The surface reconstruction algorithm of the large-scale irregular data based on laser scan has been studied. A new algorithm for two-dimensional data which is based on the Delaunay triangulation is proposed. The algorithm is divided into two parts: the generation of initial triangular mesh and optimization of the initial triangular mesh. At the very first, put the data in sequence according to the x coordinate of the data, and generate the triangular mesh directly, then optimize the triangular mesh according to the convex-concave of the quadrilateral that has the public side. The efficiency of Delaunay triangulation is improved by this method. This article describes the specific algorithm. In this paper, do projection transformation to the data obtained from the laser scanning first, the data are transformed into two-dimensional plane from the three domains; secondly divide the two-dimensional data by the improved Delaunay triangulation algorithm, and then switch to three-dimensional curved surface for blanking and linear interpolation to obtain three-dimensional images. Use the laser scanning data obtained from coal-field as an example, the algorithm has been applied to practice and has a good display.