Under complex sea conditions, ship detection from remote sensing image is easily affected by the ship wake, sea clutter, oil, and thin cloud, which may lead to poor detection results and difficulty in the detection of small ships. Herein, we propose a saliency optimization ship target detection model based on an adaptive robust background. The proposed method uses the Tophat algorithm for preprocessing of the original image to suppress interference from the ship wake and sea clutter. Further, an adaptive superpixel segmentation method is proposed to optimize the robust background detection model. An improved Otsu segmentation method based on the mean information is proposed to determine the area where the ship is located. The experimental results demonstrate that the proposed method can effectively detect the location of a ship under various sea conditions. The proposed algorithm demonstrates high detection precision (91.20%), recall (79.31%), and comprehensive evaluation index (84.00%). When compared with the existing saliency detection algorithms in ship detection, the proposed algorithm exhibits obvious advantages; therefore, it is suitable for small ship detection based on the remote sensing images under complex sea conditions.