No abstract

No abstract

No abstract

No abstract

-This paper presents a fast and robust threedimensional (3D) terrain surface reconstruction system using a stereo camera. The local featurebased 3D terrain surface reconstruction algorithm consists of two major steps: matching and terrain surface reconstruction. In this paper, extracted corners are described by multi-scale descriptors (MSDs), and the matching precision is increased by a quadratic interpolation method. In the terrain surface reconstruction step, the construction terrain surface is modeled using a 3D regular grid plane for its computational efficiency and robustness. The precision of the 3D regular grid plane is improved by inferring the 3D grid vertices from a robustly matched nearest point cloud. The computational complexity of the proposed terrain surface is simplified using 2D and 3D triangle mesh structures.In order to evaluate the proposed method, different types of 3D reconstruction methods were compared under the conditions of a practical construction site. The results of the experiment show that the MSD-based 3D terrain surface algorithm performed well in terms of robustness and computational efficiency.