Abstract: The paper discusses lines matching based on prediction by optical flow. Its contribution includes: (1) ChibaKanade algorithm for optical flow is adapted to better the estimation of low texture regions in images; (2) Hough parameters of lines are acted as similarity measure along with lines’ orientation defined by gray level. We use the measure treat with parallelclosed lines and gain credible correspondences; (3) Projection similarity measure is defined to treat with multisegment matching.