Distracted Driving Behavior Detection and Identification Based on Improved Cornernet-Saccade
Wei Zheng, Qing Qing Zhang, Zhen Hao Ni, Zhan Guo Ye, Yuan Min Hu, Zhong Jie Zhu · 2020
While driving, drivers need to be highly concentrated due to distracted driving can lead to serious accidents. In this paper, a scheme for detecting distracted driving based on an improved CornerNet-Saccade is proposed. To eliminate the risk of traffic accidents caused by distracted driving, through detecting drivers' operations when driving. The scheme includes two stages of model training and testing. First, data set is self-built, the CornerNet-Saccade model is optimized. And, then the network model is trained to detect the smoking action and eating behaviors while driving. The experimental results show that the proposed scheme can detect the improper driving of smoking action and eating behaviors in the process of driving in real time, with the accuracy of 71.1% and with a running time of 156fps. The detection accuracy is improved by 3.0% when the improved CornerNet-Saccade is guaranteed to be detected in real time.