Recognition of dangerous driving behaviors based on support Vector Machine regression

In this paper, taking the typical dangerous driving behaviors of following the preceding car too close and abnormal ways to change lane as the research object, the identification indicators are selected. Using the UC-win/Road software to build the scene, the samples to be identified were obtained, and support vector machine regression and BP neural network recognition models were constructed respectively to identify dangerous driving behaviors and compare their accuracy. The results showed that the driving behavior recognition model based on support vector regression can effectively identify the driving behavior, and the accuracy of discriminant inference was significantly higher than that based on BP neural network. both reaching over 92%. At the same time, the average prediction time was also obvious reduced. The research results have important theoretical and practical significance for the performance evaluation of behavior recognition models and the development of road traffic safety.