Research On Bus Driver’s Abnormal Behavior Analysis System Based On Deep Learning

In recent years,public transport safety problems are usually caused by accidents involving buses.It is found that the driver’s abnormal behavior is an important cause of bus accidents.This paper is supported by the independently developed intelligent monitoring system for driver abnormal behavior,using technologies based on deep learning and video and image analysis,this paper analyzes drivers’ abnormal behaviors in bus scenes.The main work contents are as follows:(1)Research on recognition algorithms of distracted driving and distracted driving based on object detection.In the bus scene,the deep learning method was used as the framework,the target detection algorithm was used to obtain the location information of the driver’s face,cigarette,mobile phone and water cup,and the discriminant conditions of distracted driving behavior and disturbed driving behavior were proposed based on the target detection results.The discriminant condition is based on the detected spatial location information of the target frame,and the result selection strategy is designed to realize the identification of distracted driving and disturbed driving based on the self-made verification data set.The experimental results show that the proposed method has high accuracy and simple structure,and can realize real-time behavior recognition on embedded core computing unit.(2)The research on the recognition algorithm of driver fatigue driving behavior is based on facial key point detection.In the actual bus scenario,taking the deep learning method as the framework,the facial key points detection algorithm is used to obtain the location of the facial key points,and the discriminant condition analysis and setting of the fatigue driving behavior based on the facial key points detection results are proposed.Based on the spatial position points of the eyes and mouth and the Pitch posture Angle located by the facial key point detection algorithm,the discriminant condition was abstracted into the features of facial fatigue,and a video was decomposed into a sequence of picture frames to realize the recognition of tired driving based on the selfmade verification data set.Experimental results show that the proposed method has good recognition accuracy and applicability.(3)This paper builds an intelligent monitoring system for bus drivers.The system uses the NVIDIA Jetson Xavier NX module for real buses to build a core computing unit that can be used for video processing and driver abnormal behavior analysis based on the above recognition algorithm.The experimental verification is based on the locally homemade data set,and the detection precision rate of driver abnormal behavior analysis algorithm in the core computing unit is more than 95%,which proves that the proposed algorithm can be applied in the intelligent monitoring system of bus drivers.