Research on Traffic Conflict Automatic Discrimination Based on Support Vector Machine

Traffic conflict discrimination is a basic technology to evaluate road traffic safety and driving assistance. In order to improve the accuracy of the traffic conflict discrimination and to achieve the level of the automatic discrimination, the traffic conflict discrimination can be divided into two stages: occurance discrimination and severity discrimination. The integrated application of traffic conflict discrimination and kinematics theory will set up critical conflict regions model centering on speed, distance and angle, and thus achieve traffic conflict discrimination; Utilising machine learning method based on support vector machine to distinguish the severity of traffic conflict and then form a complete traffic conflict automatic discrimination method, finally, the practicality and accuracy of model were proved through examples. The investigation results indicated that the method could be adopted to distinguish the probability and the severity of conflict of traffic entities, which could provide theoretical basis to traffic conflict automatic discrimination technology.