Extracting Traffic Flow Velocity with Motion Vector

Abstract

This paper describes a method of extracting traffic flow velocity with high quality and speed, using traffic surveillance camera records. To reach this aim, MPEG features are used. Motion vector (MV), which is in MPEG format, describes motion value and direction. Firstly some filter is applied to MV for noise redundancy. After this process geometric correction is implemented to the motion vector to calculate real motion value and velocity. The developed system is tested with 4 different cameras records which are provided by Metropolitan Municipality of Istanbul Traffic Control Center and it is found to be useful under different environmental conditions.