In order to improve the efficiency and accuracy of geometric active contour model-based segmentation algorithm, a novel edge detection and tracking algorithm is presented. Firstly, the gradient of an image is calculated according to vector image, and an edge indicator with adaptive threshold is proposed. Secondly, an improved evolution model using variational level set is put forward. Then, on the basis of this model, an improved edge detection algorithm is proposed, and a target tracking algorithm is designed in the framework of unscented Kalman filter. Experimental results demonstrate that the proposed algorithm not only increases the convergence rate and flexibility of active contour evolution model significantly but also possesses strong robustness to such interferences as shadow, occlusion, object deformation and background interference.