DSS-VelNet: A Deep Learning Network for Flow Velocity Estimation

Water flow velocity measurement is a critical task in hydrological monitoring, disaster warning, and water resource management. However, traditional radar-based velocity measurement methods exhibit insufficient anti-interference capability in complex river environments, are susceptible to noise, and suffer from low measurement accuracy. This paper proposes a deep learning(DL)-based Doppler-Smoothed Spectrum-to-Velocity Estimation Network (DSS-VelNet) to measure flow velocity more accurately. Unlike traditional velocity estimation methods, DSS-VelNet uses the Doppler-smoothed spectrum of radar echo signals obtained from simulations and actual measurements as input to extract data features and predict water flow velocity. The training results demonstrate that the model's predictions fit well with the actual flow velocity and exhibit strong anti-interference performance.