基于面部特征分析的疲劳驾驶检测方法

为避免疲劳驾驶,通过提取面部疲劳特征参数的方法研究了驾驶员疲劳检测技术.对SSD(single shot multi box de-tector)目标检测算法及连续自适应均值漂移跟踪算法(continuously adaptive MeanShift,CamShift)进行优化,以检测人脸区域.利用特征点定位提取面部疲劳特征参数,并基于眼睛闭合时间百分比(percentage of eyelid closure over the pupil over time,PERCLOS)设定疲劳阈值和疲劳检测策略.在实车样本集上进行试验,结果表明:优化的人脸区域定位方法对光线变化、类肤色干扰的鲁棒性更强;所提取的疲劳特征参数能有效反映驾驶员疲劳状态,平均识别准确率达到了92.2％.改进后的算法系统在基于视觉特征的疲劳驾驶检测技术中达到了较高水平,对于预防交通安全事故具有重大意义.

Fatigue Driving Detection System Based on Face Feature Analyze

In order to avoid fatigue driving, the driver fatigue detection technique was studied by extracting the facial fatigue feature parameters. A face detection method combining SSD target detection algorithm and CamShift tracking algorithm was designed. The characteristic parameters of facial fatigue were extracted by using the feature points. Based on PERCLOS criterion, the fatigue threshold and fatigue detection strategy were set. The system detection effect was evaluated on the real vehicle sample set, and the results showed that: the optimized face area localization method has stronger robustness to light change and skin-color interference; the extracted fatigue characteristic parameters can effectively reflect the fatigue state of the driver, with an average recognition accuracy of 92.2%.