Application of Matchmoving for Forensic Video Analysis with Recorded Event Data

Abstract

A fatal collision involving a pedestrian struck by a public train at a rail station located in a major U.S. metropolitan city was investigated. The train was equipped with an event data recorder that recorded valuable empirical data related to the collision, such as the train speed, braking, and acceleration inputs. However, the original digital version of the data was not available to analyze, and only a single screenshot of the data in an analog graph format was available. The subject train was equipped with multiple video cameras at various vantage points that recorded video footage of the collision. Using the process of “matchmoving” — and with the assistance of three-dimensional LiDAR scanning of the station and train — video footage was analyzed to spatially determine the location of the train in three-dimensional space. The process of matchmoving is an established scientific process used to calibrate a virtual camera to “match” the movement and optic properties of the real-world camera that captured the video. Further analysis was performed to determine the train’s kinematics (such as its speed and deceleration rates) leading up to the collision. The accuracy of the matchmoving analysis was then verified with the available event data.