In this paper we apply advanced machine vision technology to the water meter calibration. Several high speed and high resolution cameras are installed at the top of the checked meter to dynamically access the running water meter image. The meter pointer is accurately positioned using color feature modeling and background Gaussian fitting algorithms. Machine vision approaches are used to convert the captured images into digital readouts. Encountered problems such as blurred surface and air bubble interfere are overcome. The developed system realizes high efficient automatic water meter verification.