Lane changing intent identification based on logistic regression model
Jinshuan Peng, Lei Xu
Chongqing Key Lab of Traffic System & Safety in Mountain Cities, Chongqing Jiaotong University, Chongqing 400074, China
To reduce the risk of the lane changing behaviors, based on integrated collection platform, the research group conducts experiments under real road environment for the purpose of studying divers’ lane changing intent identification. On the basis of the drivers’ fixation characteristics of the rearview mirrors before changing lanes, the length of lane changing intent time window is determined. Based upon differential analysis of visual characteristics between lane keeping and lane changing intent stages, saccade numbers, visual search extent, saccade amplitude, standard deviation of head rotation angles in the horizontal direction are selected as the characteristic indexes of the identification. The logistic model is built according to feature extraction of the leaning samples, then applied to the identification process after the validity test. Results show that the identification success rate may reach 90.42%, thus verifying the feasibility and effectiveness of the logistic model to identify drivers’ lane changing intent.