Straight lines are very important intermediate level symbol in image analysis. There are many approaches to extracting straight lines, but they all have their particular drawbacks to limit their use. A robust approach to extracting straight lines is proposed in this paper. A series of methods are devised to overcome all kinds of noise in real images. It firstly thins edge points, and then tracksthem to form line support regions by using the Kalman filter, finally line support region is approximated to a line by using a robust method. The experimental results show that this line finder has a perfect performance and is suitable to a wide range of real images.