Parçacık filtresi tabanlı sürücü manevra algılama ve uyarı sistemi

Road traffic accidents are a serious socio-economic problem which causes moral and material cost. To reduce traffic incidents researchers introduce Advanced Driver Assistance Systems. Essentially, these systems provide real-time warning, information about the traffic flow and environment during driving to enhance safety. In this study, we present a driving assistant system capable of generating warning to the driver about the possible emergency situation. The main aim of the proposed vision-based driver assistant system is to accurately trigger timely warnings with respect to the driving scenarios especially possible dangerous situations. Moreover, the proposed system can be further used for evaluating the driver's driving performance. The system consists of two modules; a reference driver model and a driver evaluation system. While building the reference driving model, the particle filter is used. Basically, the particle filter probabilistically evaluates the driver by taking into account surrounding traffic environment and the interaction with other vehicles. Following that, by feeding the obtained reference model with the driver's behaviors during driving, the system gives support to the driver to improve his/her maneuvering tasks in both longitudinal and lateral direction. Experiments are conducted through a vehicle simulator. For evaluating our system, a testing set containing young, middle-aged and old drivers is used. Each driver's performance and authority is tested with and without using mobile phone while driving. The success of the drivers is measured by comparing their actions with the best optimal following action obtained from the particle filter.