Recognition Of Railway Driver's Behavior Based On 3D Skeletal Tracking Of Monocular Camera

In recent years,China's railways is at a stage of rapid development.While bringing convenience to people's travel,its safety is also facing a severe test.During the train running,the driver's improper operation will greatly reduce the safety factor of the traffic system.The current monitoring method of the Railway Bureau is to post-process the surveillance video through manual methods,and it is impossible to timely respond to and deal with abnormal train events.This paper uses computer vision to intelligently analyze driver behavior,and converts the traditional post-processing of video to a more reasonable "pre-processing",which effectively reduces the possibility of accidents.The main contents of the paper are as follows:(1)This paper introduces the status quo of detection and identification of locomotive driver behavior,and analyzes the importance of intelligent monitoring system for train driving safety.At the same time,according to the application scenario of this paper,the defects and drawbacks of the current identification methods are clarified.Combined with the status quo of human behavior recognition,a new detection and recognition algorithm is found.(2)This paper designs a foreground extraction method based on the driver's threedimensional pose in a monocular scene.First,pre-process the video collected by the train and convert it into data that meets the requirements of the system algorithm.Secondly,compare the common foreground detection methods,select the OpenPose bone extraction method that best fits the application scenario of this paper,and perform the obtained data.Smoothing and other processing;Finally,the camera imaging principle and the human skeleton model are used to restore the driver's posture information to the three-dimensional coordinate system,and the ambiguity of the monocular to three-dimensional is eliminated by combining the driving environment and the internal and external constraint information of the human body.(3)This paper identifies and classifies the driver's three-dimensional posture behavior.According to the locomotive driver's code of conduct,some of the driver's body parts must be restrained in a specific area,and the detachment area is alerted.In order to improve the efficiency of the algorithm and the accuracy of the results.By analyzing the behavior characteristics of drivers,this paper divides them into two categories: static behavior and dynamic behavior.The former combines the cab space model for constraint recognition;the latter uses the hidden Markov model to classify and identify the driver's dynamic behavior.Finally,the system is built,and the whole content of this study is summarized to further clarify the future research direction.