论文部分内容阅读
为提高汽车驾驶员注视行为的安全性,应用视频检测技术构建驾驶员不良注视特征评价体系。基于驾驶过程中多作业工况的影响以及驾驶员个体注视行为特征的多元性,提出以驾驶员2个瞳孔间的距离作为表征驾驶员左右注视的参量,以嘴巴到双眼连线中点间的距离作为表征驾驶员俯仰注视的参量,二者合成“T”特征信息。以虹膜-巩膜比例与位置特征信息表征驾驶员的斜视行为,同时结合眼睛闭合度共同表征驾驶员不良注视特征的参量集合。采用多分类支持向量机(SVM)技术,对驾驶员不良注视行为进行模式分类与融合评价。结果表明,基于这3个参量对驾驶员注视行为进行融合判断,能够甄别其不良注视行为。
In order to improve the safety of car driver’s gaze behavior, video detection technology is used to construct driver’s gaze evaluation system. Based on the influence of multi-homework and driver’s individual gaze behavior characteristics, the distance between driver’s two pupils is proposed as the parameter to characterize the driver’s left-right gaze, Distance as a parameter that characterizes the driver’s pitch gaze, the two synthesize the “T” characteristic information. The iris - sclera ratio and location feature information were used to characterize the driver ’s strabismus behavior. At the same time, the parameter set of driver’ s poor gaze characteristics was jointly characterized by eye closure. Adopting multi-class support vector machine (SVM) technology, the driver’s bad gaze behavior was classified and fused. The results show that the driver’s gaze behavior can be identified based on these three parameters, which can identify the bad gaze behavior.