中文摘要：

为了有效解决驾驶员过度疲劳引发交通事故的问题，提出了一种基于KLT（kanade—lucas—tomasi）算法的驾驶员视觉疲劳检测算法。首先在驾驶员眼睛周围选取若干个特征点，用KLT算法对这些特征点进行跟踪，从而来定位人眼，然后将定位到的人眼与模板进行比较，采用阈值的方法，计算二者的相似度，从而判断眼睛的状态。针对该方法受光照影响较大的问题，又提出背景模型对该算法作了较大改进。通过对采集到的视频流的处理，结果表明该方法有较好的实用性。

英文摘要：

To deal with the problem that overtired drivers cause traffic accidents, a KLT-based scheme for driver fatigue detection is presented. Firstly, several key points around one eye of a person are manually chosen. The KLT algorithm is used to track the points so as to locate the eye. Secondly, the tracked eye is compared with the eye served as the similarity measure template, then their similarity value is calculated by the threshold value method. So the situation of eyes is judged. In order to tackle the lighting disturblance, the background modelling technique is incorporated in the proposed scheme. Finally, byprocessingoftheacquiredvideo, the practicability of the presented method is demonstrated.