Analysis of Dashcam Video for Determination of Vehicle Speed 2020-01-0877
Using video footage to extract accurate vehicle speed data is often difficult when the available video comes from cameras whose optical parameters are not known. One means of overcoming such difficulties lies in the use of visible landmarks within a video frame whose three-dimensional locations and geometry can be determined. For certain dashcam videos, such information can be extracted from Google Earth and used for image calibration. In this work, we used a commercial videogrammetry software tool (Photomodeler Standard) with Google Earth, using video from just one camera to determine vehicle speed. Once the global origin has been selected (corresponding to the initial 3D camera position where video stream begins), the three-dimensional positions of identifiable features are located in the first image relative to that global origin. While the absolute position of the camera changes with each successive image of the video stream, its location relative to the image frame remains the same while the tracked feature locations progress downward. Conversely, the position of each identified feature (relative to the global origin) remains stationary in the environment. Therefore, the known absolute coordinates of each identified feature can be used to identify how the position of the camera changes in each successive image. The camera’s absolute location is then determined through successive frames by using verified 3D measurements of each tracked feature produced by Google Earth. We applied our approach using dashcam footage recording at a maximum of 30 frames per second, and used the calculated position to determine vehicle speed (which was compared with the reference speed provided by the satellite-based speedometer display visible in the dashcam footage). In addition, factors that affect computation error are identified and discussed.