Analysis of Dashcam Video for Determination of Vehicle Speed

Using dashcam video footage to extract reliable vehicle speed data can be challenging when the only available image stream comes from a camera whose optical parameters are unknown. One means of overcoming such difficulties uses visible landmarks and features within the video frame whose dimensions can be independently measured. While good results have been obtained by others using a Total Station or LiDAR to physically measure locations for such purposes, this approach could prove difficult if a site of interest is inaccessible (e.g. on a busy highway that cannot be shut down) or if relevent features of the target location have changed (e.g. due to construction or even restriping of lanes lines). As an alternative to direct scene measurement, it is proposed that measurement of features visible in overhead satellite images be used to dimension relevant features visible in the video. In this work, videos from individual dashcams and phone-based cameras were used to calculate average vehicle speed by tracking individual features as they pass through and then exit the image frame. Because the features used herein are limited to periodically-occurring markings such as lane lines, average vehicle speed is calculated based solely on distances referenced to these features as well as video frame rate. Although limited in scope, a major advantage of this approach is that it eliminates the need for camera calibration, because only one small patch of the image frame is used for the average speed calculation. In this study, selected video sequences were used to calculate vehicle speed at speeds ranging from 16 - 65 mph, with results corresponding to the vehicle’s speedometer display for all but the lowest frame rates.