有效检测距离为30~150 cm．脉冲耦合神经网络方法对于驾驶员疲劳状态检测可靠性较高,合理地选择硬件平台以及系统的安装位置对检测效果有重要的影响． |
| 关键词: 疲劳驾驶 神经网络 训练样本 在线识别 DM642 |
| DOI：10.11918/j.issn.0367-6234.2014.05.015 |
| 分类号:U469.72 |
| 基金项目:广东省自然科学基金资助项目(9451064101003049)； 省部产学研结合资金资助项目(2011B090400085); 花都区产学研结合资金资助项目(11HDCXY-001;11HDZD-007). |
|
| Algorithm design and hardware implementation of driver fatigue monitoring |
|
JIANG Siyu, WU Bin, QIU Shaojian, YANG Meijun
|
|
(Guangzhou College, South China University of Technology, 510800 Guangzhou, China)
|
| Abstract: |
| To reduce the loss of people′s property caused by traffic accidents because of driver fatigue, a new driver fatigue monitoring method based on non-contact is proposed. First, a camera gets driver's facial image, and then by using online method of identifying to identify the driver's facial features after pre-processing, the fatigue state of the driver can be acquired. Hardware system including TMS320DM642 as the core processor will alarm when driver is fatigue. Experimental results show that the use of pulse coupled neural network method to enhance and identify image, can determine the area of face and eye effectively, the master chip TMS320DM642 meets the requirement of detection distance between 30 cm and 150 cm when camera focal length is 8 mm. Coupled pulse neural network method has high reliability in driver fatigue detection, and choosing the right hardware platform and installation position have important influences on the detection results. |
| Key words: fatigue driving neural networks training samples online identify DM642 |
