Microscopic aggregated traffic parameter extraction against complex camera motion interference

Abstract Unmanned aerial vehicles (UAVs) attract increasing attention in the transportation community due to their low cost, wide view coverage and rapid deployment. Previous studies mainly focus on extracting traffic parameters from aerial videos with air-borne cameras in planar movements. Less focus is paid to extracting traffic parameters from the UAV-like videos recorded with multi-dimensional camera movement. To address this issue, we propose an adapted framework for obtaining traffic parameters with the UAV camera moving at different dimensions. First, the framework introduces a temporally robust global motion compensation model to compensate for UAV camera movements and obtains a stabilized traffic scenery background. Second, the kernelized correlation filter is integrated into the proposed framework to accurately track vehicles. Third, we introduce the Hough line detection to find reference markings in videos and map the image length from video to physical length in the real world. Fourth, we estimate microscopic traffic parameters, including individual vehicle speed, time headway and space headway in a traffic stream. We testify the proposed framework performance on three different videos that are collected under the interference of different camera movements. The experimental results show that the proposed method achieves an accuracy of 96.98% and 96.94%, respectively.