Stereoscopic vehicle speed measurement - System calibration and synchronization errors analysis

This paper presents a system and an algorithm for the estimation of translational speed of moving objects. The speed estimation is based on stereoscopic analysis of displacements of a region of interest on a moving object. The system is applied to vehicle speed measurement. A legacy automatic license plate recognition (ALPR) algorithm is used for license plate segmentation. The license plate is used as region of interest (RoI) and is tracked throughout a video stream. The RoI displacement and the elapsed time are measured in order to provide the vehicle average speed estimate. The system calibration is described in detail, as well as its simplifying effect on displacement measurement. The analysis of possible errors and artifacts is given and illustrated by experimental results.