Abstract: In order to realize the automatic harvesting of sea cucumber and diagnose the disease of sea cucumber, first, the problem of the image segmentation of sea cucumber under real aquaculture environment should be solved. In this paper, a new method of image segmentation of sea cucumber using GrabCut with saliency map was proposed. This method improved the traditional GrabCut algorithm, enhanced underwater images through the single scale Retinex algorithm. Based on global contrast based salient region detection method and histogram equalization, part of foreground and possible background of regional image of sea cucumber could be obtained, the mask of GrabCut algorithm can be initialized using this information. At last, GrabCut algorithm ran iterated to get the result of image segmentation. Experiment results proved that the proposed method can segment the sea cucumber images more accurately than the Otsu method, the watershed method and the traditional GrabCut algorithm, and overcome the background noise and preserve the details of the target image. The accuracy of the algorithm was 90.13%.