In order to monitor some critical maritime areas, a dynamic platform based on charge coupled device (CCD) cameras is used, a novel algorithm for ship target detection based on texture model of sea surface is presented. The sky background and horizon are detected quickly by using an energy feature in discrete cosine transform (DCT) domain of image blocks. In order to separate ship targets from the complex sea background below the horizon, the texture feature of sea surface domain under the horizon image blocks in DCT domain is extracted, and a new texture mixture model of sea surface is developed by using adaptive fuzzy c means clustering technique. Ship targets are segmented from sea background by the constructed sea surface texture model. The experimental results show that the proposed algorithm can detect ship targets quickly and steadily, especially suitable for maritime surveillance based on non-stationary monitoring platforms under the large wave sea background.