驾驶员疲劳检测中的嘴部状态研究

摘要

According to this characteristic that the mouth is one of the important information point of the face, a study of mouth state on driver fatigue detection is proposed. First it adopts the improved Adaboost algorithm to locate the face region. Then uses the iterative threshold selection method to do the two value processing in the lower half of the face, using Harris corner detection to find the corners of the mouth after get the full profile of the mouth, and corrects oblique mouth. Finally determine whether fatigue through the opening and the duration. Experimental results show that this algorithm has fast speed, and play an important role for further research of fatigue detection.%针对嘴部作为人脸的重要信息集中点之一这个特性，提出了一种用于驾驶员疲劳检测的嘴部状态的研究算法。首先采用改进的Adaboost算法精确定位人脸区域，然后利用迭代式阈值选择法对人脸下半部分进行二值化处理，得到嘴部完整轮廓后使用Harris角点检测找出嘴角，同时矫正倾斜嘴部，最后通过计算嘴部张开度和持续时间来判断是否疲劳。实验表明，该算法具有较快的速度，同时对疲劳检测的进一步研究有重要的作用。