Unmanned surface vehicle (USV) plays a more and more important role in various areas such as meteorological monitoring, maritime search and rescue, enemy reconnaissance, and precision strike. However, special features in real marine environment such as cloud clutter, sea glint, and weather conditions result in various kinds of interference in optical images, which makes it very difficult to detect the sea sky line accurately. To solve this problem, a sea sky line detection method is proposed based on gradient saliency. The line features of sea sky line are enhanced effectively through the computation of gradient saliency; other interference factors are suppressed; sea sky line detection and identification are achieved by region growing method. In the end, the proposed method is tested on optical images from “XL” USV in real marine environment and the experimental results demonstrate that the proposed method is significantly superior to other state-of-the-art methods in terms of detection rate and real-time performance.