Algorithm for Estimating Mean Traffic Speed with Uncalibrated Cameras

Publisher Website
Google Scholar
Abstract
Presented is a new algorithm for estimating speed by using a sequence of video images from an uncalibrated camera. An assertion is made that exact calibration is not necessary for estimating speed. Instead, the following are used to propose a novel method that extracts scale information and estimates speed: ( a) geometric relationships inherently available in the image; ( b) some common sense assumptions that reduce the problem to a one-dimensional geometry; ( c) frame differencing to isolate moving edges and to track vehicles between frames; and ( d) the distribution of vehicle lengths.