关键词: 交通工程, 驾驶行为识别, 支持向量机, 货车移动遮断, 山区双车道公路

Abstract: To identify passenger car driving behaviors under truck movement interruption on mountainous two-lane highways, this paper collected vehicle trajectory data from an unmanned aerial vehicle (UAV) video films and image processing afterwards. The car- following behavior, lane- change behavior and overtaking behavior of passenger cars were calibrated with the threshold criteria of passenger car headway and slope of lateral position curve. To obtain the recognition model input variables, Kruskal Wallis Test and Principal Component Analysis were used to filter and reduce the dimensionality of passenger car driving behavior parameters. A passenger car driving behaviors recognition model based on Support Vector Machine (SVM) was developed using the grid search algorithm to determine the optimal combination of parameters for the kernel function. The study also analyzed the characteristics of passenger cars driving behaviors under truck movement interruption in multiple dimensions, then trained and tested the recognition model taking a typical mountainous two-lane highway in Yunnan Province as an example. The results indicate that: (1) the passenger car speed under truck movement interruption is lower than that under the free flow condition, the speed decrease is about 20 to 30 km·h-1 . (2) The average headway of a passenger car following a truck on two-lane highway in mountainous area is 2.53 s, which is less than the prescribed minimum safe headway and significantly increases the risk of the following behavior. (3) The recognition accuracy of the proposed model is up to 98.41%, which shows good recognition ability and applicability

Key words: traffic engineering, driving behavior recognition, support vector machine (SVM), truck movement interruption, mountainous two-lane highway