Abstract: In order to enhance the floating detection ability of sea surveillance radar, an algorithm based on the feature extracted from the echo of sea surface is proposed. Firstly, some features such as relative Doppler peak, relative Doppler offset, relative Doppler entropy are extracted from the time series of the echo to verify the efficiency of separating sea clutter and small floating target in time dimension. Secondly, the feature detector is constructed which the decision region under the given false alarm rate is determined by convex hull algorithm. Finally, the efficiency of the detector is tested by real radar data, the result shows that it can fully distinguish cooperative target from sea clutter in 512 pulses when the false alarm rate is 0.01. The performance of double feature detector is better than traditional moving target detection method.