Abstract: In this paper,a novel method is proposed to automatically calibrate the cameras in expressway surveillance system under dynamic conditions.First,a GMM-based background modeling method is employed to obtain the background image and the region of interest(ROI) in which moving objects appear frequently.Then,the real lines(including straight long lines and marker lines),virtual lines(i.e.linking lines) and horizontal line are respectively detected and evaluated in the ROI by using different methods,and three vanishing points are thus obtained based on these line features.Finally,camera parameters are calculated according to the geometric relationships between the vanishing points and marker lines.Experimental results demonstrate that the proposed method is rather effective for real-time camera self-calibration in various expressway scenes with high precision and stability.