Abstract: A Terracotta Warrior blocks matching algorithm based on contour curve and feature region was proposed here in order to improve the matching accuracy and convergence rate. Firstly, the fracture surfaces and their contour curves were extracted, and each contour curve was divided into several parts, then the longest common subsequence algorithm was used to match the contour curves, achieving a coarse matching of the blocks. Secondly, the concavity or convexity of vertexes on fracture surface was calculated by volume integral invariant, and the fracture surface was divided into a number of concave and convex feature regions. Finally, the centroids of the feature regions were calculated and used to achieve fine matching of blocks with improved ICP algorithm. In the experiment, three types of algorithms were used to match the Terracotta Warrior blocks. The matching results show that the blocks matching algorithm based on contour curve and feature region can achieve complete matching and partial matching of Terracotta Warrior blocks much more accurately, and obtain a much higher iterative convergence rate in fine matching stage.