In recent years, there has been a significant increase in inspecting and evaluating transport infrastructure. Traditionally, these structural data were collected manually by measuring and redrawing the construction against design documents. In recent decades, laser scanning technology can help collect 3D data rapidly and accurately. The 3D point clouds can provide detailed texture and shape information of complex construction such as bridges. This study aims to develop a 3D mesh model for a finite element simulation from a 3D point cloud of a bridge's Pier collected by Terrestrial Laser Scanning (TLS). The point cloud is structured, and the object boundary points are generated using the marching cube algorithm. The boundary and inside points, which imply the vertex of the solid element in the 3D mesh model, are grouped as a new point cloud. The generated point cloud is input into 3D CAD, and the 3D solid model is manually created. As a result, the 3D mesh model is developed and successfully imported to ANSYS software for the structural behavior simulation. The accuracy of generated mesh model is good, with the relative error of geometric parameters being less than 4%. The distance from the point cloud to the mesh model is approximately 5 mm.

Laser scanning (LS) is an effective technology for accurately capturing point clouds of visible surfaces of objects in 3D scenes. The point clouds were subsequently used for various applications, for example, generating 2D drawings of the floor or building information models (BIM) and structural inspection. However, in practice, the products from point cloud are created mainly by using commercial software, in which the quality primarily depends on users’ experiences and may contain the error caused by technician carelessness. This paper proposed a new method to automatically extract the point clouds of the floor and create a 2D drawing of floor slabs. This method analyses features of the points within cells of a 2D cell grid in the xy plane to extract candidate points of the building and each floor, while the cell- and point-based region growing segmentations were employed to extract the final points of the floor and each edge of the floor, respectively. The proposed method was successfully tested on 7.5 million points of a concrete, two-story building with 17 m long x 7m width x 7m height.