Abstract: Line detection is an important low-level operation in computer vision that has many applications.This paper proposes a new difference of Gaussian（DOG） mode-l based line detection algorithm in which its performance is compared with that of Gaussian kernels-based detector,and explains the relation between DOG function and Gaussian kernels. Furthermore,the line detection range and 2D mechanism of the presented new detector is discussed. Experiment result in fingerprint image shows that DOG model-based detector is able to extract the skeleton of line with a certain width.