The invention discloses a GSA-based concave image segmentation method. Firstly grayscale processing is performed on an original image, then initial snake element convergence is completed by using a GSA model, and the final snake element convergence is finally completed through a greedy algorithm. According to the method, normalization processing is performed on the image, the continuous force, the bending force and the image fore are balanced, and fitting of the snake elements to the position around the target contour, especially fitting to the concave area of the image, is rapidly and uniformly performed under the premise of guaranteeing the original sequence of the snake elements so that the efficient computing advantage of the GSA model can be maintained, the contour extraction accuracy can be enhanced, and especially the problem of concave target contour extraction can be greatly processed by the method.