Citation

Kolla E, Adamová V, Vertaľ P. Sci. Justice 2022; 62(1): 94-109.

Copyright

(Copyright © 2022, Forensic Science Society, Publisher Elsevier Publishing)

DOI

10.1016/j.scijus.2021.11.001

PMID

unavailable

Abstract

Video recordings from digital cameras implemented in road vehicles present a valuable source of data concerning various dangerous traffic situations, i.e., road accidents or near-miss incidents. This data is readily available and numerous methods for forensic reconstruction of traffic situations from vehicle video were presented in the past. New alternative method for reconstruction of traffic situations from vehicle video is presented in this article. The method is based on the fusion of kinetic vehicle trajectory simulation within 3D laser scanner point cloud, projective geometry, and processing of video footage from moving vehicle camera. The method offers accurate reconstruction of general vehicle motion within relevant time domain whereby in-depth technical information about traffic incident can be quantitatively extracted from video footage. The result is physics-based 3D projection of simulated vehicle motion on the motion of real vehicle recorded by moving monocular camera. The method was validated within performed experimental test runs with respect to vehicle speed, distance travelled, acceleration/deceleration and directional quantities (yaw rate, yaw angle). The method was further applied in the reconstruction of real-world traffic events.

Language: en

Keywords

Near-miss incident; Point cloud; Road accident; Simulation; Vehicle camera; Video analysis