The classification and measuring procedure for geometrical objects of a city will be simplified if its main road network could be automatically extracted from aerial remote sensing images. In this paper, an automatic approach for extracting main city roads based on straight lines is developed, in which main city roads are modeled according to their characteristics in the image. The new method puts emphases on dealing with all kinds of noise from the imaging process, and uses Gaussian distribution to blur parameters of straight lines in order to ensure extraction robustness. All parameters of this method are acquired by theoretic analysis so that the extraction process is automatic. The experimental results show that the main road network can be extracted from real aerial remote sensing images.