A novel ultrasonic flow velocity estimation method employing the two-dimensional (2-D) frequency shift track is proposed. By representing the backscattered echoes from human blood in the 2-D frequency domain, this paper shows that the velocity profile measurement using pulsed-wave Doppler ultrasound and the moving target track in image processing are isomorphic problems. Therefore, feature template matching based on geometric moment invariants can be used to track the 2-D frequency shift from which the velocity is estimated. Then we discuss the characteristics of the moment invariants and present correlation coefficient as a tracing strategy. The experimental results show the method can overcome the disadvantage caused by frequency-dependent attenuation and get better effect in velocity profile estimation.