Weak target detection in sea clutter is always an important problem in detection theory and it is also one of the most complicated problems in radar signal processing. Sea clutter exhibits strong nonstationarity along time and range cells, which interferes gravely target detection, particularly, low-velocity small targets. Low-velocity small target detection in sea has important applications in military affairs and maritime rescue and it is always a hot-point problem in radar signal processing.At present, the nonstationary modeling on sea clutter are mainly related with the nonlinear feature description, chaotic model, as well as adaptive learning-based target detection using the neural network. These problems have been widely researched but no model can describe sea clutter in all the cases. In this paper, we try to exploit the time-frequency geometric features to detect low-velocity small targets in sea clutter, which utilizes the difference of the time-frequency geometric features between an echoes with a target and a pure sea clutter.In this paper, we first review the time-frequency analysis and time -frequency-based detection methods. Secondly, by analyzing a mass of the measured data of sea clutter, we show the difference of time-frequency geometric features between echoes with a target and pure sea clutter. Finally, we given the detection algorithm based on the time-frequency geometric features. The results of the measured data show that the proposed method can effectively find low-velocity small targets in sea clutter. This approach is a novel try to detect small targets in sea clutter.