Drivng Behavior Analysis Based On Vehicle Information Fusion

Currently, with the economy developing continuously, the amount of the vehicles increases every day. As the same time, the amount of non-professional drivers increases rapidly. Since most novice drivers are unskilled, unfamiliar with the vehicle condition and in weak awareness of traffic safety, the drivers’ personal factors and the vehicle factors have become the main reasons of traffic accidents. The driving auxiliary equipment is urgently needed to remind drivers the vehicle information in time and correct improper driving behavior.To solve this problem, this paper proposes a driving behavior analysis method based on vehicle information fusion. This method firstly respectively uses driver and vehicle as the study object. Then, it combines the information of these two parts to comprehensively analysis the current status of the driving under the certain driving condition. In this paper, the study process consists of the following three stages. The first stage uses the driver training simulators and the vehicle signal simulation platform to collect driving operation information and the on board diagnostic(OBD-II) system information. The second stage preprocesses and extract the feature of the driving operation data and the OBD-II data that have collected. In the third stage, using the information fusion technology, we proposes a driving behavior analysis method based on driver-vehicle-environment information fusion. In this stage, we respectively use vehicle information fusion technology from the feature level and the decision level. On the feature level, we firstly propose a driving behavior analysis method based on the person to directly analyze the driving state. This method uses Support Vector Machine(SVM) combining the two characteristics which are the steering wheel change rate and brake throttle pedal change rate to analyze whether the current state of the driving behavior is safe or dangous. Secondly, we propose a driving behavior analysis method based on the vehicle to indirectly analyze the driving state. This method uses the strong classifier(Ada Boost) combining three characteristics which are the relative ratio of the speed and engine speed, the relative ratio of throttle valve and engine speed and engine load to analysis whether the current state of the driving behavior is safe or dangous. On the decision level, based on the results of the feature level, the BP(Back Propagation) neture network are used to analyze and determine the current driving state. The experimental results show that the accuracy rate is over 99%, which indicates the effectiveness of our method.