驾驶人使用手机行为（DSUB）是“互联网+”时代出现的新型危险驾驶行为，研究表明DSUB可导致人的驾驶负荷增加、认知与判断能力降低、车辆操作水平下降，已迅速成为当今交通事故的主要原因之一。目前尚未出现完整的DSUB识别与预警系统。本项目拟采用双视角图像协同识别的方法构建驾驶人头动、眼动、嘴动与手动模型，提出四类DSUB细分行为的判别逻辑；深度挖掘图像信息，运用模式识别、机器学习等方法识别驾驶人注意力区域与警觉状态，评估DSUB细分行为下驾驶人应激水平的风险级别；对注意力区域与警觉状态的指标集合进行在线迭代更新，建立符合个体行为的应激水平评价档案；根据车速、纵向间距等运行参数评估车辆运行状态的风险级别，建立耦合驾驶人应激水平与车辆运行状态的DSUB智能预警策略，提升预警系统的安全度和可接受度。本项目探讨驾驶人使用手机行为识别与预警的理论与实践，为将其应用于辅助驾驶系统提供新的方法和技术支撑。

In “Internet Plus” era, driver smartphone use behavior (DSUB) becomes a new type of dangerous driving behavior. Recent research has demonstrated that DSUB could reduce driver’s cognitive ability and operation level. In such cases traffic accident would be induced easily. However, until now there is not an integrated system of DSUB recognition and pre-warning. In this project a cooperative recognition method is proposed based on dual-view image system. First, kinematic models are established for behavior parameters detection from eye movement, head pose, lip movement and hand movement. Discriminant logic is then set up for four kinds of DSUB subdivision behaviors. Second, driver attention area and his/her alert state is recognized by the approaches of image information digging technology, pattern recognition, machine learning and so on. Then the risk of driver stress levels under each DSUB subdivision behavior is evaluated by fusing attention area and alert state. Third, the index set of driver attention area and alert state, which accord with individual behavior characteristics, is updated online and then the stress level evaluation archives is built. Fourthly, the risk level of vehicle running state is evaluated according to the operating parameters including vehicle speed and longitudinal distance. In order to enhance the safety and acceptability of the system, DSUB intelligent pre-warning strategy is proposed by coupling driver stress level and vehicle running state. This project is launched for DSUB recognition and pre-warning technology. The theory and practice achievements could provide technique support to driver assistant system on vehicles.