Recognition And Prediction Of Taxi Traffic Conflicts Based On Driving Video

Traffic conflicts technology is an important part of the traffic safety, it is great significance to study the recognition and prediction of traffic conflicts for promoting the level of traffic safety. The article includes four aspects, proposing the recognition method of traffic conflicts, analyzing the distribution characteristics of traffic conflicts, building a prediction model of traffic conflicts for taxi drivers, and evaluatting the safe level of drivers.Video drive recorders were fixed on taxicabs to transcribe driving video. The traffic conflicts recognition of subjective observation and interactive verification was proposed based on analyzing the existing conflicts parameters, and the reliability of recognition was improved by training and testing the observers. 426 hours driving video was recognized through cutting the video which was traffic conflicts possibly, recognizing conflicts, and testing mutually. There were 7239 traffic conflicts altogether.In order to provide the actions properties for the conflicts prediction model, drivers were divided into three types by hierarchical clustering method, according to their behavior characteristics. In addition, the distribution characteristics of traffic conflicts were described along with the conflicts time, conflicts objects, conflicts types, conflicts places and conflicts grades, and the reasons were given for that.In order to build the traffic conflicts prediction model, firstly, the influence of drivers’ properties on traffic conflicts was analyzed, such as the gender, age and driving years. Then, the basic and behaviour properties of drivers were selected as the independent variables for the model, and the dependent variable was the number of traffic conflicts. Negative binomial model was used to build the prediction model, and the model was demarcated according to the collecting data. In addition, models were built for different conflicts time, types and grades, and the relationship between independent and dependent variables was analyzed on the basis of models. finally, the accuracy of models was tested.The grey theory model was ensured to evaluate the safe level of drivers, and the evaluation indexes were the number of traffic conflicts, the grade of traffic conflicts and accident rate. The evaluation matrix was built on the basis of indexes. In addition, the weight of indexes was determined by analytic hierarchy process. Drivers were divided four types, and the explicit values were received by drawing the cumulative frequency curve. Finally, four types of definite weighted funtions were built, and calculating the assessed value for each evaluation index, The safe levels of drivers were received according to the size of the assessed value.In order to explain the process of the prediction model and the safety evaluation, six drivers were selected to calculate their number of traffic conflicts and safe levels.