Design And Implementation Of An Abnormal Driving Event Detection System For Smartphone Vehicles

With the development of road traffic industry,the number of vehicles is increasing.Road traffic safety has become one of the major safety problems in the society.According to the survey,more than 90% of the traffic accidents are caused by improper operation of the driver,and these improper driving events are often the main cause of the accident.Therefore,it is very important to study effective detection methods of abnormal driving events.In recent years,experts and scholars at home and abroad have put forward a lot of research methods of driving events.The existing driving event detection methods often use vehicle terminal equipment,which bring some trouble to the driver's operation.With the rapid development of mobile terminals,such as smart phones,its built-in sensor functions are gradually improved,and smart phones have gradually become an ideal platform for driving event detection.We think about how the smart phone can collect relevant behavior at anytime and anywhere and the accuracy is unchanged;what kind of discriminant method can be used to better identify abnormal driving events.Therefore,in view of the above problems,this paper deeply studies the related situation of driving event identification on smart phone,summarizes the current research status at home and abroad,and gives an accurate and effective prototype system for detecting abnormal driving events of smart mobile vehicles,and answers the above questions.The main contributions of this article are as follows:First of all,this paper analyzes the drivers' driving events,and focuses on the research of the driving events such as acceleration,deceleration,the left and right channels,the brake,the right and left turn and other driving events.By using the sensor(acceleration sensor,gyroscope sensor)to collect the related state parameters of the body movement,we can analyze it and determine which one dimension component of the sensor affects driving events.In view of the ability of smart phones to perceive data anytime and anywhere,the problem of space rotation based on four elements is proposed in this paper.By converting three attitude angles of Euler angle into four elements,we can solve the rotation matrix,so that the data will not drift when the data is changed.Secondly,data filtering and endpoint detection are used to preprocess the sensor data.In the data filtering,we select the elliptical low pass filter and the sliding average filter.By comparing the effect of the filter,we choose the 5 order elliptical low pass filter as the data filtering method in this paper.Secondly,for endpoint detection,the mean value of data is used as endpoint energy value of endpoint detection.In the process of data feature extraction,we use the hybrid feature selection algorithm of SFS and SVM to select 20 eigenvalues as the input values of driving events classification.Finally,in view of the driving events such as acceleration,deceleration,turn,lane change,and brake,this paper uses two classification algorithms of OS-ELM and Lib-SVM to compare the running time and accuracy of the two,and judge the better Lib-SVM algorithm.The experimental results show that the identification rate of driving events is 93.3110%.So we chose Lib-SVM as a classification algorithm to classify these 7 driving events.For abnormal driving events,we propose to define the threshold recognition algorithm to judge,if exceeding the specified threshold,it is judged to be abnormal driving events.The experimental results show that the intelligent mobile vehicle driving event detection system proposed in this paper can effectively identify and judge the abnormal driving events,and is more convenient and quick than the previous driving event discriminant system.