Abstract: To improve the efficiency of reverse engineering, the paper presents an approach to generate rough and finishing tool-paths from massive measured data directly. The rough machining is performed by machining volumes of material in a slice-by-slice manner. To generate the rough tool-path, a robust data-structure named tool-path net (tp-net) is constructed firstly, followed by a computation of cutter-location. And then the whole tp-net is partitioned into several optimal sub-cut-areas. Finally an optimal algorithm for linking tool-paths is proposed to acquire rough tool-paths effectively. For the finishing, a triangulated mesh is constructed from the massive points firstly. To avoid interference, we compute three kinds of interference-free cutter-location, vertex, facet and edge. The experiment results indicate that the algorithm of the rough tool paths is highly efficient and requires little memory. The finishing machining can avoid interference successfully and get reliable surface precision.