FPN-D-Based Driver Smoking Behavior Detection Method

In view of the fact that a driver's smoking behavior seriously affects driving safety, feature pyramid network (FPN)-based identification method has been studied in order to reduce occurrence driver smoking. Most existing research focused on detection and recognition based movements while or smog characteristics. Therefore, probability misjudgment such methods is high, thus address this issue, present work proposes FPN detect behavior. combined with dilated convolution technique small target object image recognize their By using images collected from vehicle platform, simulation experiment was carried out by employing proposed work. The experimental results show accuracy 94.75%, recall rate 96%, precision 95.05% area under receiver operating characteristic curve 95.5%.