Abstract: In order to solve the problems of large noise and defective fracture surface in the 3D data model of Terracotta blocks, a fracture surface matching method based on geometric features of Terracotta was proposed. Firstly, the fracture surfaces were segmented and recognized from the outer surface of Terracotta blocks; Secondly, the concave and convex feature regions of the fracture surface were extracted, and the coarse matching of blocks was completed according to the similarity of feature regions; Lastly, the geometry features including curvature, surface normals and point cloud density were calculated, and the fine matching of blocks was completed using the improved iterative closest point (ICP) based on the geometry features, thus the Terracotta blocks were matching accurately. A great number of Terracotta blocks were used to validate the proposed matching method. The results show that the proposed matching method can complete the Terracotta blocks matching. Especially the proposed fine matching algorithm, compared with ICP and MICP, the matching accuracy can increase by about 40% and 10% respectively, and the time consuming can decrease by 50% and 25% respectively. So the matching method based on geometric features is an effective Terracotta blocks matching method.