In this article, solutions for redundant measured data elimination and surface reconstruction on reverse engineering are presented. In redundant data elimination, we propose a simple algorithm that can efficiently discard those redundant measured points according to the required degree of accuracy. While in the surface reconstruction, we try first convert all eliminated column or row data into spline (B-Spline, Beizer and Cubic-Spline) curves. Methods for two dimensional spline (u, v directions) curve construction are described. Non-meshed two dimensional spline curves are then blended to a surface model by formatting data points into a sparse matrix data structure. Experimental results show that the proposed characteristic points extraction method can remarkably reduce the reconstruction time only on the cost of a little extracting time and minor modeling errors. Comparisons among B-spline, Beizer and Cubic-spline on data structure, computing time, and accuracy demonstrate that the B-spline algorithm is superior to other algorithms in surface reconstruction.