定性推理生成器在驾驶疲劳检测中的应用

Application of generator for uncertainty reasoning in fatigue driving inspection

疲劳驾驶是导致交通死亡事故的原因之一。为检测识别驾驶疲劳状态,该文鉴于人的眼动行为存在随机性及模糊性的特点,建立了基于"眼睛闭合时间比"(Per-clos)和眨眼时间均值的二维云模型,根据这2个眼动参数的云模型特征,构建了14条定性规则。依据二维单规则生成器,构造了二维多规则定性推理生成器。对60例数据,进行识别判断。结果表明:平均识别率达到73.98%。同组实验数据下,使用该方法比采用K最近邻(KNN)及支持向量机(SVM)分类算法的检测率要高。随着训练样本人数增多,生成器的识别率预计可进一步提高。

To detect and identify driving fatigue state, which is one of the causes of a traffic death accident; a two-dimensional cloud model was established based on the ＂percentage of eyelid closure over the pupil over time＂ （Per-clos） and blink time mean in view of the randomness and fuzziness of eye movement. Fourteen pieces of qualitative rules were constructed based on the cloud model features of the two parameters. A generator on 2-D and multi-rules was constructed for uncertainty reasoning in fatigue detection based on a 2-D single rule generator to recognize and detect the fatigue state of an experimental data of 60 samples. The analysis results show that the average recognition rate is 73.98%. Therefore, the method has higher detection rate than classification algorithm KNN （k-nearest neighbor） and SVM （support vector machine） under the same experimental data. The recognition rate of the generator can be improved when the number of training samples increases.