In this paper, based on circle detection and fitting, an accurate long-distance positioning system using laser range finders (LRFs), is proposed. To utilize orientation-invariant property of the circle, a cylinder shaped bar is chosen as the reference target of the positioning system. After obtaining the arc shaped contour of the cylindrical target scanned from LRF, given the prior information of the radius of cylindrical target, implementing maximum likelihood estimation (MLE) for circle fitting, an accurate position of the circle center can be determined. Because the reference target is not the only object in the application scene, we have to extract the target from other parts of the scanned data from LRF. Inspired by random sample consensus (RANSAC) and Hough Transform, a modified RANSAC algorithm of detection of cylindrical reference target for positioning system is proposed. Additionally, to improve the angular resolution of LRF, a pan unit is applied in the system. Experiment results prove the feasibility of the proposed positioning system.