Active contour model, or snakes, are used extensively in digital image processing, particularly to locate object boundary, the model include geometric and parametric active models. Problems associated with initialization and convergence restrict their utility, which have been solved by Xu, who proposed Gradient Vector Flow. However, parametric active models still face the problem that the whole curve may evolve into only a side of the true boundary. This paper presents a new approach based on the parametric active models with Generalized Gradient Vector Flow(GGVF) and Dynamic Extern Force, which decreases when iterative number increases and thus makes terminal contour less dependent on the initial curve and it can speed up the iterate process. The image edge and the field that point to round from center decide the initial extern force, which can prevent curve evolution from possible wrong segmentation. The tongue edge can be detected faster and more accurately than ever in the tongue image experiments.