基于ASM和肤色模型的疲劳驾驶检测

疲劳驾驶研究中,面部关键特征精确定位与跟踪是个难点。提出了一种基于主动形状模型ASM和肤色模型的疲劳驾驶检测方法。首先,利用肤色模型检测到人脸区域为ASM提供初始定位;然后基于ASM进行人眼和嘴巴跟踪获得眼睛与嘴巴区域;再利用Canny算子对两个区域精确定位,获得疲劳检测参数;最后根据PERCLOS方法实现疲劳检测。考虑到基于HSV颜色模型的人脸检测不受姿势和角度的影响,但容易受到背景干扰,而ASM的优点是人脸关键点跟踪效果好,但初始定位困难,将二者结合实现了眼睛与嘴巴精确定位与跟踪。实验表明,眼睛检测准确率可以达到90.7%,哈欠检测准确率可以达到83.3%,疲劳检测准确率达到91.4%。

Fatigue driving detection based on ASM and skin color model

Since it is difficult to accurately locate and track facial key features in fatigue driving, we propose a new fatigue driving detection method based on the active shape model (ASM) and the skin color model. Firstly, we use the skin color model to detect the initial localization of the face region for the ASM. Secondly, eyes and mouths are tracked based on the ASM access to eye and mouth area. Thirdly, Canny operator is adopted for accurately locating of the two regions and the fatigue detection parameters are obtained. Finally, the fatigue detection is realized according to the percentage of eyelid closure over the pupil over time (PERCLOS) method. Given that the face detection based on the HSV color model is not influenced by posture and angle but is easy to be interfered by background, while the ASM has a good advantage of face key point tracking effect but has difficulty in the initial location, we combine the two methods which can achieve accurate eyes and mouth location and tracking. Experimental results show that the accuracy of eye detection can reach 90.7%, the accuracy of yawn detection can reach 83.3%, and the accuracy of fatigue detection is 91.4%.