Driving Behavior Identification Based On Multi-Sensor Of Mobile Terminal Data Fusion

Car as a great invention of modern society,has played an important role in promoting urbanization and improving people's living standard,but the resulting traffic safety problem is also serious,road traffic accident has become one of the important problems in today's society.Especially,as the core factor of road traffic safety,the safety of driving behavior of drivers has greatly affected the safety level of road traffic system.Driving behavior research has become an interdisciplinary field of computer science,psychology,control science and cognitive science.From the point of purpose of research,the driving behavior research can be divided into descriptive model and functional model.Descriptive model is used for describing the state of driving behavior and functional model is used for explaining the principle of driving behavior such as Drivers' psychological changes.In this paper,we research the former.Meanwhile,with the popularization of mobile terminal,the mobile terminal,as the new hardware carrier,contains many sensors such as GPS,inertial navigation sensor,magnetometer and camera.This makes it possible to control road traffic safety through mobile terminals.But overall,the driving behavior research on mobile terminals focuses mainly on macro-level road congestion guidance and travel route planning,the research on micro level driving behavior safety is relatively slow.Many high precision technologies tend to high cost,and due to the different driving behavior identification mechanism,the application scenario of different technologies is also limited.Based on the above background,this paper proposes using the low cost sensors in mobile terminal to process data fusion,and using fused data to identify basic driving behaviors including sharp turn,change lanes,sharp acceleration and sharp deceleration.We aim to provide a low-cost,real-time,not relying on external devices and relatively wide range of applicability driving behavior identification solution.The innovation points of this paper are as follows:(1)Initial relative attitude determination of mobile terminal.Since the mobile terminal is not fixed on the vehicle,the acceleration and angular velocity measured by the mobile terminal cannot be directly used in the description of the vehicle motion state.We proposes using magnetometer assisted accelerometer to calculate the rotation matrix from mobile terminal coordinate to carrier coordinate.Experimental result demonstrates the method is fast and effective.(2)Improved and simplified attitude and trajectory fusion algorithm.In reference to the traditional algorithm of SINS and GPS fusion,we improve and simplify classical algorithm.Posture fusion: using the accelerometer combined with low-pass filter to obtain pitch Angle observation which has higher precision than pitch angle obtained from GPS of mobile terminal.Trajectory fusion: we simplify 15 parameters 3d trajectory fusion to 8 parameters 2d plane trajectory fusion which reduce computation effectively.The experiment show that the combined attitude data and trajectory data combine the characteristics of dual system and have better robustness.(3)Driving behavior identification based on fusion attitude and trajectory data.Based on the fusion trajectory and attitude data,this paper proposes a method of combining threshold value method and support vector machine to identify the basic driving behaviors and to some extent,improved the accuracy of driving behavior recognition.