This paper introduces mainstream ultrasonic velocity measurement methods, and details the fundamental principles and calculation formulas of the transit-time method and the Doppler method. It focuses particularly on the ill-posed inverse problem inherent in ultrasonic velocity field reconstruction, and provides an in-depth analysis of the mechanisms, strengths, and inherent ill-posedness of classical inversion algorithms including the least squares method, Tikhonov regularization, and truncated singular value decomposition (TSVD). The paper summarizes key physical-signal joint processing strategies for mitigating significant ultrasonic beam drift and low signal-to-noise ratio (SNR). Prospectively, it proposes that the integration of physics-informed deep learning, computational fluid dynamics (CFD)-acoustics simulation coupling, and system-on-chip integration are core directions for advancing the technology towards high precision, adaptability, and miniaturization. This work aims to provide a reference for further breakthroughs and the engineering application of ultrasonic velocity measurement technology.