Extraction of Vehicle Turning Trajectories at Signalized Intersections Using Convolutional Neural Networks

• DOI: 10.1007/s13369-020-04546-y
• Publication Date: 2020-05-04T17:03:47Z
• Authors: Osama Abdeljaber; Adel Younis; Wael Alhajyaseen
• Keywords: Jerk; Tracking (education)

Abstract This paper aims at developing a convolutional neural network (CNN)-based tool that can automatically detect the left-turning vehicles (right-hand traffic rule) signalized intersections and extract their trajectories from recorded video. The proposed uses region-based CNN trained over limited number of video frames to moving vehicles. Kalman filters are then used track detected trajectories. achieved an acceptable accuracy level when verified against manually extracted trajectories, with average error 16.5 cm. Furthermore, using vehicle tracking method were demonstrate applicability minimum-jerk principle reproduce variations in vehicles’ paths. effort presented this be regarded as way forward toward maximizing potential use deep learning safety applications.