Based on the idea of generating NC tool-path from the point cloud directly, the topology reconstruction methods of scattered point cloud are emphatically studied in this paper. The scattered point cloud is divided at equal intervals and stored in link list structure in the process of topology reconstruction. In each data region, the points are reduced and projected to a middle slice plane, obtaining a set of scattered points. A double-mapping algorithm based on orthogonal projection theory, which aims to sort the scattered points on a slice plane, is presented. Each point and its neighborhood can easily be searched by this sorting method and the limitation of the method using the natural adjacent sequence of points is also researched. It aims to develop a judge criterion of sorting sequence by computing the vector locating trend of the current point with its neighborhood to further improve the double-mapping algorithm and the topology of the slice data is fully reconstructed.