B-spline fitting is a technology of fitting a curve with a set of points which scatter along a line or a curve in 2D or 3D space. In this paper, we use B-spline to represent road lines in images of 36-pair Mobile Mapping System (MMS) image sequences, and road line information can be collected rapidly and cost-effectively. Edge detection and template matching are used to extract data points on the target road lines, and multi-image collinearity is used to determine 3D data point coordinates. Least-squares B-spline curve fits in 2D and 3D spaces, and it is used to detect and remove blunders. It is an automatic procedure of road line extraction, due to results can provide information for recursive processes. The results show that our algorithm can successfully extract 3D road lines.